{"id": "49c489b00d67a57a3f147cb9877b0e2d5beac0528ca0ba0557188ef83f34424c", "sources": ["arxiv"], "title": "PACT: Preserving Anchored Cores in Task-vectors for Model Merging", "abstract": "Model merging has emerged as a training-free alternative to multi-task learning, aiming to combine multiple task-specific fine-tuned models into a single multi-task model. Most existing model merging approaches follow the Task Arithmetic paradigm, which decomposes fine-tuned weights into pre-trained parameters and task vectors, and performs merging exclusively in the task-vector space. The effectiveness of this paradigm implicitly relies on the assumption that task-specific knowledge is encoded solely within task vectors. We argue that this assumption generally does not hold due to the intrinsic task preferences of pre-trained models. Specifically, we identify \\textbf{Load-Bearing Wall (LBW) dimensions}, namely some task-critical knowledge that remains embedded in the pre-trained weights rather than being fully transferred into task vectors. We characterize LBW dimensions from both scalar-weight and subspace perspectives, thereby covering the major paradigms of existing model merging methods. Our analysis reveals that, by ignoring LBW dimensions, task-vector-based approaches fail to fully resolve task conflicts and may inadvertently damage task-specific knowledge encoded in the pre-trained model, leading to degradation. To address this issue, we propose PACT, which preserves the anchored task-specific cores (i.e., LBW dimensions) within task vectors by aligning their orthogonal complements with the subspace of the pre-trained weights. These aligned subspace components are then removed from the task vectors before applying existing model merging algorithms. Furthermore, we develop an efficient variant based on randomized SVD to improve scalability. PACT can be seamlessly integrated with existing methods. Extensive experiments across multiple benchmarks demonstrate that PACT consistently enhances mainstream model merging approaches and establishes new state-of-the-art performance.", "authors": ["Ningyuan Shi", "Zhipeng Zhou", "Hao Wang", "Chunyan Miao", "Peilin Zhao"], "categories": ["cs.LG"], "fields_of_study": [], "published_date": "2026-06-17", "url": "https://arxiv.org/abs/2606.18627", "pdf_url": "https://arxiv.org/pdf/2606.18627v1", "arxiv_id": "2606.18627", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "395f544d427cb59b539ac8d673e2264256ff312294d4e3e6ad6012bd549c970d", "sources": ["arxiv"], "title": "Essential Subspace Merging for Multi-Task Learning", "abstract": "Model merging aims to enable multi-task learning by integrating the capabilities of multiple models fine-tuned from the same pre-trained checkpoint into a single model. Its core challenge is inter-task interference among task-specific parameter updates. In this paper, we analyze the output shifts induced by task updates and observe that their energy is concentrated in a small number of principal directions. We call the subspace spanned by these directions the essential subspace. In contrast, most remaining directions carry little task-relevant energy, but their accumulation across multiple task updates can cause severe interference during merging. Motivated by this observation, we propose Essential Subspace Decomposition (ESD), which decomposes each task update according to the principal components of its activation shift. Based on ESD, we introduce Essential Subspace Merging (ESM), a training-free static merging method that orthogonalizes and fuses essential components into one compact multi-task model. We further extend ESM to ESM++, a training-free dynamic merging method that decomposes task-specific residuals into low-rank experts and selects the most relevant expert through prototype-based routing during forward inference. Extensive experiments across multiple task sets and model scales demonstrate that ESM and ESM++ effectively preserves task knowledge while reducing inter-task interference.", "authors": ["Longhua Li", "Lei Qi", "Xin Geng", "Qi Tian"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": [], "published_date": "2026-06-17", "url": "https://arxiv.org/abs/2606.19164", "pdf_url": "https://arxiv.org/pdf/2606.19164v1", "arxiv_id": "2606.19164", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "2df46214da6c29caf600f0569b82aabac9c2df46536dadd091d8a22ac969162d", "sources": ["arxiv"], "title": "Enhancing Multilingual Reasoning via Steerable Model Merging", "abstract": "Model merging is an effective technique for composing the capabilities of a multilingual model and a reasoning model. It has achieved promising generalization in multilingual reasoning tasks by aligning feature spaces of different models. However, the merged single model often fails to address the conflicts between source models, leading to suboptimal performance. In other words, the one-size-fits-all merging strategy may not align with the characteristics of different inputs which may require prioritizing certain models over others. To this end, we propose a Steerable Model Merging (ST-Merge) framework to modulate the contribution of each source model. To realize this idea, we introduce a gated cross-attention mechanism to weight or filter the two attended source models in an adaptive manner. Extensive experiments demonstrate that ST-Merge consistently outperforms multiple strong baselines on four multilingual reasoning benchmarks across 21 different languages.", "authors": ["Zhuoran Li", "Rui Xu", "Jian Yang", "Junnan Liu", "Zhijun Chen", "Qianren Mao", "Hongcheng Guo", "Jiaheng Liu", "Likang Xiao", "Ming Li", "Xiaojie Wang"], "categories": ["cs.CL"], "fields_of_study": [], "published_date": "2026-06-17", "url": "https://arxiv.org/abs/2606.19002", "pdf_url": "https://arxiv.org/pdf/2606.19002v1", "arxiv_id": "2606.19002", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "28890da352d2d2348f1e45e05f8bb8bd71ab5fd0755926ee200db6c818f73759", "sources": ["arxiv"], "title": "Task-Restricted Symmetries in Recurrent Weight Space", "abstract": "Recurrent networks can contain substantial functional redundancy in weight space: changing a recurrent matrix may leave the input-output rollout nearly unchanged on a task distribution, while similar-scale changes can destroy the same behavior. We study this redundancy in one-layer tanh RNNs using ordered real Schur coordinates. The Schur form separates spectral blocks from directed nonnormal couplings, giving a diagnostic basis for structured ablations that keep the input and readout maps fixed. In a fixed-length copy task, selected nonnormal Schur couplings can be removed with little loss in some trained solutions, whereas other couplings are necessary for accurate autonomous replay. Across flip-flop, sine generation, and context-dependent integration, the loss-preserving ablation profile varies across tasks and trained solutions. These results identify candidate approximate functional invariances, not universal symmetries of recurrent weight space. Schur-coordinate ablations provide a practical diagnostic for which structured perturbations preserve a trained recurrent solution and which ones disrupt its computation.", "authors": ["Simon Dräger"], "categories": ["cs.LG"], "fields_of_study": [], "published_date": "2026-06-16", "url": "https://arxiv.org/abs/2606.18457", "pdf_url": "https://arxiv.org/pdf/2606.18457v1", "arxiv_id": "2606.18457", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "1b591adcb860ed122ac1b487df99fba218e8b7b048bbe8543967d98e2280a855", "sources": ["arxiv", "semantic_scholar"], "title": "Post-Hoc Merging is Not Enough: Many-Shot Model Merging with Loss-Gap Balancing", "abstract": "Model merging has become a practical post-training strategy for building a single multi-task large language model (LLM) by combining multiple task-specialized models. However, most existing approaches rely on post-hoc merging, in which task-specific models are merged only once after training. This one-shot aggregation often suffers from task interference, leading to information erasure across individual tasks. In this work, we show that replacing post-hoc merging with an iterative many-shot merging protocol is effective in improving multi-task performance. Building on this insight, we propose METIS, Mitigating Erasure from Task Interference for Stable many-shot merging. METIS is a loss-aware many-shot merging method that addresses information erasure in post-hoc merging through task-wise loss-gap weighting and consensus-based masking. Notably, METIS exhibits significant performance improvement on the worst-performing task, effectively mitigating information erasure. (Project page: https://imkyungjin.github.io/METIS/)", "authors": ["Kyungjin Im", "Miru Kim", "Chanin Eom", "Minhae Kwon"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-15", "url": "https://arxiv.org/abs/2606.16501", "pdf_url": "https://arxiv.org/pdf/2606.16501v1", "arxiv_id": "2606.16501", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "cf332f398fe5ffabea6f8897e661103c111d4ddbfcf5a148d042ce3f50248124", "sources": ["arxiv", "semantic_scholar"], "title": "From Parameters to Feature Space: Task Arithmetic for Backdoor Mitigation in Model Merging", "abstract": "Model merging (MM) has gained significant attention as a cost-effective approach to integrate multiple task-specific models into a unified model. However, recent work reveals that MM is highly susceptible to backdoor attacks. Existing defenses based on task arithmetic often fail to eliminate backdoors without substantially degrading clean-task performance, owing to their reliance on direct parameter-space editing. To address this gap, we propose Linear Feature Path Minimization (LFPM), a backdoor mitigation framework for model merging, which introduces an anti-backdoor task vector into the backdoored merged model. Unlike prior approaches, LFPM formulates the backdoor robustness of the merged model from a unified feature-space perspective under the Cross-Task Linearity (CTL) framework, which leverages the approximate linearity of features across tasks. This perspective guides the optimization of the anti-backdoor task to suppress backdoors while preserving clean-task performance. Furthermore, we introduce an effective optimization mechanism based on gradient accumulation and loss path-integral, ensuring robust backdoor suppression along the interpolation path. Extensive experiments demonstrate that LFPM consistently exhibits strong robustness against backdoor attacks in both full fine-tuning and Parameter-Efficient Fine-Tuning (PEFT) settings.", "authors": ["Zhenqian Zhu", "Yamin Hu", "Yiya Diao", "Weixiang Li", "Haodong Li", "Wenjian Luo"], "categories": ["cs.CR", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-10", "url": "https://arxiv.org/abs/2606.12498", "pdf_url": "https://arxiv.org/pdf/2606.12498v1", "arxiv_id": "2606.12498", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "bd802d69ca3ac24ad1e580769fb44be3f2888cf6f4e649521cb82710bdb8a7a5", "sources": ["arxiv", "semantic_scholar"], "title": "Closed-Form Spectral Regularization for Multi-Task Model Merging", "abstract": "Model merging combines several independently fine-tuned experts into a single multi-task model without any training data, reducing the storage, serving, and decentralized-development costs of large foundation models. State-of-the-art merging methods formulate merging as a layer-wise quadratic interference minimization problem. Although this problem admits an exact closed-form pseudoinverse solution, that solution underperforms hundreds of iterations of gradient descent in practice. The iterative loop dominates the cost of the pipeline, yet its effectiveness has remained unexplained. We revisit this regime and show that the iterative solver does not primarily act as an optimizer; rather, it serves as an implicit spectral regularizer for an ill-posed normal equation, where small-eigenvalue directions of the per-layer interference operator amplify proxy noise. Building on this finding, we formalize multi-task model merging as a noisy linear inverse problem and propose a spectral filtering estimator parameterized by a per-direction filter. We instantiate this estimator with SWUDI, a closed-form method that combines a soft exponential filter, which matches the gradient-flow trajectory of iterative descent, with a hard top-K truncation that suppresses noise-amplifying small-eigenvalue directions. Furthermore, we propose SWUDI-A, an adaptive variant that replaces the global rank hyperparameter with per-layer rank rules, further improving robustness across architectures. Both variants share a single symmetric eigendecomposition per linear layer and require no training data or optimizer state. Across four general benchmarks and a multimodal merging benchmark spanning VQA, Geometry, Chart, OCR, Grounding, and modality merging, our proposed spectral solvers match or outperform state-of-the-art merging methods. Crucially, they reduce wall-clock time by 28-72x and peak GPU memory by up to 50%.", "authors": ["Yongxian Wei", "Runxi Cheng", "Xingxuan Zhang", "Li Shen", "Chun Yuan", "Peng Cui", "Dacheng Tao"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-05", "url": "https://arxiv.org/abs/2606.07289", "pdf_url": "https://arxiv.org/pdf/2606.07289v1", "arxiv_id": "2606.07289", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "147dd55918a7551461e87f20ec491db53a9fceb48bdc3a31e2173ad7b1ae29d9", "sources": ["arxiv", "semantic_scholar"], "title": "TaDA: Calibrated Probe Gating for Task-Domain LoRA Merging", "abstract": "Combining a task LoRA adapter with a domain LoRA adapter into a single unified model is a practical yet largely unexplored challenge. Existing methods treat both adapters as symmetric peers, applying uniform weights across all layers. We argue that task and domain adapters exhibit a consistent depth-dependent asymmetry across transformer architectures. Domain dominance increases with layer depth, while shallower layers retain stronger task-relevant signals. Motivated by this observation, we propose $\\textbf{TaDA}$ ($\\textbf{Ta}$sk-$\\textbf{D}$omain LoR$\\textbf{A}$ Merging), a training-free algorithm that exploits this structure through calibrated probe-guided per-layer gating and per-component subspace-aware merging. The gating assigns individual weights per layer and projection type using a probe signal proved invariant to adapter weight magnitude. The merging discards conflicting singular directions before combining the remaining components. $\\textbf{TaDA}$ produces a standard rank-$r$ LoRA adapter with zero inference overhead. On six scientific QA benchmarks with Llama-2-7B, TaDA achieves an average accuracy of 0.452, outperforming DARE-TIES by +3.6 percentage points and obtaining the best result on all six benchmarks. On six image classification benchmarks with ViT-L/16, TaDA reaches 85.9\\% average accuracy, improving over the strongest merging baseline while leading in three of the six individual benchmarks.", "authors": ["Huy Quoc To", "Fuyi Li", "Guangyan Huang", "Ming Liu"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-03", "url": "https://arxiv.org/abs/2606.05016", "pdf_url": "https://arxiv.org/pdf/2606.05016v1", "arxiv_id": "2606.05016", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "4e49f8d87029823b017e56d722de2f7801937a2a0c1695f02e83aaed3e48dd2b", "sources": ["arxiv", "semantic_scholar"], "title": "RogueMerge: Robust and Unified Attacks against LLM Model Merging", "abstract": "Model merging composes specialized capabilities into a single LLM by aggregating task vectors sourced from unverified public platforms, exposing a critical supply-chain attack surface: Because any malicious behavior can be encoded into a task vector, and merging grants third-party vectors direct write access to model weights, an attacker-provided task vector can enable or amplify diverse downstream threats. Prior work studies only backdoor attacks against model merging for classifiers using static arithmetic heuristics, which fail to effectively handle diverse attacks on generative LLMs for three reasons. (i) LLMs rely on autoregressive decoding, where the minor parameter drift introduced by merging compounds across tokens and rapidly degrades the attack. (ii) Attackers have no knowledge of the victim's merging configurations, causing a static attack vector optimized in isolation to be easily diluted or destroyed. (iii) Practical threat induction must generalize to attack prompts unseen during optimization, which static vectors cannot adequately encode. We present RogueMerge, the first principled, unified framework that addresses all three challenges. To handle autoregressive generation, we replace static arithmetic with a joint optimization that explicitly enforces attack success after merging. To handle unknown merging settings, we formulate attack injection as a stochastic min-max problem and solve it via meta-learning-style simulation. To generalize across heterogeneous attack prompts, we employ distributionally robust optimization and derive a tractable first-order Taylor approximation at LLM scale, with a provable error bound. Across four threats, six merging algorithms, and over 170 merged LLMs, RogueMerge consistently outperforms existing attacks. It also remains stable across diverse merging settings and resists standard defenses.", "authors": ["Jinghuai Zhang", "Yetian He", "Kunlin Cai", "Han Zhao", "Fnu Suya", "Yuan Tian"], "categories": ["cs.CR", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-02", "url": "https://arxiv.org/abs/2606.03344", "pdf_url": "https://arxiv.org/pdf/2606.03344v1", "arxiv_id": "2606.03344", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "456376c706209431c598f624319e233b0fd495ede4262d4232c71457c3843746", "sources": ["arxiv", "semantic_scholar"], "title": "Decentralized Instruction Tuning: Conflict-Aware Splitting and Weight Merging", "abstract": "Instruction tuning aligns large language models, including multimodal ones, with diverse user intents, but scaling to heterogeneous mixtures is hindered by gradient interference and bandwidth-heavy synchronization. We ask whether these two bottlenecks can be addressed jointly by training parts of the mixture independently and reconciling them once in parameter space. We develop a local quadratic theory inside a shared flat basin that yields three results: weight merging produces a curvature-weighted variance reduction; PCA-aligned conflict splitting maximizes this gain along high-curvature directions; and merging additionally acts as spectral filtering with implicit norm regularization. These results directly motivate MERIT, a decentralized merge-ready instruction-tuning pipeline that estimates dataset-level gradient conflicts, partitions the mixture along the top PCA conflict axes, fine-tunes each partition independently with no inter-partition communication, and merges once via token-weighted averaging. On Qwen2.5-VL-3B with 136 Vision-FLAN tasks, MERIT improves the 8-benchmark average from 54.3 (joint training) to 57.0. The same recipe scales to a 7B model on a 1.6M-example, 176-source mixture -- matching or exceeding centralized joint training with minimal cost overhead -- and transfers to text-only FLAN. Our code is available at https://github.com/naver-ai/merit.", "authors": ["Minsik Choi", "Geewook Kim"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-01", "url": "https://arxiv.org/abs/2606.01717", "pdf_url": "https://arxiv.org/pdf/2606.01717v1", "arxiv_id": "2606.01717", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/naver-ai/merit", "venue": null, "quality_score": 0.65}
{"id": "5078b7222424a78bbcb406dd8820db45bbaf87f409cd09627c0404c5a277f551", "sources": ["arxiv", "semantic_scholar"], "title": "ThinkSwitch: Context Distillation with LoRA and Weight Interpolation for Specific-Purpose Reasoning Tasks", "abstract": "Large language models often improve on difficult tasks by spending inference-time compute on a reasoning trace before producing the final answer. That extra computation can be useful, but it also raises latency, token cost, and deployment complexity. We introduce \\textbf{ThinkSwitch}, a low-compute procedure for co-training paired instruct and thinking checkpoints. Starting from compatible Qwen3-4B instruct and thinking models, each iteration asks the thinking checkpoint to generate answers, removes the reasoning trace, distills the answer-only pairs into the instruct checkpoint with QLoRA, and reconstructs a thinking checkpoint with spherical weight interpolation. The only human-supplied inputs are task prompts; the labels are generated by the model itself. On a 30-question AIME 2026 evaluation, ThinkSwitch improves the instruct checkpoint from 10/30 to 20/30 and the thinking checkpoint from 14/30 to 22/30. On a 30-question PubMedQA subset, it improves the instruct checkpoint from 13/30 to 18/30 and the thinking checkpoint from 18/30 to 25/30. The complete experiment uses 15 training prompts per domain and costs \\$2.86 on a single cloud RTX 3070. The results are small-scale, but they indicate that targeted distillation loops can move part of the benefit of explicit reasoning into weights while preserving a separate thinking mode.", "authors": ["Dhruv Saini", "Rohan Pandey"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-31", "url": "https://arxiv.org/abs/2606.01080", "pdf_url": "https://arxiv.org/pdf/2606.01080v1", "arxiv_id": "2606.01080", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "aee8014d45f7f9672028b1ebd4be4ef199a587f45b75e86b1b226c6390100878", "sources": ["arxiv", "semantic_scholar"], "title": "Saliency-Aware Model Merging", "abstract": "Model merging aims to consolidate multiple task-specific models fine-tuned on different datasets into a unified architecture that performs cross-domain proficiency. Current data-free model merging methods often struggle to scale as they rely on simple parameter-level heuristics that ignore inter-layer dependencies and non-uniform distribution of expertise. This work proposes SA-Merging, which is built upon connectivity-based saliency formulations from structural pruning (e.g., SynFlow) and extends them to the data-free model merging setting. We define a saliency score over task vectors relative to a shared base model, and further introduce merge-aware modulation that incorporates agreement across experts to mitigate task interference. Based on this formulation, an iterative saliency-aware merging procedure progressively removes non-informative updates while preserving end-to-end connectivity. Furthermore, we extend SA-Merging to introduce rank-wise saliency decomposition for LoRAs without compromising their structural integrity. Extensive experiments on vision and language tasks demonstrate the effectiveness of our saliency-based approach, further reducing the gap between data-free and test-time adaptation methods.", "authors": ["Jungin Park", "Jiyoung Lee", "Kwanghoon Sohn"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-30", "url": "https://arxiv.org/abs/2606.00511", "pdf_url": "https://arxiv.org/pdf/2606.00511v1", "arxiv_id": "2606.00511", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "412e8b6e48974ff122c31dc2b22eef04d34eb995c98ab6d0a5cbc6399dcc2ec5", "sources": ["arxiv", "semantic_scholar"], "title": "Access Sets Matter: Budgeting Expert Reads for Scalable Weight-Space Model Merging", "abstract": "Weight-space model merging is usually formulated as an algebraic operation on checkpoints, yet at LLM scale the limiting resource is often the set of expert weights that must be read. We introduce MergePipe, a budget-aware execution layer that casts LLM merging as an \\emph{expert access-set} problem: given a merge operator and a checkpoint family in a shared weight coordinate system, choose which expert delta blocks to access under an explicit I/O budget. MergePipe indexes parameter blocks, builds deterministic access plans, and executes the induced budgeted merge with replayable manifests. The plan is budget-sound by construction and recovers the full-read merge at full budget; for fixed-coefficient additive operators, the omitted-update error is bounded by the norm of omitted deltas. Across Qwen and Llama merging workloads, MergePipe reduces expert-read I/O by up to an order of magnitude and achieves up to $11\\times$ speedups. Representative budget sweeps show $O(10^{-3})$ parameter deviation from full-read merges and no monotonic degradation on downstream benchmarks.", "authors": ["Yuanyi Wang", "Yanggan Gu", "Su Lu", "Yifan Yang", "Zhaoyi Yan", "Congkai Xie", "Jianmin Wu", "Hongxia Yang"], "categories": ["cs.LG", "eess.SY"], "fields_of_study": ["Computer Science", "Engineering"], "published_date": "2026-05-28", "url": "https://arxiv.org/abs/2605.29489", "pdf_url": "https://arxiv.org/pdf/2605.29489v1", "arxiv_id": "2605.29489", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "12df446b2647a3b452f7fe785de9055821053696a351e665f0c00b57c56a720d", "sources": ["arxiv", "semantic_scholar"], "title": "Model Merging by Output-Space Projection", "abstract": "Model merging combines fine-tuned checkpoints into a single multi-task model without retraining. Existing methods - such as task arithmetic, model soups, TIES, and DARE - are computationally efficient and empirically successful, but rely on heuristic design choices and lack formal optimality guarantees. We show that merging can be formulated as a convex quadratic programme over residual updates, yielding weights that minimise a squared-output calibration objective using calibration inputs and fine-tuned model outputs, and subsuming existing methods as special cases. Our framework yields a closed-form diagnostic - the fraction of residual energy captured by a chosen basis - that predicts downstream merge quality using only the calibration set. Empirically, the QP matches or outperforms existing methods in the single-layer setting, and we characterise when the optimal basis provides significant gains over the cheaper diagonal QP. We extend to multi-layer merging via a sequential layer-wise algorithm and demonstrate consistent gains across language and vision benchmarks.", "authors": ["Bethan Evans", "Benjamin Etheridge", "Stephen Roberts", "Jared Tanner"], "categories": ["cs.LG", "cs.IT"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2026-05-27", "url": "https://arxiv.org/abs/2605.29101", "pdf_url": "https://arxiv.org/pdf/2605.29101v1", "arxiv_id": "2605.29101", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "67b9cba3834464d319f7ef9b44251449cc6198e7fb911bab4dc7d715ce1a9991", "sources": ["arxiv", "semantic_scholar"], "title": "What-If World: A Causal Benchmark for General World Models in Embodied Scenarios", "abstract": "Video generation models are increasingly used as world simulators for tasks like driving and robotic manipulation. What matters in these settings is not whether a single video looks right, but whether the model's output changes when its input changes. We test this by giving a model two prompts describing the same scene with one physical detail varied, and checking whether the two videos diverge the way physics predicts. The wording difference between the prompts is small by design, since only one variable is changed, but the correct physical difference is not. A model that misses this can still produce two videos that each look plausible individually, and existing benchmarks score videos one at a time and cannot detect this failure. We introduce What-If World, 319 such prompt pairs built on real frames from nuScenes and DROID, organized by a taxonomy of six physical variables shared across driving and manipulation. Each pair is scored with APEO, a four-part rubric checking whether each video follows its prompt (Adherence), is physically consistent (Physics), preserves the shared scene (Environment), and ends in the correct difference (Outcome). Across nine state-of-the-art models, no system exceeds 52% on the paired score, and open-source models cluster near 28%. Every model tested fails on a large fraction of causal interventions, indicating substantial room before these models can reliably support action-conditioned simulation or model-based planning. Where models do score well, performance appears to track the visual prominence of the intervention rather than the tractability of its underlying physics. Some visually subtle interventions score as low as 14.2%, while visually pronounced ones reach 40.4%.", "authors": ["Kunlin Cai", "Rui Song", "Jinghuai Zhang", "Kaiyuan Zhang", "Pranav Bodapati", "Alicia Yu", "Fnu Suya", "Mohammad Rostami", "Jiaqi Ma", "Yuan Tian"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-26", "url": "https://arxiv.org/abs/2605.27589", "pdf_url": "https://arxiv.org/pdf/2605.27589v1", "arxiv_id": "2605.27589", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": null, "quality_score": 0.65}
{"id": "a54dc89832d4feea79b4c2458de043b65a0686836f484fe0362be31cd84b4e92", "sources": ["arxiv", "semantic_scholar"], "title": "Model Merging on Loss Landscape: A Geometry Perspective", "abstract": "Model merging offers a promising avenue for knowledge integration and parallel development without retraining. Yet, existing methods either ignore the geometry of the loss landscape or rely on intractable full-space Hessian approximations. We propose EpiMer, a framework that casts model merging as solving the Fréchet mean on a Riemannian manifold and restricts the computation to a low-rank subspace spanned by the task vectors. With the expected Hessian as the metric, we reveal a connection between local curvature and epistemic uncertainty of the parameters. Our theoretical analysis decomposes the merging error bound into the subspace Fréchet variance and the residual energy, and provides a closed-form characterization of when curvature-aware merging provably outperforms flat-geometry methods. In addition, our framework unifies both curvature-aware methods and recent spectral methods as special cases of the subspace Fréchet mean with different geometric metrics. Merging fine-tuned CLIP-ViT models on eight image classification tasks, Epistemic Merging strictly outperforms the baselines on all three CLIP-ViT backbones at matched rank, improving the across-task average accuracy and worst-task accuracy on every backbone.", "authors": ["Juanwu Lu", "Anand Bhaskar", "Brian Axelrod", "Ekaterina Tolstaya", "Tristan Emrich"], "categories": ["cs.LG", "cs.AI", "stat.ML"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2026-05-26", "url": "https://arxiv.org/abs/2605.26693", "pdf_url": "https://arxiv.org/pdf/2605.26693v1", "arxiv_id": "2605.26693", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "2620bf3007c1fffa0d289e418af177aba55134e727d5a7148ef8f40fbd958dc4", "sources": ["arxiv", "semantic_scholar"], "title": "On the Limits of Model Merging for Multilinguality in Pre-Training", "abstract": "Endowing models with consistent multilingual performance can be achieved by mixing pre-training data, or post-training approaches such as language-specific model merging. In this work, we test whether merging can be applied to monolingually pre-trained models. We conduct a controlled study on the efficacy of mixed, merged, and monolingual pre-training setups. We find that while monolingual pre-training results in strong in-language performance, merging any combination of monolingual models leads to performance collapse due to interference. Our analysis suggests representational similarity is a prerequisite for model merging. We therefore conclude that the flexibility of merging in fine-tuning does not extend trivially to language-specific pre-training.", "authors": ["Seth Aycock", "Fedor Vitiugin", "Aleksandr Umnov", "Christof Monz", "Khalil Sima'an"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-25", "url": "https://arxiv.org/abs/2605.25846", "pdf_url": "https://arxiv.org/pdf/2605.25846v1", "arxiv_id": "2605.25846", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "05faab4447d90c658ab877b93dc73703d4326f5591fd5014731e03717ac3696c", "sources": ["arxiv", "semantic_scholar"], "title": "Causal Physics Steering in Video World Models via Concept Activation Vectors", "abstract": "Video world models learn representations of physical dynamics, but controlling their physical expectations at inference time remains an open problem. Recent interpretability work identified a Physics Emergence Zone (PEZ), a group of middle transformer layers in VideoMAE where physical plausibility is represented separately from other visual features. However, it remained unclear whether this structure could be used to directly control the model's physics reasoning. We present physics steering, a training-free method that uses the weight vector of a linear probe at a PEZ layer as a Concept Activation Vector (CAV) and injects it into hidden states during inference. This shifts the model's physical expectations without changing any model weights. On the IntPhys benchmark, this intervention reliably shifts the model's plausibility judgment in either direction, depending on the steering sign. The effect appears only when the intervention is applied within the Physics Emergence Zone, suggesting that the relevant physics representation is localized there. We further find that physics is encoded separately from motion direction, and that different intuitive physics principles occupy distinct directions within this representation space. Together, these results show that physical reasoning in VideoMAE is not only readable, but also directly steerable.", "authors": ["Nahid Alam"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-23", "url": "https://arxiv.org/abs/2605.24322", "pdf_url": "https://arxiv.org/pdf/2605.24322v1", "arxiv_id": "2605.24322", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "2d9ae35b0561209cbb693daaff9e23b48160e3d2adf8b2dc4d05cc9cb6bf4984", "sources": ["arxiv", "semantic_scholar"], "title": "GEM-4D: Geometry-Enhanced Video World Models for Robot Manipulation", "abstract": "Video world models can generate realistic futures from a single instruction, but they often fail to track the same physical points consistently across time. As a result, the generated videos appear plausible, yet lack the physical grounding required for reliable action execution, such as robot manipulation. We present GEM-4D, a geometry-grounded video world model that resolves this limitation by injecting dense 4D correspondence supervision distilled from a pretrained geometry foundation model into the video generative backbone during training. This supervision enables the model to jointly capture appearance and geometric structure while retaining a single-stream architecture with no additional inference cost. We further introduce an inverse dynamics module that converts correspondence-consistent video rollouts into executable robot trajectories, enabling direct deployment in both real-world and simulated manipulation. GEM-4D achieves state-of-the-art performance on both video prediction and geometric consistency across both simulation and realistic scenarios and improves real-world manipulation success from 61% to 81%. Additional results are available at https://gem-4d.github.io/.", "authors": ["Kaichen Zhou", "Yuzhen Chen", "Fangneng Zhan", "Hang Hua", "Grace Chen", "Xinhai Chang", "Ao Qu", "Yilun Du", "Zhuang Liu", "Paul Pu Liang", "Mengyu Wang"], "categories": ["cs.CV", "cs.RO"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-20", "url": "https://arxiv.org/abs/2605.22882", "pdf_url": "https://arxiv.org/pdf/2605.22882v3", "arxiv_id": "2605.22882", "doi": null, "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "1dc8d26a5e3de54d49e2b766ae1778c983910bfa265dcdb3482e7e1403e251fe", "sources": ["arxiv", "semantic_scholar"], "title": "Tunable MAGMAX: Preference-Aware Model Merging for Continual Learning", "abstract": "Continual learning (CL) aims to train models sequentially on multiple tasks while mitigating catastrophic forgetting of previously learned knowledge. Recent advances in large pre-trained models (LPMs) and model merging techniques, such as MAGMAX, have demonstrated effective CL performance by combining task-specific parameters. However, existing methods primarily focus on average performance across all tasks and do not adequately address how to construct models accommodating different deployment environments or varying user preferences. This paper proposes a model merging framework, termed Tunable MAGMAX, which enables preference-aware control of task-specific performance in CL. Our method introduces a preference vector that controls the number of elements selected from each task vector during model merging, allowing us to adjust the merged model performance according to their deployment needs. We further propose a method for automatically constructing appropriate preference vectors by leveraging small amounts of target environment data and datasets from model training tasks, thereby eliminating the need for manual specification. The experimental result on CL benchmark tasks demonstrates that Tunable MAGMAX effectively controls task-wise performance and successfully adapts merged models to various target environments. The proposed Tunable MAGMAX achieves superior or comparable performance to baseline methods, making it a practical solution for deploying CL models to various environments where the preferences of each task performance differ.", "authors": ["Kei Hiroshima", "Kento Uchida", "Shinichi Shirakawa"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-20", "url": "https://arxiv.org/abs/2605.20803", "pdf_url": "https://arxiv.org/pdf/2605.20803v1", "arxiv_id": "2605.20803", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "e7b8fedf7f260f82ca10f1f5cac674f55430d72fa9565399e574385de9cf7939", "sources": ["arxiv", "semantic_scholar"], "title": "Unlocking the Potential of Continual Model Merging: An ODE Perspective", "abstract": "Continual Model Merging (CMM) enables rapid customization of foundation models by sequentially incorporating task-adapted models without repeated retraining. However, existing merging rules usually update the deployed model through fixed algebraic or projection-based operations, providing limited control over how much previously accumulated knowledge should be retained relative to the incoming task model. This limitation leads to unstable retention and performance degradation in long task streams, and becomes more pronounced when tasks have heterogeneous utilities. We propose ODE-driven Merging (ODE-M), a controllable framework that formulates each continual merge as a trajectory in parameter space rather than a one-step endpoint update. Motivated by mode connectivity, ODE-M constructs a barrier-aware trajectory using a rectified time-dependent velocity field, where lightweight first-order feedback from a small calibration set suppresses loss-increasing motion while preserving progress toward the incoming model. The next merged model is then obtained by selecting an operating point along this trajectory through a utility-aware time schedule, providing an explicit mechanism for balancing retained historical knowledge and incoming task expertise. Extensive experiments on standard CMM benchmarks show that ODE-M consistently improves over strong continual merging baselines across CLIP ViT backbones, stream lengths, and heterogeneous task-utility settings.", "authors": ["Lihong Lin", "Haidong Kang"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-19", "url": "https://arxiv.org/abs/2605.19409", "pdf_url": "https://arxiv.org/pdf/2605.19409v3", "arxiv_id": "2605.19409", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "aa6c8164da3fdde4a93cb23ca38cc6dc81f2015e9e31df2f32fcbd76fb0ae033", "sources": ["arxiv", "semantic_scholar"], "title": "Distilling Linearized Behavior into Non-Linear Fine-Tuning for Effective Task Arithmetic", "abstract": "Task vector composition has emerged as a promising paradigm for editing pre-trained models, enabling model merging through addition and unlearning through subtraction. Fine-tuning in the tangent space of a pre-trained model (linear fine-tuning) has proven effective, as it produces task vectors that are naturally disentangled and resistant to interference. However, linearized models suffer from limited expressivity during training and incur higher computational costs at inference time, which restrict their practical applicability. In this work, we bridge the gap between linear and standard non-linear fine-tuning. We show that linearity with respect to weight perturbations, a property defined in parameter space, can be enforced through constraints in activation space during training. Concretely, we distill hidden representations from a curvature-regularized linearized teacher into a non-linear student trained via conventional fine-tuning. We find that the resulting model inherits key properties of linearized models for task arithmetic, enabling effective composition of task vectors and achieving strong performance across vision and language benchmarks without incurring any inference-time overhead.", "authors": ["Thomas Sommariva", "Francesca Morandi", "Simone Calderara", "Angelo Porrello"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-18", "url": "https://arxiv.org/abs/2605.18993", "pdf_url": "https://arxiv.org/pdf/2605.18993v2", "arxiv_id": "2605.18993", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "d17440f0bf7bba9cbaf5113028c821a5f47ce81a85eb4f2ee19a2a437b20ca3d", "sources": ["arxiv", "semantic_scholar"], "title": "Dynamic Model Merging Made Slim", "abstract": "Model merging enables the reuse of fine-tuned models without joint training or access to original data. Dynamic merging further improves flexibility by selectively activating task-relevant parameters and efficiently composing experts across multiple tasks. However, existing dynamic methods either maintain a full shared model with tiny experts or allocate excessive capacity to experts, leading to suboptimal accuracy--efficiency trade-offs. To address this, we propose DiDi-Merging, a slim dynamic merging framework that leverages differentiable rank allocation to balance shared and expert parameters. By formulating parameter budgeting as differentiable rank optimization in low-rank modules and introducing a data-free refinement step to recover task fidelity, DiDi-Merging matches prior dynamic baselines at only 1.24x the parameters of a single fine-tuned model and surpasses them at 1.4x, substantially more compact than methods requiring > 2x storage. DiDi-Merging applies across vision, language, and multimodal tasks.", "authors": ["Guodong Du", "Wanyu Lin"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-17", "url": "https://arxiv.org/abs/2605.18904", "pdf_url": "https://arxiv.org/pdf/2605.18904v1", "arxiv_id": "2605.18904", "doi": null, "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "6613298ee5f564f65b60c55ffbcb7d084fe4918cb2546935c9bc0a29ebd0569b", "sources": ["arxiv", "semantic_scholar"], "title": "E-PMQ: Expert-Guided Post-Merge Quantization with Merged-Weight Anchoring", "abstract": "Low-resource deployment constraints have made model quantization essential for deploying neural networks while preserving performance. Meanwhile, model merging has become an increasingly practical low-resource strategy for integrating multiple task- or domain-specialized experts into a single model without joint training or multi-model serving. Together, quantization and model merging enable an efficient low-resource deployment pipeline by integrating multiple experts into one low-bit model. We formulate this setting as Post-Merge Quantization (PMQ). We show that directly applying post-training quantization (PTQ) to a merged model is unreliable because two distinct deviations are coupled: the quantization deviation introduced by low-bit reconstruction and the expert-relative merging deviation inherited from model merging. To mitigate these deviations, we propose E-PMQ, an expert-guided PMQ framework that uses source expert weights to provide expert- guided output targets during layer-wise calibration, together with merged-weight anchoring to stabilize the calibration and preserve the integrated behavior of the merged model. On CLIP-ViT-B/32 eight-task merging, E-PMQ improves 4-bit GPTQ from 65.0% to 73.6% under Task Arithmetic and from 69.1% to 74.8% under TIES-Merging. On harder settings, E-PMQ improves GPTQ from 34.8% to 76.7% on 20-task CLIP-ViT-L/14 and from 78.26% to 83.34% on FLAN-T5- base GLUE. These results demonstrate that E-PMQ enables effective post-merge quantization and low-bit deployment.", "authors": ["Wenjun Wang", "Yanggan Gu", "Shuo Cai", "Yuanyi Wang", "Pengkai Wang", "Jianmin Wu", "Hongxia Yang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-16", "url": "https://arxiv.org/abs/2605.16882", "pdf_url": "https://arxiv.org/pdf/2605.16882v1", "arxiv_id": "2605.16882", "doi": null, "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "c839223e80b83e549e5d2ea221266f31441717b2990ec107efe364fb1e82d2de", "sources": ["arxiv", "semantic_scholar"], "title": "Bayesian Model Merging", "abstract": "Model merging aims to combine multiple task-specific expert models into a single model without joint retraining, offering a practical alternative to multi-task learning when data access or computational budget is limited. Existing methods, however, face two key limitations: (1) they overlook the valuable inductive bias of strong anchor models and estimate the merged weights from scratch, and (2) they rely on a shared hyperparameter setting across different modules of the network, lacking a global optimization strategy. This paper introduces Bayesian Model Merging (BMM), a plug-and-play bi-level optimization framework, where the inner level formulates the model merging as an activation-based Bayesian regression under a strong prior induced by an anchor model, yielding an efficient closed-form solution; and the outer level leverages a Bayesian optimization procedure to search module-specific hyperparameters globally based on a small validation set. Furthermore, we reveal a key alignment between activation statistics and task vectors, enabling us to derive a data-free variant of BMM that estimates the Gram matrix for regression without any auxiliary data. Across extensive benchmarks, including up to 20-task merging in vision and 5-task merging in language, BMM consistently outperforms all plug-and-play anchor baselines (e.g., TA, WUDI-Merging, and TSV). In particular, on the ViT-L/14 benchmark for 8-task merging, a single merged model reaches 95.1, closely matching the average performance of eight task-specific experts (95.8).", "authors": ["Kaiyang Li", "Shaobo Han", "Qing Su", "Shihao Ji"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-13", "url": "https://arxiv.org/abs/2605.12843", "pdf_url": "https://arxiv.org/pdf/2605.12843v1", "arxiv_id": "2605.12843", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "4930a71f466f4ebff3edc2a1306f6215235eb5ef4828386386f16ac96a8fe3cb", "sources": ["arxiv", "semantic_scholar"], "title": "FeatCal: Feature Calibration for Post-Merging Models", "abstract": "Model merging combines task experts into one model and avoids joint training, retraining, or deploying many expert models, but the merged model often still underperforms task experts. We study this performance gap through feature drift, the difference between features produced by the merged model and by the expert on the same input. Our theory decomposes this drift into upstream propagation and local mismatch, tracks how it propagates and combines through later layers in forward order, and links final feature drift to output drift. This view motivates FeatCal, which uses a small calibration set to calibrate the merged model weights layer by layer in forward order, reducing feature drift while staying close to merged weights and preserving the benefits of model merging. FeatCal uses an efficient closed-form solution to update model weights, with no gradient descent, iterative optimization, or extra modules. On the main CLIP and GLUE benchmarks, FeatCal beats Surgery and ProbSurgery, the closest post-merging calibration baselines: 85.5% vs. 77.0%/78.8% on CLIP-ViT-B/32 Task Arithmetic (TA) and 85.2% vs. 83.7%/82.2% on FLAN-T5-base GLUE. On CLIP-ViT-B/32, 8 examples per task reach 82.9%, and 256 examples per task take 53 seconds, about 4x faster than both baselines, showing better sample efficiency and lower calibration cost.", "authors": ["Yanggan Gu", "Shuo Cai", "Zihao Wang", "Wenjun Wang", "Yuanyi Wang", "Pengkai Wang", "Sirui Huang", "Su Lu", "Jianmin Wu", "Hongxia Yang"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-13", "url": "https://arxiv.org/abs/2605.13030", "pdf_url": "https://arxiv.org/pdf/2605.13030v1", "arxiv_id": "2605.13030", "doi": null, "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "4d47028b83a40d304b240a66abfd251dc54660ac5f485f8f02dcbda8592f01b8", "sources": ["arxiv", "semantic_scholar"], "title": "EHR-RAGp: Retrieval-Augmented Prototype-Guided Foundation Model for Electronic Health Records", "abstract": "Electronic Health Records (EHR) contain rich longitudinal patient information and are widely used in predictive modeling applications. However, effectively leveraging historical data remains challenging due to long trajectories, heterogeneous events, temporal irregularity, and the varying relevance of past clinical context. Existing approaches often rely on fixed windows or uniform aggregation, which can obscure clinically important signals. In this work, we introduce EHR-RAGp, a retrieval-augmented foundation model that dynamically integrates the most relevant patient history across diverse clinical event types. We propose a prototype-guided retrieval module that acts as an alignment mechanism and estimates the relevance of retrieved historical chunks with respect to a given prediction task, guiding the model towards the most informative context. Across multiple clinical prediction tasks, EHR-RAGp consistently outperforms state-of-the-art EHR foundation models and transformer-based baselines. Furthermore, integrating EHR-RAGp with existing clinical foundation models yields substantial performance gains. Overall, EHR-RAGp provides a scalable and efficient framework for leveraging long-range clinical context to improve downstream performance.", "authors": ["Saeed Shurrab", "Mariam Al-Omari", "Dana El Samad", "Farah E. Shamout"], "categories": ["cs.IR", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-12", "url": "https://arxiv.org/abs/2605.12335", "pdf_url": "https://arxiv.org/pdf/2605.12335v1", "arxiv_id": "2605.12335", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "49e1be355fac19775b08c9c2a98481aab55608d95b3781251826db4ae4199ab4", "sources": ["arxiv", "semantic_scholar"], "title": "Revitalizing the Beginning: Avoiding Storage Dependency for Model Merging in Continual Learning", "abstract": "Model merging provides a compelling paradigm for integrating specialized expertise into a unified multi-task model, a goal that aligns naturally with the sequential knowledge acquisition in continual learning (CL). However, the requirement for preserving diverse forms of previous knowledge conflicts with the storage limitations inherent to CL. In this paper, we systematically analyze existing model merging methods under the constraints of CL. We find that current methods prioritize global alignment, which often leads to the accumulation and amplification of task-specific errors within the continuous data stream; and the vanishing gradients at the onset of subsequent tasks frequently cause optimization to stagnate. These leave the merged model in a suboptimal state at the beginning of the next training phase. To address these challenges, we propose Trajectory Regularized Merging (TRM), a framework that reformulates the merging phase as an optimization process within an augmented trajectory subspace. Our framework integrates three synergistic objectives including task alignment, prediction consistency, and gradient responsiveness to concurrently preserve merged model's historical stability and re-activate optimization dynamics. Extensive experimental results demonstrate that our method achieves state-of-the-art performance across multiple benchmarks.", "authors": ["Xi Wang", "Cheng Deng"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-08", "url": "https://arxiv.org/abs/2605.08311", "pdf_url": "https://arxiv.org/pdf/2605.08311v1", "arxiv_id": "2605.08311", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "c4c7e408fe306af6d068f85164a09e6c07113b5c3bf2991a68803b3725a67dd7", "sources": ["arxiv", "semantic_scholar"], "title": "A Tutorial for Evaluating Cure Model Appropriateness", "abstract": "In survival analysis, traditional models assume all individuals will eventually experience the event of interest. However, advances in therapeutics have led to multiple clinical contexts with potentially curative therapies, and in these contexts, certain individuals may never experience the event. Statisticians have developed cure models as a methodology to address this challenge. Nonetheless, despite significant statistical advances in cure models, we have seen more limited uptake in biomedical applications, and we hypothesize that this is caused by limited guidance in the appropriate application of cure models. Cure models require specific identifiability conditions for valid parameter estimation, and previous reports have demonstrated significant issues with the inappropriate application of cure models. Existing tutorials for cure models focus on model implementation and either assume or provide only limited guidance on whether cure modeling is appropriate for the given dataset. This tutorial addresses this gap by describing a systematic procedure that integrates clinical judgment, visual inspection of Kaplan-Meier curves, and quantitative evaluation. We provide a worked example using data from a randomized clinical trial in acute myeloid leukemia, and we also summarize findings from a series of other datasets of hematopoietic cell transplantation to suggest broad practical guidance for choosing to apply cure models. By systematically evaluating cure model appropriateness before fitting these models, researchers can achieve more reliable survival analysis and improved clinical decision-making.", "authors": ["A Tutorial for Evaluating Cure Model Appropriateness Geethanjalee Mudunkotuwa", "Durbadal Ghosh", "Subodh Selukar"], "categories": ["stat.ME"], "fields_of_study": ["Mathematics"], "published_date": "2026-05-06", "url": "https://arxiv.org/abs/2605.04999", "pdf_url": "https://arxiv.org/pdf/2605.04999v2", "arxiv_id": "2605.04999", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "b0888e2f3515294cf1c12a315703628619062902b93a6bf304294b87f36a687b", "sources": ["arxiv", "semantic_scholar"], "title": "Model Merging: Foundations and Algorithms", "abstract": "Modern deep learning usually treats models as separate artifacts: trained independently, specialized for particular purposes, and replaced when improved versions appear. This thesis studies model merging as an alternative paradigm: combining independently trained neural networks directly in weight space, with little or no optimization and without requiring access to the original training data. The thesis considers two main regimes. In the single-task setting, where models share an objective but differ in initialization, we introduce C$^2$M$^3$, a cycle-consistent merging algorithm based on Frank-Wolfe optimization. C$^2$M$^3$ aligns multiple networks into a shared, reference-free parameter space, making weight averaging meaningful without privileging any individual model. In the multi-task setting, where models are fine-tuned for different downstream tasks from a common pretrained initialization, we first develop a theoretical account of task vectors as approximate gradients. This explains both the effectiveness and the limitations of task arithmetic. Building on this view, we show that task vectors inherit the low-rank structure of gradients and introduce Task Singular Vectors (TSV), a decomposition that enables compression and interference reduction through TSV-Merge. We then present MASS, an input-adaptive routing method that uses TSV geometry to select task-relevant subspaces at inference time. Finally, we introduce MERGE$^3$, an evolutionary merging framework that uses Item Response Theory to reduce evaluation costs by up to 50$\\times$ while preserving solution quality. Together, these contributions provide theoretical and algorithmic foundations for model merging, supporting a paradigm in which learned capabilities can be composed, reused, and extended across models.", "authors": ["Donato Crisostomi"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-02", "url": "https://arxiv.org/abs/2605.01580", "pdf_url": "https://arxiv.org/pdf/2605.01580v1", "arxiv_id": "2605.01580", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "39cc3cc4d4dcdcbf8a4b9a2e71005daba831fe658bb604e0a881471aa03607b0", "sources": ["arxiv", "semantic_scholar"], "title": "Auto-FlexSwitch: Efficient Dynamic Model Merging via Learnable Task Vector Compression", "abstract": "Model merging has attracted attention as an effective path toward multi-task adaptation by integrating knowledge from multiple task-specific models. Among existing approaches, dynamic merging mitigates performance degradation caused by conflicting parameter updates across tasks by flexibly combining task-specific parameters at inference time, thereby maintaining high performance. However, these methods require storing independent parameters for each task, resulting in prohibitive storage overhead. To address this issue, we first experimentally demonstrate that the fine-tuned weight increments (referred to as task vectors) exhibit an impulse-like activation pattern and high robustness to low-bit representations. Driven by this insight, we propose T-Switch, which decomposes task vectors into three compact components: a binary sparse mask, a sign vector, and a scalar scaling factor, achieving high-fidelity approximation at high compression ratios. We then introduce Auto-Switch, a training-free merging scheme that automatically composes task vectors via feature similarity retrieval. Building on this, we develop Auto-Switch, a training-free merging scheme that automatically assembles task vectors through feature similarity retrieval. Furthermore, to transform task vector sparsification and quantization from static rules to adaptive learning, we propose FlexSwitch, a learnable framework which jointly optimizes the compression strategy for each model unit via Learnable Gating Sparsification (LGS) and Bit-width Adaptive Selection (BAS), while employing the Sparsity-Aware Storage Strategy (SASS) to select the optimal storage encoding structure. Finally, by incorporating a K-Nearest Neighbor (KNN) inference scheme with a learnable low-rank metric, we present Auto-FlexSwitch, a dynamic model merging approach that supports highly efficient task vector compression.", "authors": ["Junqi Gao", "Dazhi Zhang", "Zhichang Guo", "Biqing Qi", "Yi Ran", "Wangmeng Zuo"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-30", "url": "https://arxiv.org/abs/2604.28109", "pdf_url": "https://arxiv.org/pdf/2604.28109v1", "arxiv_id": "2604.28109", "doi": "10.48550/arXiv.2604.28109", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.55}
{"id": "ad97dee1a4e25cf7bd714d775cb31ba4027eec524dd379dbade962f7e3bad9ac", "sources": ["arxiv", "semantic_scholar"], "title": "Embedding Arithmetic: A Lightweight, Tuning-Free Framework for Post-hoc Bias Mitigation in Text-to-Image Models", "abstract": "Modern text-to-image (T2I) models amplify harmful societal biases, challenging their ethical deployment. We introduce an inference-time method that reliably mitigates social bias while keeping prompt semantics and visual context (background, layout, and style) intact. This ensures context persistency and provides a controllable parameter to adjust mitigation strength, giving practitioners fine-grained control over fairness-coherence trade-offs. Using Embedding Arithmetic, we analyze how bias is structured in the embedding space and correct it without altering model weights, prompts, or datasets. Experiments on FLUX 1.0-Dev and Stable Diffusion 3.5-Large show that the conditional embedding space forms a complex, entangled manifold rather than a grid of disentangled concepts. To rigorously assess semantic preservation beyond the circularity and bias limitations of of CLIP scores, we propose the Concept Coherence Score (CCS). Evaluated against this robust metric, our lightweight, tuning-free method significantly outperforms existing baselines in improving diversity while maintaining high concept coherence, effectively resolving the critical fairness-coherence trade-off. By characterizing how models represent social concepts, we establish geometric understanding of latent space as a principled path toward more transparent, controllable, and fair image generation.", "authors": ["Venkatesh Thirugnana Sambandham", "Torsten Schön"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-20", "url": "https://arxiv.org/abs/2604.18167", "pdf_url": "https://arxiv.org/pdf/2604.18167v1", "arxiv_id": "2604.18167", "doi": "10.48550/arXiv.2604.18167", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/cvims/EMBEDDING-ARITHMETIC}", "venue": "arXiv.org", "quality_score": 0.85}
{"id": "983090ef75a7f9605b4908be64c04f9e7cd9d79d0f98c584f65806a7d147ad40", "sources": ["arxiv", "semantic_scholar"], "title": "Understanding and Enforcing Weight Disentanglement in Task Arithmetic", "abstract": "Task arithmetic provides an efficient, training-free way to edit pre-trained models, yet lacks a fundamental theoretical explanation for its success. The existing concept of ``weight disentanglement\" describes the ideal outcome of non-interfering task composition but does not reveal its underlying cause. Crucially, what intrinsic properties of the pre-trained model ($θ_0$) or the task vectors ($τ_t$) enable this disentanglement remains underexplored. In this paper, we introduce Task-Feature Specialization (TFS), a model's ability to allocate distinct internal features to different tasks, as the fundamental principle. We first prove that TFS is a sufficient condition for weight disentanglement. More importantly, we find that TFS also gives rise to an observable geometric consequence: weight vector orthogonality. This positions TFS as the common cause for both the desired functional outcome (disentanglement) and a measurable geometric property (orthogonality). This relationship provides the key insight for our method: since the abstract TFS property is intractable to enforce directly, we can instead promote weight disentanglement by shaping its concrete geometric consequence, orthogonality. Therefore, we propose OrthoReg, a simple and effective regularization method that actively enforces an internal orthogonal structure on weight updates ($ΔW$) that constitute $τ_t$ during fine-tuning. And we theoretically prove that OrthoReg promotes disentanglement. Extensive experiments demonstrate that OrthoReg consistently and significantly enhances the performance of various task arithmetic methods. Code is available at \\href{https://github.com/RL-MIND/OrthoReg}{https://github.com/RL-MIND/OrthoReg}.", "authors": ["Shangge Liu", "Yuehan Yin", "Lei Wang", "Qi Fan", "Yinghuan Shi", "Wenbin Li", "Yang Gao", "Dacheng Tao"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-18", "url": "https://arxiv.org/abs/2604.17078", "pdf_url": "https://arxiv.org/pdf/2604.17078v1", "arxiv_id": "2604.17078", "doi": "10.48550/arXiv.2604.17078", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/RL-MIND/OrthoReg}{https://github.com/RL-MIND/OrthoReg}", "venue": "arXiv.org", "quality_score": 0.8465}
{"id": "810634cf4ae51a3ac5284f6fe396bc40a5580303ae4037458ae13dbcb06af677", "sources": ["arxiv", "semantic_scholar"], "title": "Task Alignment: A simple and effective proxy for model merging in computer vision", "abstract": "Efficiently merging several models fine-tuned for different tasks, but stemming from the same pretrained base model, is of great practical interest. Despite extensive prior work, most evaluations of model merging in computer vision are restricted to image classification using CLIP, where different classification datasets define different tasks. In this work, our goal is to make model merging more practical and show its relevance on challenging scenarios beyond this specific setting. In most vision scenarios, different tasks rely on trainable and usually heterogeneous decoders. Differently from previous studies with frozen decoders, where merged models can be evaluated right away, the non-trivial cost of decoder training renders hyperparameter selection based on downstream performance impractical. To address this, we introduce the task alignment proxy, and show how it can be used to speed up hyperparameter selection by orders of magnitude while retaining performance. Equipped with the task alignment proxy, we extend the applicability of model merging to multi-task vision models beyond CLIP-based classification.", "authors": ["Pau de Jorge", "César Roberto de Souza", "Björn Michele", "Mert Bülent Sarıyıldız", "Philippe Weinzaepfel", "Florent Perronnin", "Diane Larlus", "Yannis Kalantidis"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-14", "url": "https://arxiv.org/abs/2604.12935", "pdf_url": "https://arxiv.org/pdf/2604.12935v1", "arxiv_id": "2604.12935", "doi": "10.48550/arXiv.2604.12935", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5431}
{"id": "92d17abab81ff5ac8976c3d7e6d482b91e5f8fff55e74a3f803824cba484ff75", "sources": ["arxiv", "semantic_scholar"], "title": "One Model to Translate Them All? A Journey to Mount Doom for Multilingual Model Merging", "abstract": "Weight-space model merging combines independently fine-tuned models without accessing original training data, offering a practical alternative to joint training. While merging succeeds in multitask settings, its behavior in multilingual contexts remains poorly understood. We systematically study weight-space merging for multilingual machine translation by fully fine-tuning language model on large-scale bilingual corpora and evaluating standard merging strategies. Our experiments reveal that merging degrades performance, especially when target languages differ. To explain this failure, we analyze internal representations using span-conditioned neuron selectivity and layer-wise centered kernel alignment. We find that language-specific neurons concentrate in embedding layers and upper transformer blocks, while intermediate layers remain largely shared across languages. Critically, fine-tuning redistributes rather than sharpens language selectivity: neurons for supervised and related languages become less exclusive, while those for unsupervised languages grow more isolated. This redistribution increases representational divergence in higher layers that govern generation. These findings suggest that multilingual fine-tuning may reshape geometry in ways that reduce compatibility with standard weight-space merging assumptions. Our work thus provides an explanation for why merging fails in multilingual translation scenarios.", "authors": ["Baban Gain", "Asif Ekbal", "Trilok Nath Singh"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-03", "url": "https://arxiv.org/abs/2604.02881", "pdf_url": "https://arxiv.org/pdf/2604.02881v1", "arxiv_id": "2604.02881", "doi": "10.48550/arXiv.2604.02881", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5305}
{"id": "e78f568ec4f4de1e66aacc1a647a9724d46e1db0e6ecd4de2a0c54c0df7018bf", "sources": ["arxiv", "semantic_scholar"], "title": "Countering Catastrophic Forgetting of Large Language Models for Better Instruction Following via Weight-Space Model Merging", "abstract": "Large language models have been adopted in the medical domain for clinical documentation to reduce clinician burden. However, studies have reported that LLMs often \"forget\" a significant amount of instruction-following ability when fine-tuned using a task-specific medical dataset, a critical challenge in adopting general-purpose LLMs for clinical applications. This study presents a model merging framework to efficiently adapt general-purpose LLMs to the medical domain by countering this forgetting issue. By merging a clinical foundation model (GatorTronLlama) with a general instruct model (Llama-3.1-8B-Instruct) via interpolation-based merge methods, we seek to derive a domain-adapted model with strong performance on clinical tasks while retaining instruction-following ability. Comprehensive evaluation across medical benchmarks and five clinical generation tasks (e.g., radiology and discharge summarization) shows that merged models can effectively mitigate catastrophic forgetting, preserve clinical domain expertise, and retain instruction-following ability. In addition, our model merging strategies demonstrate training efficiency, achieving performance on par with fully fine-tuned baselines under severely constrained supervision (e.g., 64-shot vs. 256-shot). Consequently, weight-space merging constitutes a highly scalable solution for adapting open-source LLMs to clinical applications, facilitating broader deployment in resource-constrained healthcare environments.", "authors": ["Mengxian Lyu", "Cheng Peng", "Ziyi Chen", "Mengyuan Zhang", "Jieting Li Lu", "Yonghui Wu"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-02", "url": "https://arxiv.org/abs/2604.01538", "pdf_url": "https://arxiv.org/pdf/2604.01538v1", "arxiv_id": "2604.01538", "doi": "10.48550/arXiv.2604.01538", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.8181}
{"id": "f65ff7e2ce2f761f8bea3dfa8b6abb18a456892fd1c93b0cb11808af831dfcfc", "sources": ["arxiv", "semantic_scholar"], "title": "Yet Even Less Is Even Better For Agentic, Reasoning, and Coding LLMs", "abstract": "Training effective software engineering agents requires large volumes of task-specific trajectories, incurring substantial data construction costs. Inspired by the \"Less-Is-More\" hypothesis in mathematical reasoning, we investigate its extension to agentic scenarios and propose an end-to-end training framework that achieves superior agentic capabilities with fewer but higher-quality training trajectories. This is achieved via STITCH (Sliding-memory Trajectory Inference and Task Chunking Heuristic), a coarse-to-fine mechanism that filters low-value noise and retains decision-critical tokens to maximize training signal quality. We conduct experiments across multiple agent frameworks (e.g., mini-SWE-agent, MSWE-agent), model scales (30B to 355B), and multilingual settings (Python, Java, and ArkTS). On SWE-bench Verified, models trained with STITCH achieve up to 63.16% relative improvement over base models. On Multi-SWE-bench (Java), MiniMax-M2.5-STITCH achieves 43.75% with our CodeArts Agent scaffold (+16.67%). On HarmonyOS (ArkTS), GLM-4.7-STITCH improves the compilation pass rate to 61.31% (+43.34%) with less than 1K training trajectories. Our results confirm that the \"Less-Is-More\" paradigm generalizes effectively to complex agentic tasks across diverse languages and model scales.", "authors": [" CodeArts Model Team", "Yang Ye", "Jingyuan Tan", "Tianyue Jiang", "Ruizhe Ye", "Qiankun He", "Jiarui Yang", "Jian Dong", "Sicong Liang", "Chongjian Yue", "Peibai Xu", "Lufan Lu", "Shiguan Pang", "Taotao Qian", "Junbao Hu", "Yuechan Hao", "Ensheng Shi", "Qi Zhang", "Yi Hao", "Na Fan", "Xin Tan", "Shuai Yao", "Zhiwei Shen", "Zongchen Li", "Yanlin Wang", "Chong Chen", "Yuchi Ma"], "categories": ["cs.SE"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-01", "url": "https://arxiv.org/abs/2604.00824", "pdf_url": "https://arxiv.org/pdf/2604.00824v3", "arxiv_id": "2604.00824", "doi": "10.48550/arXiv.2604.00824", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5282}
{"id": "5c142e2cc97364a477a5751796bb3145d048a3a33452ce2c4487484b411bf719", "sources": ["arxiv", "semantic_scholar"], "title": "Robust Language Identification for Romansh Varieties", "abstract": "The Romansh language has several regional varieties, called idioms, which sometimes have limited mutual intelligibility. Despite this linguistic diversity, there has been a lack of documented efforts to build a language identification (LID) system that can distinguish between these idioms. Since Romansh LID should also be able to recognize Rumantsch Grischun, a supra-regional variety that combines elements of several idioms, this makes for a novel and interesting classification problem. In this paper, we present a LID system for Romansh idioms based on an SVM approach. We evaluate our model on a newly curated benchmark across two domains and find that it reaches an average in-domain accuracy of 97%, enabling applications such as idiom-aware spell checking or machine translation. Our classifier is publicly available.", "authors": ["Charlotte Model", "Sina Ahmadi", "Jannis Vamvas"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-16", "url": "https://arxiv.org/abs/2603.15969", "pdf_url": "https://arxiv.org/pdf/2603.15969v2", "arxiv_id": "2603.15969", "doi": "10.48550/arXiv.2603.15969", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5099}
{"id": "e6046277e3441301722c53b9153a9b30a0c0165164610b33386bc5a8e0aafc3f", "sources": ["arxiv", "semantic_scholar"], "title": "Resolving Interference (RI): Disentangling Models for Improved Model Merging", "abstract": "Model merging has shown that multitask models can be created by directly combining the parameters of different models that are each specialized on tasks of interest. However, models trained independently on distinct tasks often exhibit interference that degrades the merged model's performance. To solve this problem, we formally define the notion of Cross-Task Interference as the drift in the representation of the merged model relative to its constituent models. Reducing cross-task interference is key to improving merging performance. To address this issue, we propose our method, Resolving Interference (RI), a light-weight adaptation framework which disentangles expert models to be functionally orthogonal to the space of other tasks, thereby reducing cross-task interference. RI does this whilst using only unlabeled auxiliary data as input (i.e., no task-data is needed), allowing it to be applied in data-scarce scenarios. RI consistently improves the performance of state-of-the-art merging methods by up to 3.8% and generalization to unseen domains by up to 2.3%. We also find RI to be robust to the source of auxiliary input while being significantly less sensitive to tuning of merging hyperparameters. Our codebase is available at: https://github.com/pramesh39/resolving_interference", "authors": ["Pratik Ramesh", "George Stoica", "Arun Iyer", "Leshem Choshen", "Judy Hoffman"], "categories": ["cs.LG", "cs.AI", "cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-13", "url": "https://arxiv.org/abs/2603.13467", "pdf_url": "https://arxiv.org/pdf/2603.13467v1", "arxiv_id": "2603.13467", "doi": "10.48550/arXiv.2603.13467", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/pramesh39/resolving_interference", "venue": "arXiv.org", "quality_score": 0.7827}
{"id": "fe57823867561511731633bcaa9cc3339d991cdde48cc5784e2b4c35349c9f1b", "sources": ["arxiv", "semantic_scholar"], "title": "Mortgage Burnout and Selection Effects in Heterogeneous Cox Hazard Models", "abstract": "We study the aggregate hazard rate of a heterogeneous population whose individual event intensities are modeled as Cox (doubly stochastic) processes. In the deterministic hazard setting, the observed pool hazard is the survival weighted mean of the individual hazards, and its time derivative equals the mean individual hazard drift minus a variance term. This yields a transparent structural explanation of burnout in mortgage pools. We extend this perspective to stochastic intensity models. The observed pool hazard remains a survival-weighted mean, but now evolves as an Ito process whose drift contains the mean drift of the individual hazards and a negative selection term driven by cross-sectional dispersion, together with a diffusion term inherited from the common factor. We formulate the general identity and discuss special cases relevant to mortgage prepayment modeling.", "authors": ["Andrew Lesniewski"], "categories": ["q-fin.MF", "econ.GN", "stat.ME"], "fields_of_study": ["Economics", "Mathematics"], "published_date": "2026-03-12", "url": "https://arxiv.org/abs/2603.12422", "pdf_url": "https://arxiv.org/pdf/2603.12422v3", "arxiv_id": "2603.12422", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.3216}
{"id": "a30c8672c6849d29fb693d80d07ae96594564ace209e679e5ce86faabace6313", "sources": ["arxiv", "semantic_scholar"], "title": "An Empirical Study and Theoretical Explanation on Task-Level Model-Merging Collapse", "abstract": "Model merging unifies independently fine-tuned LLMs from the same base, enabling reuse and integration of parallel development efforts without retraining. However, in practice we observe that merging does not always succeed: certain combinations of task-specialist models suffer from catastrophic performance degradation after merging. We refer to this failure mode as merging collapse. Intuitively, collapse arises when the learned representations or parameter adjustments for different tasks are fundamentally incompatible, so that merging forces destructive interference rather than synergy. In this paper, we identify and characterize the phenomenon of task-level merging collapse, where certain task combinations consistently trigger huge performance degradation across all merging methods. Through extensive experiments and statistical analysis, we demonstrate that representational incompatibility between tasks is strongly correlated with merging collapse, while parameter-space conflict metrics show minimal correlation, challenging conventional wisdom in model merging literature. We provide a theoretical explanation on this phenomenon through rate-distortion theory with a dimension-dependent bound, establishing fundamental limits on task mergeability regardless of methodology.", "authors": ["Yuan Cao", "Dezhi Ran", "Yuzhe Guo", "Mengzhou Wu", "Simin Chen", "Linyi Li", "Wei Yang", "Tao Xie"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-10", "url": "https://arxiv.org/abs/2603.09463", "pdf_url": "https://arxiv.org/pdf/2603.09463v1", "arxiv_id": "2603.09463", "doi": "10.48550/arXiv.2603.09463", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.503}
{"id": "37f13bce44fd265b6a320b12f9c050d116b89cc93a79a83687c452d9bb5a7b55", "sources": ["arxiv", "semantic_scholar"], "title": "Model Merging in the Era of Large Language Models: Methods, Applications, and Future Directions", "abstract": "Model merging combines the parameters of multiple neural networks into a single model without additional training. As fine-tuned large language models (LLMs) proliferate, merging offers a computationally efficient alternative to ensembles and full retraining, enabling practitioners to compose specialized capabilities at minimal cost. This survey examines model merging in the LLM era through the \\textbf{FUSE} taxonomy, organized along \\textbf{F}oundations, \\textbf{U}nification Strategies, \\textbf{S}cenarios, and \\textbf{E}cosystem. We first establish the theoretical underpinnings of merging, including loss landscape geometry and mode connectivity, then systematically review the algorithmic space spanning weight averaging, task vector arithmetic, sparsification-enhanced methods, mixture-of-experts architectures, and evolutionary optimization. We further examine downstream applications across multi-task learning, safety alignment, domain specialization, and federated learning, and survey the supporting ecosystem of tools and evaluation benchmarks. Finally, we identify key open challenges and future directions, aiming to equip researchers and practitioners with a structured foundation for advancing model merging.", "authors": ["Mingyang Song", "Mao Zheng"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-10", "url": "https://arxiv.org/abs/2603.09938", "pdf_url": "https://arxiv.org/pdf/2603.09938v2", "arxiv_id": "2603.09938", "doi": "10.48550/arXiv.2603.09938", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.503}
{"id": "db3afe44be8b623f6135a2f763b6005dca6f6d1e7fcf58df8df321f7d774096c", "sources": ["arxiv", "semantic_scholar"], "title": "DC-Merge: Improving Model Merging with Directional Consistency", "abstract": "Model merging aims to integrate multiple task-adapted models into a unified model that preserves the knowledge of each task. In this paper, we identify that the key to this knowledge retention lies in maintaining the directional consistency of singular spaces between merged multi-task vector and individual task vectors. However, this consistency is frequently compromised by two issues: i) an imbalanced energy distribution within task vectors, where a small fraction of singular values dominate the total energy, leading to the neglect of semantically important but weaker components upon merging, and ii) the geometric inconsistency of task vectors in parameter space, which causes direct merging to distort their underlying directional geometry. To address these challenges, we propose DC-Merge, a method for directional-consistent model merging. It first balances the energy distribution of each task vector by smoothing its singular values, ensuring all knowledge components are adequately represented. These energy-balanced vectors are then projected onto a shared orthogonal subspace to align their directional geometries with minimal reconstruction error. Finally, the aligned vectors are aggregated in the shared orthogonal subspace and projected back to the original parameter space. Extensive experiments on vision and vision-language benchmarks show that DC-Merge consistently achieves state-of-the-art performance in both full fine-tuning and LoRA settings. The implementation code is available at https://github.com/Tobeginwith/DC-Merge.", "authors": ["Han-Chen Zhang", "Zi-Hao Zhou", "Mao-Lin Luo", "Shimin Di", "Min-Ling Zhang", "Tong Wei"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-06", "url": "https://arxiv.org/abs/2603.06242", "pdf_url": "https://arxiv.org/pdf/2603.06242v2", "arxiv_id": "2603.06242", "doi": "10.48550/arXiv.2603.06242", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/Tobeginwith/DC-Merge", "venue": "arXiv.org", "quality_score": 0.7703}
{"id": "016ba6b7f09a96c77fbcb5f32eb427ad06f939de0673edd78cb9f8225d5dec4f", "sources": ["arxiv", "semantic_scholar"], "title": "Magic partition functions: Sign smoothing convolutions with Dirichlet invertible arithmetic functions", "abstract": "Sign changes in sums of arithmetic functions and their inverses are a subtle topic with room to grow new results. Suppose that $S_f(x) := \\sum_{n \\leq x} f(n)$ is the summatory function of some arithmetic function $f$ such that $f(1) \\neq 1$. There are known lower bounds on the limiting growth of $V(S_f, Y)$ -- the number of sign changes of $S_f(y)$ on the interval $y \\in (0, Y]$ as $Y \\rightarrow \\infty$. We observe a partition theoretic sign smoothing by discrete convolution of the local oscillatory properties of the Dirichlet inverse of $f$, $S_{f^{-1}}(x)$. These so-called invertible ``magic partition function`` encodings lead to a sequence of convolution sums which have predictable sign properties provided the sequence of $f(n)$ ($f^{-1}(n)$, respectively) has reasonable asymptotic upper bounds with respect to $n$.", "authors": ["Maxie Dion Schmidt"], "categories": ["math.NT"], "fields_of_study": ["Mathematics"], "published_date": "2026-03-06", "url": "https://arxiv.org/abs/2603.06890", "pdf_url": "https://arxiv.org/pdf/2603.06890v1", "arxiv_id": "2603.06890", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.3172}
{"id": "40b92de5d2aedd4870494e7f9ab1b1cfbbcc73e90e3db0aec6121eb9bd96a90a", "sources": ["arxiv", "semantic_scholar"], "title": "Bridging Domains through Subspace-Aware Model Merging", "abstract": "Model merging integrates multiple task-specific models into a single consolidated one. Recent research has made progress in improving merging performance for in-distribution or multi-task scenarios, but domain generalization in model merging remains underexplored. We investigate how merging models fine-tuned on distinct domains affects generalization to unseen domains. Through an analysis of parameter competition in the task matrix using singular value decomposition, we show that merging models trained under different distribution shifts induces stronger conflicts between their subspaces compared to traditional multi-task settings. To mitigate this issue, we propose SCORE (Subspace COnflict-Resolving mErging), a method designed to alleviate such singular subspace conflicts. SCORE finds a shared orthogonal basis by computing the principal components of the concatenated leading singular vectors of all models. It then projects each task matrix into the shared basis, pruning off-diagonal components to remove conflicting singular directions. SCORE consistently outperforms, on average, existing model merging approaches in domain generalization settings across a variety of architectures and model scales, demonstrating its effectiveness and scalability.", "authors": ["Levy Chaves", "Chao Zhou", "Rebekka Burkholz", "Eduardo Valle", "Sandra Avila"], "categories": ["cs.LG", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-06", "url": "https://arxiv.org/abs/2603.05768", "pdf_url": "https://arxiv.org/pdf/2603.05768v2", "arxiv_id": "2603.05768", "doi": "10.48550/arXiv.2603.05768", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4984}
{"id": "b581d341a74672acab3484e932bcbd2ece749a7a612d03a5f423f853eb112465", "sources": ["arxiv", "semantic_scholar"], "title": "Model Medicine: A Clinical Framework for Understanding, Diagnosing, and Treating AI Models", "abstract": "Model Medicine is the science of understanding, diagnosing, treating, and preventing disorders in AI models, grounded in the principle that AI models -- like biological organisms -- have internal structures, dynamic processes, heritable traits, observable symptoms, classifiable conditions, and treatable states. This paper introduces Model Medicine as a research program, bridging the gap between current AI interpretability research (anatomical observation) and the systematic clinical practice that complex AI systems increasingly require. We present five contributions: (1) a discipline taxonomy organizing 15 subdisciplines across four divisions -- Basic Model Sciences, Clinical Model Sciences, Model Public Health, and Model Architectural Medicine; (2) the Four Shell Model (v3.3), a behavioral genetics framework empirically grounded in 720 agents and 24,923 decisions from the Agora-12 program, explaining how model behavior emerges from Core--Shell interaction; (3) Neural MRI (Model Resonance Imaging), a working open-source diagnostic tool mapping five medical neuroimaging modalities to AI interpretability techniques, validated through four clinical cases demonstrating imaging, comparison, localization, and predictive capability; (4) a five-layer diagnostic framework for comprehensive model assessment; and (5) clinical model sciences including the Model Temperament Index for behavioral profiling, Model Semiology for symptom description, and M-CARE for standardized case reporting. We additionally propose the Layered Core Hypothesis -- a biologically-inspired three-layer parameter architecture -- and a therapeutic framework connecting diagnosis to treatment.", "authors": ["Jihoon Jeong"], "categories": ["cs.AI", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-05", "url": "https://arxiv.org/abs/2603.04722", "pdf_url": "https://arxiv.org/pdf/2603.04722v2", "arxiv_id": "2603.04722", "doi": "10.48550/arXiv.2603.04722", "citation_count": 2, "influential_citation_count": 2, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.7685}
{"id": "e12060224d6c67308d7f2a09f5a6038d0ad1760d870936f88650fc5c6b868993", "sources": ["arxiv", "semantic_scholar"], "title": "BD-Merging: Bias-Aware Dynamic Model Merging with Evidence-Guided Contrastive Learning", "abstract": "Model Merging (MM) has emerged as a scalable paradigm for multi-task learning (MTL), enabling multiple task-specific models to be integrated without revisiting the original training data. Despite recent progress, the reliability of MM under test-time distribution shift remains insufficiently understood. Most existing MM methods typically assume that test data are clean and distributionally aligned with both the training and auxiliary sources. However, this assumption rarely holds in practice, often resulting in biased predictions with degraded generalization. To address this issue, we present BD-Merging, a bias-aware unsupervised model merging framework that explicitly models uncertainty to achieve adaptive reliability under distribution shift. First, BD-Merging introduces a joint evidential head that learns uncertainty over a unified label space, capturing cross-task semantic dependencies in MM. Second, building upon this evidential foundation, we propose an Adjacency Discrepancy Score (ADS) that quantifies evidential alignment among neighboring samples. Third, guided by ADS, a discrepancy-aware contrastive learning mechanism refines the merged representation by aligning consistent samples and separating conflicting ones. Combined with general unsupervised learning, this process trains a debiased router that adaptively allocates task-specific or layer-specific weights on a per-sample basis, effectively mitigating the adverse effects of distribution shift. Extensive experiments across diverse tasks demonstrate that BD-Merging achieves superior effectiveness and robustness compared to state-of-the-art MM baselines.", "authors": ["Yuhan Xie", "Chen Lyu"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-04", "url": "https://arxiv.org/abs/2603.03920", "pdf_url": "https://arxiv.org/pdf/2603.03920v2", "arxiv_id": "2603.03920", "doi": "10.48550/arXiv.2603.03920", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4961}
{"id": "9140caa560fbd31b5d80aa941eba5427a59e7abe3257ee8ebb13c1f8cf9ab61c", "sources": ["arxiv", "semantic_scholar"], "title": "ACE-Merging: Data-Free Model Merging with Adaptive Covariance Estimation", "abstract": "Model merging aims to combine multiple task-specific expert models into a single model while preserving generalization across diverse tasks. However, interference among experts, especially when they are trained on different objectives, often leads to significant performance degradation. Despite recent progress, resolving this interference without data access, retraining, or architectural modification remains a fundamental challenge. This paper provides a theoretical analysis demonstrating that the input covariance of each task, which is a key factor for optimal merging, can be implicitly estimated from the parameter differences of its fine-tuned model, even in a fully data-free setting. Building on this insight, we introduce \\acem, an Adaptive Covariance Estimation framework that effectively mitigates inter-task interference. Our approach features a principled, closed-form solution that contrasts with prior iterative or heuristic methods. Extensive experiments on both vision and language benchmarks demonstrate that \\acem sets a new state-of-the-art among data-free methods. It consistently outperforms existing baselines; for example, \\acem achieves an average absolute improvement of 4\\% over the previous methods across seven tasks on GPT-2. Owing to its efficient closed-form formulation, \\acem delivers superior performance with a modest computational cost, providing a practical and theoretically grounded solution for model merging.", "authors": ["Bo Xu", "Haotian Wu", "Hehai Lin", "Weiquan Huang", "Beier Zhu", "Yao Shu", "Chengwei Qin"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-03", "url": "https://arxiv.org/abs/2603.02945", "pdf_url": "https://arxiv.org/pdf/2603.02945v2", "arxiv_id": "2603.02945", "doi": "10.48550/arXiv.2603.02945", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.495}
{"id": "c283172d605ef6c0b753cd0c6548ef754c6c1c256bdbf23b9c2531bdbd219265", "sources": ["arxiv", "semantic_scholar"], "title": "Leveraging Model Soups to Classify Intangible Cultural Heritage Images from the Mekong Delta", "abstract": "The classification of Intangible Cultural Heritage (ICH) images in the Mekong Delta poses unique challenges due to limited annotated data, high visual similarity among classes, and domain heterogeneity. In such low-resource settings, conventional deep learning models often suffer from high variance or overfit to spurious correlations, leading to poor generalization. To address these limitations, we propose a robust framework that integrates the hybrid CoAtNet architecture with model soups, a lightweight weight-space ensembling technique that averages checkpoints from a single training trajectory without increasing inference cost. CoAtNet captures both local and global patterns through stage-wise fusion of convolution and self-attention. We apply two ensembling strategies - greedy and uniform soup - to selectively combine diverse checkpoints into a final model. Beyond performance improvements, we analyze the ensembling effect through the lens of bias-variance decomposition. Our findings show that model soups reduces variance by stabilizing predictions across diverse model snapshots, while introducing minimal additional bias. Furthermore, using cross-entropy-based distance metrics and Multidimensional Scaling (MDS), we show that model soups selects geometrically diverse checkpoints, unlike Soft Voting, which blends redundant models centered in output space. Evaluated on the ICH-17 dataset (7,406 images across 17 classes), our approach achieves state-of-the-art results with 72.36% top-1 accuracy and 69.28% macro F1-score, outperforming strong baselines including ResNet-50, DenseNet-121, and ViT. These results underscore that diversity-aware checkpoint averaging provides a principled and efficient way to reduce variance and enhance generalization in culturally rich, data-scarce classification tasks.", "authors": ["Quoc-Khang Tran", "Minh-Thien Nguyen", "Nguyen-Khang Pham"], "categories": ["cs.CV", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-02", "url": "https://arxiv.org/abs/2603.02181", "pdf_url": "https://arxiv.org/pdf/2603.02181v1", "arxiv_id": "2603.02181", "doi": "10.32913/mic-ict-research.v2025.n3.1395", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.3143}
{"id": "efd5ab0e7601576782c61cce15281a78fc4b8ddee519aabb849d90cd60e549b4", "sources": ["arxiv", "semantic_scholar"], "title": "Model Merging in the Essential Subspace", "abstract": "Model merging aims to integrate multiple task-specific fine-tuned models derived from a shared pre-trained checkpoint into a single multi-task model without additional training. Despite extensive research, task interference remains a major obstacle that often undermines the performance of merged models. In this paper, we propose ESM (Essential Subspace Merging) , a robust framework for effective model merging. We begin by performing Principal Component Analysis (PCA) on feature shifts induced by parameter updates. The resulting principal directions span an essential subspace that dominantly influences feature representations. Each task's parameter update matrix is projected onto its respective essential subspace for low-rank decomposition before merging. This methodology mitigates inter-task interference while preserving core task-specific functionality. Furthermore, we introduce a multi-level polarized scaling strategy that amplifies parameters containing critical knowledge and suppresses redundant ones, preventing essential knowledge from being overwhelmed during fusion. Extensive experiments across multiple task sets and model scales demonstrate that our method achieves state-of-the-art performance in multi-task model merging.", "authors": ["Longhua Li", "Lei Qi", "Qi Tian", "Xin Geng"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-23", "url": "https://arxiv.org/abs/2602.20208", "pdf_url": "https://arxiv.org/pdf/2602.20208v1", "arxiv_id": "2602.20208", "doi": "10.48550/arXiv.2602.20208", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4858}
{"id": "5117d328859e8ea7be7aa8a22eabc7717ef438946e98d7f9f29f0ed6a7e71cc3", "sources": ["arxiv", "semantic_scholar"], "title": "Interpolation in Proof Theory", "abstract": "This chapter provides a comprehensive overview of proof-theoretic methods for establishing interpolation properties across a range of logics, including classical, intuitionistic, modal, and substructural logics. Central to the discussion are two foundational techniques: Maehara's method for Craig interpolation and Pitts' method for uniform interpolation. The chapter demonstrates how these methods lead to results on the existence of well-behaved proof systems in the contemporary framework of universal proof theory and how they provide a road map for constructing interpolation proofs using modern proof formalisms. The emphasis of the chapter is on constructive, modular, and syntax-driven techniques that illuminate deeper connections between interpolation properties and proof systems.", "authors": ["Iris van der Giessen", "Raheleh Jalali", "Roman Kuznets"], "categories": ["cs.LO"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-18", "url": "https://arxiv.org/abs/2602.16318", "pdf_url": "https://arxiv.org/pdf/2602.16318v1", "arxiv_id": "2602.16318", "doi": "10.48550/arXiv.2602.16318", "citation_count": 3, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4801}
{"id": "9cccff0d7842f751feb9afa193c1a35a1620568eef4f41ce9df3cc39517ac554", "sources": ["arxiv", "semantic_scholar"], "title": "Beyond Parameter Arithmetic: Sparse Complementary Fusion for Distribution-Aware Model Merging", "abstract": "Model merging has emerged as a promising paradigm for composing the capabilities of large language models by directly operating in weight space, enabling the integration of specialized models without costly retraining. However, existing merging methods largely rely on parameter-space heuristics, which often introduce severe interference, leading to degraded generalization and unstable generation behaviors such as repetition and incoherent outputs. In this work, we propose Sparse Complementary Fusion with reverse KL (SCF-RKL), a novel model merging framework that explicitly controls functional interference through sparse, distribution-aware updates. Instead of assuming linear additivity in parameter space, SCF-RKL measures the functional divergence between models using reverse Kullback-Leibler divergence and selectively incorporates complementary parameters. This mode-seeking, sparsity-inducing design effectively preserves stable representations while integrating new capabilities. We evaluate SCF-RKL across a wide range of model scales and architectures, covering both reasoning-focused and instruction-tuned models. Extensive experiments on 24 benchmarks spanning advanced reasoning, general reasoning and knowledge, instruction following, and safety demonstrate, vision classification that SCF-RKL consistently outperforms existing model merging methods while maintaining strong generalization and generation stability.", "authors": ["Weihong Lin", "Lin Sun", "Qilong Shi", "Aomufei Yuan", "Yuxuan Tian", "Zhengyang Wang", "Guangxiang Zhao", "Xiangzheng Zhang", "Tong Yang"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-12", "url": "https://arxiv.org/abs/2602.11717", "pdf_url": "https://arxiv.org/pdf/2602.11717v1", "arxiv_id": "2602.11717", "doi": "10.48550/arXiv.2602.11717", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4732}
{"id": "7dd364e9eb6e0b67bf0d0306f9f9b865dda40acf9b091cc718b58b19c537f8c6", "sources": ["arxiv", "semantic_scholar"], "title": "Craig Interpolation in Program Verification", "abstract": "Craig interpolation is used in program verification for automating key tasks such as the inference of loop invariants and the computation of program abstractions. This chapter covers some of the most important techniques that have been developed in this context over the last years, focusing on two aspects: the derivation of Craig interpolants modulo the theories and data types used in verification and the basic design of verification algorithms applying interpolation.", "authors": ["Philipp Rümmer"], "categories": ["cs.LO"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-09", "url": "https://arxiv.org/abs/2602.08532", "pdf_url": "https://arxiv.org/pdf/2602.08532v1", "arxiv_id": "2602.08532", "doi": "10.48550/arXiv.2602.08532", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4698}
{"id": "c21aeafc7629df476c4c2bb500d537f78feac59fd8c1430fa9a67d84ff3402cd", "sources": ["arxiv", "semantic_scholar"], "title": "M-Loss: Quantifying Model Merging Compatibility with Limited Unlabeled Data", "abstract": "Training of large-scale models is both computationally intensive and often constrained by the availability of labeled data. Model merging offers a compelling alternative by directly integrating the weights of multiple source models without requiring additional data or extensive training. However, conventional model merging techniques, such as parameter averaging, often suffer from the unintended combination of non-generalizable features, especially when source models exhibit significant weight disparities. Comparatively, model ensembling generally provides more stable and superior performance that aggregates multiple models by averaging outputs. However, it incurs higher inference costs and increased storage requirements. While previous studies experimentally showed the similarities between model merging and ensembling, theoretical evidence and evaluation metrics remain lacking. To address this gap, we introduce Merging-ensembling loss (M-Loss), a novel evaluation metric that quantifies the compatibility of merging source models using very limited unlabeled data. By measuring the discrepancy between parameter averaging and model ensembling at layer and node levels, M-Loss facilitates more effective merging strategies. Specifically, M-Loss serves both as a quantitative criterion of the theoretical feasibility of model merging, and a guide for parameter significance in model pruning. Our theoretical analysis and empirical evaluations demonstrate that incorporating M-Loss into the merging process significantly improves the alignment between merged models and model ensembling, providing a scalable and efficient framework for accurate model consolidation.", "authors": ["Tiantong Wang", "Yiyang Duan", "Haoyu Chen", "Tiantong Wu", "Wei Yang Bryan Lim"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-09", "url": "https://arxiv.org/abs/2602.08564", "pdf_url": "https://arxiv.org/pdf/2602.08564v1", "arxiv_id": "2602.08564", "doi": "10.1609/aaai.v40i31.39854", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/languangduan/mLoss", "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.726}
{"id": "0b258e7e0e3659d5a5ebd84a017f3459d1658a407b3bb58fbef072157e616fd7", "sources": ["arxiv", "semantic_scholar"], "title": "Definability and Interpolation in Philosophy", "abstract": "This paper is a historical tour of occurrences of the Craig interpolation theorem and the Beth definability theorem in philosophy since the 1950s. We identify the notion of dependence as one major red thread behind these, and include some new technical results, in particular, on logical system translations and generalized definability", "authors": ["Johan van Benthem"], "categories": ["cs.LO"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-08", "url": "https://arxiv.org/abs/2602.07907", "pdf_url": "https://arxiv.org/pdf/2602.07907v1", "arxiv_id": "2602.07907", "doi": "10.48550/arXiv.2602.07907", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4686}
{"id": "e7819db6fdbfdbee89da2a0c4e0238496b8c5b49064b31f30cea7737204e0c7c", "sources": ["arxiv", "semantic_scholar"], "title": "Fine-Grained Model Merging via Modular Expert Recombination", "abstract": "Model merging constructs versatile models by integrating task-specific models without requiring labeled data or expensive joint retraining. Although recent methods improve adaptability to heterogeneous tasks by generating customized merged models for each instance, they face two critical limitations. First, the instance-specific merged models lack reusability, restricting the exploitation of high-quality merging configurations and efficient batch inference. Second, these methods treat each task-specific model as a monolithic whole, overlooking the diverse mergeability of homologous components such as attention and multilayer perceptron layers, and the differing merging sensitivities across components. To address these limitations, we propose MERGE (\\underline{M}odular \\underline{E}xpert \\underline{R}ecombination for fine-\\underline{G}rained m\\underline{E}rging), a method that enables component-wise model merging and input-aware, on-demand module recombination at inference. MERGE formulates component-wise merging as a bi-objective optimization problem that balances cross-task performance and storage efficiency, and develops a surrogate-assisted evolutionary algorithm to efficiently identify Pareto-optimal merging configurations. These high-quality configurations underpin a reusable modular expert library, from which a lightweight routing network dynamically activates and recombines modular experts to assemble input-specific models and enable efficient inference under storage constraints. Extensive experiments across various model scales, task types, and fine-tuning strategies demonstrate that MERGE consistently outperforms strong baselines and generalizes effectively.", "authors": ["Haiyun Qiu", "Xingyu Wu", "Liang Feng", "Kay Chen Tan"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-06", "url": "https://arxiv.org/abs/2602.06552", "pdf_url": "https://arxiv.org/pdf/2602.06552v1", "arxiv_id": "2602.06552", "doi": "10.48550/arXiv.2602.06552", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4664}
{"id": "bfe6a4b4d2b0af2b24edfa14388688ea651303824707410cb8f9892310326f57", "sources": ["arxiv", "semantic_scholar"], "title": "Orthogonal Model Merging", "abstract": "Merging finetuned Large Language Models (LLMs) has become increasingly important for integrating diverse capabilities into a single unified model. However, prevailing model merging methods rely on linear arithmetic in Euclidean space, which often destroys the intrinsic geometric properties of pretrained weights, such as hyperspherical energy. To address this, we propose Orthogonal Model Merging (OrthoMerge), a method that performs merging operations on the Riemannian manifold formed by the orthogonal group to preserve the geometric structure of the model's weights. By mapping task-specific orthogonal matrices learned by Orthogonal Finetuning (OFT) to the Lie algebra, OrthoMerge enables a principled yet efficient integration that takes into account both the direction and intensity of adaptations. In addition to directly leveraging orthogonal matrices obtained by OFT, we further extend this approach to general models finetuned with non-OFT methods (i.e., low-rank finetuning, full finetuning) via an Orthogonal-Residual Decoupling strategy. This technique extracts the orthogonal components of expert models by solving the orthogonal Procrustes problem, which are then merged on the manifold of the orthogonal group, while the remaining linear residuals are processed through standard additive merging. Extensive empirical results demonstrate the effectiveness of OrthoMerge in mitigating catastrophic forgetting and maintaining model performance across diverse tasks.", "authors": ["Sihan Yang", "Kexuan Shi", "Weiyang Liu"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-05", "url": "https://arxiv.org/abs/2602.05943", "pdf_url": "https://arxiv.org/pdf/2602.05943v1", "arxiv_id": "2602.05943", "doi": "10.48550/arXiv.2602.05943", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4652}
{"id": "03cce1c5e4d2cd288c2ab05df72c16cdc45616146d85b9a95803a227ce8e4f89", "sources": ["arxiv", "semantic_scholar"], "title": "When Shared Knowledge Hurts: Spectral Over-Accumulation in Model Merging", "abstract": "Model merging combines multiple fine-tuned models into a single model by adding their weight updates, providing a lightweight alternative to retraining. Existing methods primarily target resolving conflicts between task updates, leaving the failure mode of over-counting shared knowledge unaddressed. We show that when tasks share aligned spectral directions (i.e., overlapping singular vectors), a simple linear combination repeatedly accumulates these directions, inflating the singular values and biasing the merged model toward shared subspaces. To mitigate this issue, we propose Singular Value Calibration (SVC), a training-free and data-free post-processing method that quantifies subspace overlap and rescales inflated singular values to restore a balanced spectrum. Across vision and language benchmarks, SVC consistently improves strong merging baselines and achieves state-of-the-art performance. Furthermore, by modifying only the singular values, SVC improves the performance of Task Arithmetic by 13.0%. Code is available at https://github.com/lyymuwu/SVC.", "authors": ["Yayuan Li", "Ze Peng", "Jian Zhang", "Jintao Guo", "Yue Duan", "Yinghuan Shi"], "categories": ["cs.LG", "cs.AI", "cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-05", "url": "https://arxiv.org/abs/2602.05536", "pdf_url": "https://arxiv.org/pdf/2602.05536v2", "arxiv_id": "2602.05536", "doi": "10.48550/arXiv.2602.05536", "citation_count": 3, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/lyymuwu/SVC", "venue": "arXiv.org", "quality_score": 0.719}
{"id": "c660043a721b5b119eaa4310f588f2050d1b82cfe0306898dea8c6d906875faf", "sources": ["arxiv", "semantic_scholar"], "title": "Model-Dowser: Data-Free Importance Probing to Mitigate Catastrophic Forgetting in Multimodal Large Language Models", "abstract": "Fine-tuning Multimodal Large Language Models (MLLMs) on task-specific data is an effective way to improve performance on downstream applications. However, such adaptation often leads to a degradation in generalization on pretrained tasks, a phenomenon known as Catastrophic Forgetting. Existing methods that aim to mitigate this issue either become ineffective when fine-tuning deeper layers of the language decoder or scale poorly with increasing model size. To address these limitations, we propose Model-Dowser, a novel sparse fine-tuning approach for MLLMs. Model-Dowser measures a principled importance score for each model parameter with respect to pretrained generalization (prior to downstream adaptation) by jointly considering weight magnitudes, input activations, and output sensitivities. During fine-tuning, Model-Dowser selectively preserves high-importance parameters and updates the remaining. Comprehensive experiments on two representative MLLMs, LLaVA and NVILA, demonstrate that Model-Dowser effectively mitigates catastrophic forgetting and consistently outperforms prior methods, while remaining resource-efficient and scalable to multi-billion-parameter models.", "authors": ["Hyeontaek Hwang", "Nguyen Dinh Son", "Daeyoung Kim"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-04", "url": "https://arxiv.org/abs/2602.04509", "pdf_url": "https://arxiv.org/pdf/2602.04509v7", "arxiv_id": "2602.04509", "doi": "10.48550/arXiv.2602.04509", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4641}
{"id": "64eef9dd34cb982c11b268d9a395f77d488a6be603f8ec523276114a5859e4e7", "sources": ["arxiv", "semantic_scholar"], "title": "Self-Soupervision: Cooking Model Soups without Labels", "abstract": "Model soups are strange and strangely effective combinations of parameters. They take a model (the stock), fine-tune it into multiple models (the ingredients), and then mix their parameters back into one model (the soup) to improve predictions. While all known soups require supervised learning, and optimize the same loss on labeled data, our recipes for Self-Soupervision generalize soups to self-supervised learning (SSL). Our Self-Souping lets us flavor ingredients on new data sources, e.g. from unlabeled data from a task for transfer or from a shift for robustness. We show that Self-Souping on corrupted test data, then fine-tuning back on uncorrupted train data, boosts robustness by +3.5% (ImageNet-C) and +7% (LAION-C). Self-Soupervision also unlocks countless SSL algorithms to cook the diverse ingredients needed for more robust soups. We show for the first time that ingredients can differ in their SSL hyperparameters -- and more surprisingly, in their SSL algorithms. We cook soups of MAE, MoCoV3, MMCR, and LeJEPA ingredients that are more accurate than any single SSL ingredient.", "authors": ["Anthony Fuller", "James R. Green", "Evan Shelhamer"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-02", "url": "https://arxiv.org/abs/2602.02890", "pdf_url": "https://arxiv.org/pdf/2602.02890v2", "arxiv_id": "2602.02890", "doi": "10.48550/arXiv.2602.02890", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/antofuller/self_soupervision", "venue": "arXiv.org", "quality_score": 0.7136}
{"id": "85b3ebd467ba4360f31d987da572e76812044e55777a1ab27ea7d40841b7f359", "sources": ["arxiv", "semantic_scholar"], "title": "AutoMerge: Search-Based Model Merging Framework for Effective Model Reuse", "abstract": "Software reuse has long been recognized as a critical and widely studied topic in software engineering, offering substantial benefits in reducing development costs, improving software quality, and enhancing operational efficiency. This paradigm extends into deep learning through model reuse. Recently, model merging has emerged in the domain of large language models (LLMs) as a training-free approach that takes multiple task-specific models with the same architecture as source models and merges them without retraining, enhancing model reuse within LLMs. However, no prior work has systematically investigated whether such an approach can be effectively applied to other deep learning models with different architectures across domains. To bridge this gap, we present the first systematic study that evaluates five model merging techniques on three distinct model architectures across three domains: LLMs, image classification, and autonomous driving. Our findings reveal that directly applying existing model merging techniques leads to highly inconsistent results and falls notably short of their success within LLMs. Moreover, a single model merging technique often fails to handle the heterogeneous structural properties within a model, limiting its applicability to different model architectures across domains. Furthermore, the effectiveness of model merging techniques is highly sensitive to hyperparameter configurations, thereby constraining their potential for broader adoption. Inspired by these insights, we propose AutoMerge, a novel search-based model merging framework that first segments complex models into multiple heterogeneous blocks and then systematically explores the merging space to identify the merging technique and its hyperparameter configuration.", "authors": ["You Lu", "Jiyang Zhang", "Bihuan Chen", "Chaofeng Sha", "Dingji Wang", "Xin Peng"], "categories": ["cs.SE"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-30", "url": "https://arxiv.org/abs/2601.22748", "pdf_url": "https://arxiv.org/pdf/2601.22748v1", "arxiv_id": "2601.22748", "doi": "10.48550/arXiv.2601.22748", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4583}
{"id": "f2f794df11319183e552813d69f3ab08cb7c7e11c930aaffd41fdd33706ce502", "sources": ["arxiv", "semantic_scholar"], "title": "Per-parameter Task Arithmetic for Unlearning in Large Language Models", "abstract": "In large language model (LLM) unlearning, private information is required to be removed. Task arithmetic unlearns by subtracting a specific task vector (TV)--defined as the parameter difference between a privacy-information-tuned model and the original model. While efficient, it can cause over-forgetting by disrupting parameters essential for retaining other information. Motivated by the observation that each parameter exhibits different importance for forgetting versus retention, we propose a per-parameter task arithmetic (PerTA) mechanism to rescale the TV, allowing per-parameter adjustment. These weights quantify the relative importance of each parameter for forgetting versus retention, estimated via gradients (i.e., PerTA-grad) or the diagonal Fisher information approximation (i.e., PerTA-fisher). Moreover, we discuss the effectiveness of PerTA, extend it to a more general form, and provide further analysis. Extensive experiments demonstrate that PerTA consistently improves upon standard TV, and in many cases surpasses widely used training-based unlearning methods in both forgetting effectiveness and overall model utility. By retaining the efficiency of task arithmetic while mitigating over-forgetting, PerTA offers a principled and practical framework for LLM unlearning.", "authors": ["Chengyi Cai", "Zesheng Ye", "Jiangchao Yao", "Jianzhong Qi", "Bo Han", "Xiaolu Zhang", "Feng Liu", "Jun Zhou"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-29", "url": "https://arxiv.org/abs/2601.22030", "pdf_url": "https://arxiv.org/pdf/2601.22030v1", "arxiv_id": "2601.22030", "doi": "10.48550/arXiv.2601.22030", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4572}
{"id": "73c76cc32d5ab353f4ed6b6d4281fea131e0071f25e5aa5c90962a36908692c2", "sources": ["arxiv", "semantic_scholar"], "title": "SOUP: Token-level Single-sample Mix-policy Reinforcement Learning for Large Language Models", "abstract": "On-policy reinforcement learning (RL) methods widely used for language model post-training, like Group Relative Policy Optimization (GRPO), often suffer from limited exploration and early saturation due to low sampling diversity. While off-policy data can help, current approaches that mix entire trajectories cause significant policy mismatch and instability. In this work, we propose the $\\textbf{S}$ingle-sample Mix-p$\\textbf{O}$licy $\\textbf{U}$nified $\\textbf{P}$aradigm (SOUP), a framework that unifies off- and on-policy learning within individual samples at the token level. It confines off-policy influence to the prefix of a generated sequence sampled from historical policies, while the continuation is generated on-policy. Through token-level importance ratios, SOUP effectively leverages off-policy information while preserving training stability. Extensive experiments demonstrate that SOUP consistently outperforms standard on-policy training and existing off-policy extensions. Our further analysis clarifies how our fine-grained, single-sample mix-policy training can improve both exploration and final performance in LLM RL.", "authors": ["Lei Yang", "Wei Bi", "Chenxi Sun", "Renren Jin", "Deyi Xiong"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-29", "url": "https://arxiv.org/abs/2601.21476", "pdf_url": "https://arxiv.org/pdf/2601.21476v1", "arxiv_id": "2601.21476", "doi": "10.48550/arXiv.2601.21476", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4572}
{"id": "1bbe87ba20db336e216b7930120005615a6c26b8f8d4fe98d7dca21b8e94cbc6", "sources": ["arxiv", "semantic_scholar"], "title": "Multi-task Code LLMs: Data Mix or Model Merge?", "abstract": "Recent research advocates deploying smaller, specialized code LLMs in agentic frameworks alongside frontier models, sparking interest in efficient strategies for multi-task learning that balance performance, constraints, and costs. We compare two approaches for creating small, multi-task code LLMs: data mixing versus model merging. We conduct extensive experiments across two model families (Qwen Coder and DeepSeek Coder) at two scales (2B and 7B parameters), fine-tuning them for code generation and code summarization tasks. Our evaluation on HumanEval, MBPP, and CodeXGlue benchmarks reveals that model merging achieves the best overall performance at larger scale across model families, retaining 96% of specialized model performance on code generation tasks while maintaining summarization capabilities. Notably, merged models can even surpass individually fine-tuned models, with our best configuration of Qwen Coder 2.5 7B model achieving 92.7% Pass@1 on HumanEval compared to 90.9% for its task-specific fine-tuned equivalent. At a smaller scale we find instead data mixing to be a preferred strategy. We further introduce a weight analysis technique to understand how different tasks affect model parameters and their implications for merging strategies. The results suggest that careful merging and mixing strategies can effectively combine task-specific capabilities without significant performance degradation, making them ideal for resource-constrained deployment scenarios.", "authors": ["Mingzhi Zhu", "Boris Sobolev", "Rahul Krishna", "Raju Pavuluri", "Stacy Patterson", "Michele Merler"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-28", "url": "https://arxiv.org/abs/2601.21115", "pdf_url": "https://arxiv.org/pdf/2601.21115v1", "arxiv_id": "2601.21115", "doi": "10.48550/arXiv.2601.21115", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.456}
{"id": "32e0c9120fa185cfe7cb9859c407bbcd3ff747aa423e8e2a584cc8e85d3b6556", "sources": ["arxiv", "semantic_scholar"], "title": "Behavior Knowledge Merge in Reinforced Agentic Models", "abstract": "Reinforcement learning (RL) is central to post-training, particularly for agentic models that require specialized reasoning behaviors. In this setting, model merging offers a practical mechanism for integrating multiple RL-trained agents from different tasks into a single generalist model. However, existing merging methods are designed for supervised fine-tuning (SFT), and they are suboptimal to preserve task-specific capabilities on RL-trained agentic models. The root is a task-vector mismatch between RL and SFT: on-policy RL induces task vectors that are highly sparse and heterogeneous, whereas SFT-style merging implicitly assumes dense and globally comparable task vectors. When standard global averaging is applied under this mismatch, RL's non-overlapping task vectors that encode critical task-specific behaviors are reduced and parameter updates are diluted. To address this issue, we propose Reinforced Agent Merging (RAM), a distribution-aware merging framework explicitly designed for RL-trained agentic models. RAM disentangles shared and task-specific unique parameter updates, averaging shared components while selectively preserving and rescaling unique ones to counteract parameter update dilution. Experiments across multiple agent domains and model architectures demonstrate that RAM not only surpasses merging baselines, but also unlocks synergistic potential among agents to achieve performance superior to that of specialized agents in their domains.", "authors": ["Xiangchi Yuan", "Dachuan Shi", "Chunhui Zhang", "Zheyuan Liu", "Shenglong Yao", "Soroush Vosoughi", "Wenke Lee"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-20", "url": "https://arxiv.org/abs/2601.13572", "pdf_url": "https://arxiv.org/pdf/2601.13572v1", "arxiv_id": "2601.13572", "doi": "10.48550/arXiv.2601.13572", "citation_count": 6, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4469}
{"id": "8c344cd2626c159a853f5559148041fbd02b5e3a6786b03c83f5538680a06a44", "sources": ["arxiv", "semantic_scholar"], "title": "Will it Merge? On The Causes of Model Mergeability", "abstract": "Model merging has emerged as a promising technique for combining multiple fine-tuned models into a single multitask model without retraining. However, the factors that determine whether merging will succeed or fail remain poorly understood. In this work, we investigate why specific models are merged better than others. To do so, we propose a concrete, measurable definition of mergeability. We investigate several potential causes for high or low mergeability, highlighting the base model knowledge as a dominant factor: Models fine-tuned on instances that the base model knows better are more mergeable than models fine-tuned on instances that the base model struggles with. Based on our mergeability definition, we explore a simple weighted merging technique that better preserves weak knowledge in the base model.", "authors": ["Adir Rahamim", "Asaf Yehudai", "Boaz Carmeli", "Leshem Choshen", "Yosi Mass", "Yonatan Belinkov"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-10", "url": "https://arxiv.org/abs/2601.06672", "pdf_url": "https://arxiv.org/pdf/2601.06672v1", "arxiv_id": "2601.06672", "doi": "10.48550/arXiv.2601.06672", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4354}
{"id": "60c611a8cf880669efa34be0297a7de91185e586e17895eee264693bf5f928e5", "sources": ["arxiv", "semantic_scholar"], "title": "Model Merging via Multi-Teacher Knowledge Distillation", "abstract": "Model merging has emerged as a lightweight alternative to joint multi-task learning (MTL), yet the generalization properties of merged models remain largely unexplored. Establishing such theoretical guarantees is non-trivial, as the merging process typically forbids access to the original training data and involves combining fine-tuned models trained on fundamentally heterogeneous data distributions. Without a principled understanding of these dynamics, current methods often rely on heuristics to approximate the optimal combination of parameters. This dependence is most critical in coefficient scaling, the weighting factors that modulate the magnitude of each fine-tuned model's contribution to the shared parameter. However, without a principled objective to guide their selection, these methods lead to brittle performance and are highly sensitive to scaling initialization. We address this gap by (i) establishing a novel flatness-aware PAC-Bayes generalization bound specifically for the model merging setting. This analysis introduces a \"cross-task heterogeneity\" term that formally captures the mismatch between diverse fine-tuned model priors and the target multi-task distributions. Guided by this theoretical insight, (ii) we frame model merging as multi-teacher knowledge distillation on scarce, unlabeled data. We formally demonstrate that minimizing the student-teacher Kullback-Leibler divergence directly tightens the upper bound on the merged model's excess risk. Guided by the flatness-aware bound derived, (iii) we operationalize this objective via SAMerging, a method that employs Sharpness-Aware Minimization (SAM) to find flat minima. Empirically, SAMerging establishes a new state of the art across vision and NLP benchmarks, achieving remarkable performance. The code is available at https://github.com/arshandalili/SAMerging.", "authors": ["Seyed Arshan Dalili", "Mehrdad Mahdavi"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-24", "url": "https://arxiv.org/abs/2512.21288", "pdf_url": "https://arxiv.org/pdf/2512.21288v1", "arxiv_id": "2512.21288", "doi": "10.48550/arXiv.2512.21288", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/arshandalili/SAMerging", "venue": "arXiv.org", "quality_score": 0.6428}
{"id": "1b5d0401b1e0b3674ad18977ff1823de10087727f5f161c57ee398e7292ac6d2", "sources": ["arxiv", "semantic_scholar"], "title": "MAGIC: Achieving Superior Model Merging via Magnitude Calibration", "abstract": "The proliferation of pre-trained models has given rise to a wide array of specialised, fine-tuned models. Model merging aims to merge the distinct capabilities of these specialised models into a unified model, requiring minimal or even no additional training. A core objective of model merging is to ensure the merged model retains the behavioural characteristics of the specialised models, typically achieved through feature alignment. We identify that features consist of two critical components: direction and magnitude. Prior research has predominantly focused on directional alignment, while the influence of magnitude remains largely neglected, despite its pronounced vulnerability to perturbations introduced by common merging operations (e.g., parameter fusion and sparsification). Such perturbations to magnitude inevitably lead to feature deviations in the merged model from the specialised models, resulting in subsequent performance degradation. To address this, we propose MAGnItude Calibration (MAGIC), a plug-and-play framework that rectifies layer-wise magnitudes in feature and weight spaces, with three variants. Specifically, our Feature Space Calibration (FSC) realigns the merged model's features using a small set of unlabelled data, while Weight Space Calibration (WSC) extends this calibration to the weight space without requiring additional data. Combining these yields Dual Space Calibration (DSC). Comprehensive experiments demonstrate that MAGIC consistently boosts performance across diverse Computer Vision tasks (+4.3% on eight datasets) and NLP tasks (+8.0% on Llama) without additional training. Our code is available at: https://github.com/lyymuwu/MAGIC", "authors": ["Yayuan Li", "Jian Zhang", "Jintao Guo", "Zihan Cheng", "Lei Qi", "Yinghuan Shi", "Yang Gao"], "categories": ["cs.LG", "cs.AI", "cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-22", "url": "https://arxiv.org/abs/2512.19320", "pdf_url": "https://arxiv.org/pdf/2512.19320v1", "arxiv_id": "2512.19320", "doi": "10.48550/arXiv.2512.19320", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/lyymuwu/MAGIC", "venue": "arXiv.org", "quality_score": 0.6393}
{"id": "9fd992bbeca9afbab3147779a5ff8af5ff92576ee4456e9f814bbcf15d758bbc", "sources": ["arxiv", "semantic_scholar"], "title": "Uniform Interpolation", "abstract": "Uniform interpolation is a strengthening of interpolation that holds for certain propositional logics. The starting point of this chapter is a theorem of A. Pitts, which shows that uniform interpolation holds for intuitionistic propositional logic. We outline how this theorem may be proved semantically via the definability of bisimulation quantifiers, and how it generalizes to an open mapping theorem between Esakia spaces. We also discuss connections between uniform interpolation and research in categorical logic, algebra, and model theory.", "authors": ["Sam van Gool"], "categories": ["math.LO", "cs.LO"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2025-12-17", "url": "https://arxiv.org/abs/2512.15391", "pdf_url": "https://arxiv.org/pdf/2512.15391v3", "arxiv_id": "2512.15391", "doi": "10.48550/arXiv.2512.15391", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4079}
{"id": "bf66ba2d409e7fa116abe1d3dd3dcee8a47426a45c77acd8bf42e007c4d00ee7", "sources": ["arxiv", "semantic_scholar"], "title": "Per-Axis Weight Deltas for Frequent Model Updates", "abstract": "Serving many task-specialized LLM variants is often limited by the large size of fine-tuned checkpoints and the resulting cold-start latency. Since fine-tuned weights differ from their base model by relatively small structured residuals, a natural approach is to represent them as compressed deltas. We propose a simple 1-bit delta scheme that stores only the sign of the weight difference together with lightweight per-axis (row/column) FP16 scaling factors, learned from a small calibration set. This design preserves the compactness of 1-bit deltas while more accurately capturing variation across weight dimensions, leading to improved reconstruction quality over scalar alternatives. From a systems perspective, a streamlined loader that transfers packed deltas in a single operation per module reduces cold-start latency and storage overhead, with artifacts several times smaller than a full FP16 checkpoint. The method is drop-in, requires minimal calibration data, and maintains inference efficiency by avoiding dense reconstruction. Our experimental setup and source code are available at https://github.com/kuiumdjiev/Per-Axis-Weight-Deltas-for-Frequent-Model-Updates.", "authors": ["Stefan Kuyumdzhiev", "Radostin Cholakov"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-16", "url": "https://arxiv.org/abs/2512.19720", "pdf_url": "https://arxiv.org/pdf/2512.19720v1", "arxiv_id": "2512.19720", "doi": "10.48550/arXiv.2512.19720", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/kuiumdjiev/Per-Axis-Weight-Deltas-for-Frequent-Model-Updates", "venue": "arXiv.org", "quality_score": 0.6286}
{"id": "11b40cfbec786eafba8795f92876a45ffc4ac6f78049ede64d82dfe00782309d", "sources": ["arxiv", "semantic_scholar"], "title": "Interpolation in Knowledge Representation", "abstract": "Craig interpolation and uniform interpolation have many applications in knowledge representation, including explainability, forgetting, modularization and reuse, and even learning. At the same time, many relevant knowledge representation formalisms do in general not have Craig or uniform interpolation, and computing interpolants in practice is challenging. We have a closer look at two prominent knowledge representation formalisms, description logics and logic programming, and discuss theoretical results and practical methods for computing interpolants.", "authors": ["Jean Christoph Jung", "Patrick Koopmann", "Matthias Knorr"], "categories": ["cs.AI", "cs.LO"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-09", "url": "https://arxiv.org/abs/2512.08833", "pdf_url": "https://arxiv.org/pdf/2512.08833v1", "arxiv_id": "2512.08833", "doi": "10.48550/arXiv.2512.08833", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3987}
{"id": "f9a84988676292f9bdeb0de68ce40ee993dad14ba5d8aa9cb77eee7f337c2297", "sources": ["arxiv", "semantic_scholar"], "title": "An Empirical Survey of Model Merging Algorithms for Social Bias Mitigation", "abstract": "Large language models (LLMs) are known to inherit and even amplify societal biases present in their pre-training corpora, threatening fairness and social trust. To address this issue, recent work has explored ``editing'' LLM parameters to mitigate social bias with model merging approaches; however, there is no empirical comparison. In this work, we empirically survey seven algorithms: Linear, Karcher Mean, SLERP, NuSLERP, TIES, DELLA, and Nearswap, applying 13 open weight models in the GPT, LLaMA, and Qwen families. We perform a comprehensive evaluation using three bias datasets (BBQ, BOLD, and HONEST) and measure the impact of these techniques on LLM performance in downstream tasks of the SuperGLUE benchmark. We find a trade-off between bias reduction and downstream performance: methods achieving greater bias mitigation degrade accuracy, particularly on tasks requiring reading comprehension and commonsense and causal reasoning. Among the merging algorithms, Linear, SLERP, and Nearswap consistently reduce bias while maintaining overall performance, with SLERP at moderate interpolation weights emerging as the most balanced choice. These results highlight the potential of model merging algorithms for bias mitigation, while indicating that excessive debiasing or inappropriate merging methods may lead to the degradation of important linguistic abilities.", "authors": ["Daiki Shirafuji", "Tatsuhiko Saito", "Yasutomo Kimura"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-02", "url": "https://arxiv.org/abs/2512.02689", "pdf_url": "https://arxiv.org/pdf/2512.02689v1", "arxiv_id": "2512.02689", "doi": "10.48550/arXiv.2512.02689", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Pacific Asia Conference on Language, Information and Computation", "quality_score": 0.3907}
{"id": "2f154303d35a995619f92c14c3a0c2fe90b167e7f511e42636cc057cb775ddeb", "sources": ["arxiv", "semantic_scholar"], "title": "Interpolation in Non-Classical Logics", "abstract": "This chapter surveys some of the main results on interpolation in several of the most prominent families of non-classical logics. Special attention is given to the distinction between the two most commonly studied variants of interpolation--namely, Craig interpolation and deductive interpolation. Our discussion focuses primarily on how these properties present in families of logical systems taken as a whole, particularly those comprising all axiomatic extensions of any of several notable non-classical logics. We consider a range of important examples: superintuitionistic and modal logics, fuzzy logics, paraconsistent logics, relevant logics, and substructural logics.", "authors": ["Wesley Fussner"], "categories": ["math.LO", "cs.LO"], "fields_of_study": ["Mathematics", "Computer Science"], "published_date": "2025-12-01", "url": "https://arxiv.org/abs/2512.01600", "pdf_url": "https://arxiv.org/pdf/2512.01600v1", "arxiv_id": "2512.01600", "doi": "10.48550/arXiv.2512.01600", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3896}
{"id": "0c5159109d17abf48e407b94f9ab445211d9b7ed1e2fd79040f42f292d2b45da", "sources": ["arxiv", "semantic_scholar"], "title": "Stay Unique, Stay Efficient: Preserving Model Personality in Multi-Task Merging", "abstract": "Model merging has emerged as a promising paradigm for enabling multi-task capabilities without additional training. However, existing methods often experience substantial performance degradation compared with individually fine-tuned models, even on similar tasks, underscoring the need to preserve task-specific information. This paper proposes Decomposition, Thresholding, and Scaling (DTS), an approximation-based personalized merging framework that preserves task-specific information with minimal storage overhead. DTS first applies singular value decomposition to the task-specific information and retains only a small subset of singular values and vectors. It then introduces a novel thresholding strategy that partitions singular vector elements into groups and assigns a scaling factor to each group. To enable generalization to unseen tasks, we further extend DTS with a variant that fuses task-specific information in a data-free manner based on the semantic similarity of task characteristics. Extensive experiments demonstrate that DTS consistently outperforms state-of-the-art baselines while requiring only 1\\% additional storage per task. Furthermore, experiments on unseen tasks show that the DTS variant achieves significantly better generalization performance. Our code is available at https://github.com/krumpguo/DTS.", "authors": ["Kuangpu Guo", "Yuhe Ding", "Jian Liang", "Zilei Wang", "Ran He"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-01", "url": "https://arxiv.org/abs/2512.01461", "pdf_url": "https://arxiv.org/pdf/2512.01461v1", "arxiv_id": "2512.01461", "doi": "10.48550/arXiv.2512.01461", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/krumpguo/DTS", "venue": "arXiv.org", "quality_score": 0.6021}
{"id": "6f68ef8a58ff5ee806cb91118686181058b1796afffc9f680e7f6c167524398b", "sources": ["arxiv", "semantic_scholar"], "title": "Merge and Bound: Direct Manipulations on Weights for Class Incremental Learning", "abstract": "We present a novel training approach, named Merge-and-Bound (M&B) for Class Incremental Learning (CIL), which directly manipulates model weights in the parameter space for optimization. Our algorithm involves two types of weight merging: inter-task weight merging and intra-task weight merging. Inter-task weight merging unifies previous models by averaging the weights of models from all previous stages. On the other hand, intra-task weight merging facilitates the learning of current task by combining the model parameters within current stage. For reliable weight merging, we also propose a bounded update technique that aims to optimize the target model with minimal cumulative updates and preserve knowledge from previous tasks; this strategy reveals that it is possible to effectively obtain new models near old ones, reducing catastrophic forgetting. M&B is seamlessly integrated into existing CIL methods without modifying architecture components or revising learning objectives. We extensively evaluate our algorithm on standard CIL benchmarks and demonstrate superior performance compared to state-of-the-art methods.", "authors": ["Taehoon Kim", "Donghwan Jang", "Bohyung Han"], "categories": ["cs.CV", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-26", "url": "https://arxiv.org/abs/2511.21490", "pdf_url": "https://arxiv.org/pdf/2511.21490v1", "arxiv_id": "2511.21490", "doi": "10.48550/arXiv.2511.21490", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3839}
{"id": "1c5a372283c92d5cc4d8f0c5bc464528373ef5216f69c37f36f25e1a8465a520", "sources": ["arxiv", "semantic_scholar"], "title": "A Systematic Study of In-the-Wild Model Merging for Large Language Models", "abstract": "Model merging combines multiple fine-tuned checkpoints into a single model without additional training, offering an attractive approach to reusing models and efficiently improving performance. However, it remains unclear whether the advantages reported for settings where all merged experts have distinct roles and are tuned on clearly separated tasks also hold in settings where the merged experts do not have clearly distinct roles, but are trained on overlapping or even conflicting objectives. To evaluate this setting, we present a large-scale, systematic evaluation of \"in-the-wild\" model merging of heterogeneous experts, that may have been trained on overlapping or conflicting objectives. Concretely, we evaluate six state-of-the-art merging methods, including recent subspace methods, across four open-weight LLMs, twelve fine-tuned checkpoints per base model, and sixteen standard LLM benchmarks. Evaluating through standardized benchmarks, we measure both the probability that a model merged from a heterogeneous set of experts outperforms the base model and we measure relative gains over the best individual checkpoint. Our results show that the oldest and simplest method, Task Arithmetic, is the only approach that reliably yields performance gains on LLMs in this \"in-the-wild\" setting. Other interference-aware and subspace merging methods typically do not result in notable improvements over the base model. Our findings indicate that current merging techniques mostly do not enable extracting useful weight updates from heterogeneous and potentially conflicting versions. This motivates the design of LLM-specific merging algorithms and merging-aware fine-tuning methods.", "authors": ["Oğuz Kağan Hitit", "Leander Girrbach", "Zeynep Akata"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-26", "url": "https://arxiv.org/abs/2511.21437", "pdf_url": "https://arxiv.org/pdf/2511.21437v2", "arxiv_id": "2511.21437", "doi": null, "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Transactions on Machine Learning Research (03/2026)", "quality_score": 0.3839}
{"id": "88c66ad8c9d2d8aac7aa6523e2ab20ae18e196dacefd68c80362764eb6bd514d", "sources": ["arxiv", "semantic_scholar"], "title": "Merging without Forgetting: Continual Fusion of Task-Specific Models via Optimal Transport", "abstract": "Merging models fine-tuned for different tasks into a single unified model has become an increasingly important direction for building versatile, efficient multi-task systems. Existing approaches predominantly rely on parameter interpolation in weight space, which we show introduces significant distribution shift in the feature space and undermines task-specific knowledge. In this paper, we propose OTMF (Optimal Transport-based Masked Fusion), a novel model merging framework rooted in optimal transport theory to address the distribution shift that arises from naive parameter interpolation. Instead of directly aggregating features or weights, OTMF aligns the semantic geometry of task-specific models by discovering common masks applied to task vectors through optimal transport plans. These masks selectively extract transferable and task-agnostic components while preserving the unique structural identities of each task. To ensure scalability in real-world settings, OTMF further supports a continual fusion paradigm that incrementally integrates each new task vector without revisiting previous ones, maintaining a bounded memory footprint and enabling efficient fusion across a growing number of tasks. We conduct comprehensive experiments on multiple vision and language benchmarks, and results show that OTMF achieves state-of-the-art performance in terms of both accuracy and efficiency. These findings highlight the practical and theoretical value of our approach to model merging.", "authors": ["Zecheng Pan", "Zhikang Chen", "Ding Li", "Min Zhang", "Sen Cui", "Hongshuo Jin", "Luqi Tao", "Yi Yang", "Deheng Ye", "Yu Zhang", "Tingting Zhu", "Tianling Ren"], "categories": ["cs.LG", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-24", "url": "https://arxiv.org/abs/2511.19561", "pdf_url": "https://arxiv.org/pdf/2511.19561v1", "arxiv_id": "2511.19561", "doi": "10.48550/arXiv.2511.19561", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3816}
{"id": "22bccf603496fbbc62e67e95feede8f7aa8aad425ff061d8b955282a518080df", "sources": ["arxiv", "semantic_scholar"], "title": "Escaping Optimization Stagnation: Taking Steps Beyond Task Arithmetic via Difference Vectors", "abstract": "Current methods for editing pre-trained models face significant challenges, primarily high computational costs and limited scalability. Task arithmetic has recently emerged as a promising solution, using simple arithmetic operations-addition and negation-based on task vectors which are the differences between fine-tuned and pre-trained model weights, to efficiently modify model behavior. However, the full potential of task arithmetic remains underexplored, primarily due to limited mechanisms for overcoming optimization stagnation. To address this challenge, we introduce the notion of difference vector, a generalized form of task vectors derived from the historical movements during optimization. Using difference vectors as directed perturbations, we propose the Difference Vector-based Anisotropic Scaling Iterative algorithm (DV-BASI) to enable a continuous optimization process for task arithmetic methods without relying on any additional modules or components. Notably, by leveraging escapability and directional advantages of difference vectors, the average performance on different tasks of the multi-task model merged by DV-BASI may even outperform models individually fine-tuned. Based on this observation, we extend the application of difference vectors to a feasible fine-tuning method for single-task models. On the practical side, DV-BASI allows expressive searching directions with few learnable parameters and forms a scalable framework. We also integrate DV-BASI with task arithmetic methods and advanced optimization techniques to achieve state-of-the-art performance on both supervised and unsupervised evaluation protocols.", "authors": ["Jinping Wang", "Zhiqiang Gao", "Dinggen Zhang", "Zhiwu Xie"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-22", "url": "https://arxiv.org/abs/2511.17987", "pdf_url": "https://arxiv.org/pdf/2511.17987v1", "arxiv_id": "2511.17987", "doi": "10.48550/arXiv.2511.17987", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.3793}
{"id": "f7ece0a771f4beea649c8d8a2c8fac58c1def5e7d6fc5618e35828cd3666cd75", "sources": ["arxiv", "semantic_scholar"], "title": "Task Addition and Weight Disentanglement in Closed-Vocabulary Models", "abstract": "Task arithmetic has recently emerged as a promising method for editing pre-trained \\textit{open-vocabulary} models, offering a cost-effective alternative to standard multi-task fine-tuning. However, despite the abundance of \\textit{closed-vocabulary} models that are not pre-trained with language supervision, applying task arithmetic to these models remains unexplored. In this paper, we deploy and study task addition in closed-vocabulary image classification models. We consider different pre-training schemes and find that \\textit{weight disentanglement} -- the property enabling task arithmetic -- is a general consequence of pre-training, as it appears in different pre-trained closed-vocabulary models. In fact, we find that pre-trained closed-vocabulary vision transformers can also be edited with task arithmetic, achieving high task addition performance and enabling the efficient deployment of multi-task models. Finally, we demonstrate that simple linear probing is a competitive baseline to task addition. Overall, our findings expand the applicability of task arithmetic to a broader class of pre-trained models and open the way for more efficient use of pre-trained models in diverse settings.", "authors": ["Adam Hazimeh", "Alessandro Favero", "Pascal Frossard"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-18", "url": "https://arxiv.org/abs/2511.14569", "pdf_url": "https://arxiv.org/pdf/2511.14569v1", "arxiv_id": "2511.14569", "doi": "10.48550/arXiv.2511.14569", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3747}
{"id": "956fb0caa34a99fb85a2586c42a37fbf4c566a25a1ef054bd3d390aada0ea8cd", "sources": ["arxiv", "semantic_scholar"], "title": "MergeSlide: Continual Model Merging and Task-to-Class Prompt-Aligned Inference for Lifelong Learning on Whole Slide Images", "abstract": "Lifelong learning on Whole Slide Images (WSIs) aims to train or fine-tune a unified model sequentially on cancer-related tasks, reducing the resources and effort required for data transfer and processing, especially given the gigabyte-scale size of WSIs. In this paper, we introduce MergeSlide, a simple yet effective framework that treats lifelong learning as a model merging problem by leveraging a vision-language pathology foundation model. When a new task arrives, it is: 1) defined with class-aware prompts, 2) fine-tuned for a few epochs using an MLP-free backbone, and 3) merged into a unified model using an orthogonal continual merging strategy that preserves performance and mitigates catastrophic forgetting. For inference under the class-incremental learning (CLASS-IL) setting, where task identity is unknown, we introduce Task-to-Class Prompt-aligned (TCP) inference. Specifically, TCP first identifies the most relevant task using task-level prompts and then applies the corresponding class-aware prompts to generate predictions. To evaluate MergeSlide, we conduct experiments on a stream of six TCGA datasets. The results show that MergeSlide outperforms both rehearsal-based continual learning and vision-language zero-shot baselines. Code and data are available at https://github.com/caodoanh2001/MergeSlide.", "authors": ["Doanh C. Bui", "Ba Hung Ngo", "Hoai Luan Pham", "Khang Nguyen", "Maï K. Nguyen", "Yasuhiko Nakashima"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-17", "url": "https://arxiv.org/abs/2511.13099", "pdf_url": "https://arxiv.org/pdf/2511.13099v1", "arxiv_id": "2511.13099", "doi": "10.1109/WACV61042.2026.00472", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/caodoanh2001/MergeSlide", "venue": "IEEE Workshop/Winter Conference on Applications of Computer Vision", "quality_score": 0.5773}
{"id": "1f43f2ec73aaf6558699cfbe33e8e51dd0cb00f6c2239d0d043640eda523af40", "sources": ["arxiv", "semantic_scholar"], "title": "A Novel Hierarchical Integration Method for Efficient Model Merging in Medical LLMs", "abstract": "Large Language Models (LLMs) face significant challenges in distributed healthcare, including consolidating specialized domain knowledge across institutions while maintaining privacy, reducing computational overhead, and preventing catastrophic forgetting during model updates.This paper presents a systematic evaluation of six parameter-space merging techniques applied to two architecturally compatible medical LLMs derived from the Mistral-7B base model. We introduce a novel hierarchical method that combines selective Optimal Transport (OT) alignment for attention layers with cosine similarity-weighted interpolation, designed to address permutation variance while minimizing computational overhead for edge deployment scenarios. Our study evaluates Task Arithmetic, Linear Averaging, DARE-TIES, DELLA, Breadcrumbs, and our Hierarchical approach across five medical benchmarks. Results demonstrate that architecturally compatible models benefit significantly from simple averaging methods, with Task Arithmetic achieving 45.80% accuracy on MedQA, outperforming complex pruning-based approaches. These findings offer critical insights for the deployment of distributed medical AI in resource-constrained IoT environments, where computational efficiency and model compatibility are paramount. Our work establishes that for architecturally compatible models, simple averaging provides a robust and computationally efficient baseline for knowledge consolidation, offering a pragmatic path forward for scalable medical AI systems.", "authors": ["Prakrit Timilsina", "Anuj Nepal", "Rajan Kadel", "Robin Doss"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-17", "url": "https://arxiv.org/abs/2511.13373", "pdf_url": "https://arxiv.org/pdf/2511.13373v1", "arxiv_id": "2511.13373", "doi": "10.1109/SmartIoT66867.2025.00031", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Conferences on Smart Internet of Things", "quality_score": 0.3735}
{"id": "2bf4488fb9d4c127e6f476c66ff2a1d8a566046d75fbde200fb561814b74f6e2", "sources": ["arxiv", "semantic_scholar"], "title": "Defending Unauthorized Model Merging via Dual-Stage Weight Protection", "abstract": "The rapid proliferation of pretrained models and open repositories has made model merging a convenient yet risky practice, allowing free-riders to combine fine-tuned models into a new multi-capability model without authorization. Such unauthorized model merging not only violates intellectual property rights but also undermines model ownership and accountability. To address this issue, we present MergeGuard, a proactive dual-stage weight protection framework that disrupts merging compatibility while maintaining task fidelity. In the first stage, we redistribute task-relevant information across layers via L2-regularized optimization, ensuring that important gradients are evenly dispersed. In the second stage, we inject structured perturbations to misalign task subspaces, breaking curvature compatibility in the loss landscape. Together, these stages reshape the model's parameter geometry such that merged models collapse into destructive interference while the protected model remains fully functional. Extensive experiments on both vision (ViT-L-14) and language (Llama2, Gemma2, Mistral) models demonstrate that MergeGuard reduces merged model accuracy by up to 90% with less than 1.5% performance loss on the protected model.", "authors": ["Wei-Jia Chen", "Min-Yen Tsai", "Cheng-Yi Lee", "Chia-Mu Yu"], "categories": ["cs.CV", "cs.CR"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-14", "url": "https://arxiv.org/abs/2511.11851", "pdf_url": "https://arxiv.org/pdf/2511.11851v3", "arxiv_id": "2511.11851", "doi": "10.48550/arXiv.2511.11851", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3701}
{"id": "af45a91f5eea018489d72d95ab3226dd0d3c129796702912eba7598774ca41a1", "sources": ["arxiv", "semantic_scholar"], "title": "Do Not Merge My Model! Safeguarding Open-Source LLMs Against Unauthorized Model Merging", "abstract": "Model merging has emerged as an efficient technique for expanding large language models (LLMs) by integrating specialized expert models. However, it also introduces a new threat: model merging stealing, where free-riders exploit models through unauthorized model merging. Unfortunately, existing defense mechanisms fail to provide effective protection. Specifically, we identify three critical protection properties that existing methods fail to simultaneously satisfy: (1) proactively preventing unauthorized merging; (2) ensuring compatibility with general open-source settings; (3) achieving high security with negligible performance loss. To address the above issues, we propose MergeBarrier, a plug-and-play defense that proactively prevents unauthorized merging. The core design of MergeBarrier is to disrupt the Linear Mode Connectivity (LMC) between the protected model and its homologous counterparts, thereby eliminating the low-loss path required for effective model merging. Extensive experiments show that MergeBarrier effectively prevents model merging stealing with negligible accuracy loss.", "authors": ["Qinfeng Li", "Miao Pan", "Jintao Chen", "Fu Teng", "Zhiqiang Shen", "Ge Su", "Hao Peng", "Xuhong Zhang"], "categories": ["cs.CR", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-13", "url": "https://arxiv.org/abs/2511.10712", "pdf_url": "https://arxiv.org/pdf/2511.10712v2", "arxiv_id": "2511.10712", "doi": "10.48550/arXiv.2511.10712", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.5702}
{"id": "b22815202aa071278bc54ea9838a5d2caf6242a66e15042a8478e74146fe6059", "sources": ["arxiv", "semantic_scholar"], "title": "LT-Soups: Bridging Head and Tail Classes via Subsampled Model Soups", "abstract": "Real-world datasets typically exhibit long-tailed (LT) distributions, where a few head classes dominate and many tail classes are severely underrepresented. While recent work shows that parameter-efficient fine-tuning (PEFT) methods like LoRA and AdaptFormer preserve tail-class performance on foundation models such as CLIP, we find that they do so at the cost of head-class accuracy. We identify the head-tail ratio, the proportion of head to tail classes, as a crucial but overlooked factor influencing this trade-off. Through controlled experiments on CIFAR100 with varying imbalance ratio ($ρ$) and head-tail ratio ($η$), we show that PEFT excels in tail-heavy scenarios but degrades in more balanced and head-heavy distributions. To overcome these limitations, we propose LT-Soups, a two-stage model soups framework designed to generalize across diverse LT regimes. In the first stage, LT-Soups averages models fine-tuned on balanced subsets to reduce head-class bias; in the second, it fine-tunes only the classifier on the full dataset to restore head-class accuracy. Experiments across six benchmark datasets show that LT-Soups achieves superior trade-offs compared to both PEFT and traditional model soups across a wide range of imbalance regimes.", "authors": ["Masih Aminbeidokhti", "Subhankar Roy", "Eric Granger", "Elisa Ricci", "Marco Pedersoli"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-11", "url": "https://arxiv.org/abs/2511.10683", "pdf_url": "https://arxiv.org/pdf/2511.10683v2", "arxiv_id": "2511.10683", "doi": "10.48550/arXiv.2511.10683", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3667}
{"id": "d880ec144875557a21089ac082477e66e598253f377293425d17a849f77b811f", "sources": ["arxiv", "semantic_scholar"], "title": "Steering Language Models with Weight Arithmetic", "abstract": "Providing high-quality feedback to Large Language Models (LLMs) on a diverse training distribution can be difficult and expensive, and providing feedback only on a narrow distribution can result in unintended generalizations. To better leverage narrow training data, we propose contrastive weight steering, a simple post-training method that edits the model parameters using weight arithmetic. We isolate a behavior direction in weight-space by subtracting the weight deltas from two small fine-tunes -- one that induces the desired behavior and another that induces its opposite -- and then add or remove this direction to modify the model's weights. We apply this technique to mitigate sycophancy and induce misalignment, and find that weight steering often generalizes further than activation steering, achieving stronger out-of-distribution behavioral control before degrading general capabilities. We also show that, in the context of task-specific fine-tuning, weight steering can partially mitigate undesired behavioral drift: it can reduce sycophancy and under-refusals introduced during fine-tuning while preserving task performance gains. Finally, we provide preliminary evidence that emergent misalignment can be detected by measuring the similarity between fine-tuning updates and an \"evil\" weight direction, suggesting that it may be possible to monitor the evolution of weights during training and detect rare misaligned behaviors that never manifest during training or evaluations.", "authors": ["Constanza Fierro", "Fabien Roger"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-07", "url": "https://arxiv.org/abs/2511.05408", "pdf_url": "https://arxiv.org/pdf/2511.05408v2", "arxiv_id": "2511.05408", "doi": "10.48550/arXiv.2511.05408", "citation_count": 8, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3621}
{"id": "6baa54d76f87e54f2662c813a06ce971062d0e3752c096e81291d0fe3fe31fc5", "sources": ["arxiv", "semantic_scholar"], "title": "Polarization-resolved imaging improves eye tracking", "abstract": "Polarization-resolved near-infrared imaging adds a useful optical contrast mechanism to eye tracking by measuring the polarization state of light reflected by ocular tissues in addition to its intensity. In this paper we demonstrate how this contrast can be used to enable eye tracking. Specifically, we demonstrate that a polarization-enabled eye tracking (PET) system composed of a polarization--filter--array camera paired with a linearly polarized near-infrared illuminator can reveal trackable features across the sclera and gaze-informative patterns on the cornea, largely absent in intensity-only images. Across a cohort of 346 participants, convolutional neural network based machine learning models trained on data from PET reduced the median 95th-percentile absolute gaze error by 10--16\\% relative to capacity-matched intensity baselines under nominal conditions and in the presence of eyelid occlusions, eye-relief changes, and pupil-size variation. These results link light--tissue polarization effects to practical gains in human--computer interaction and position PET as a simple, robust sensing modality for future wearable devices.", "authors": ["Mantas Žurauskas", "Tom Bu", "Sanaz Alali", "Beyza Kalkanli", "Derek Shi", "Fernando Alamos", "Gauresh Pandit", "Christopher Mei", "Ali Behrooz", "Ramin Mirjalili", "Dave Stronks", "Alexander Fix", "Dmitri Model"], "categories": ["cs.CV", "physics.optics"], "fields_of_study": ["Computer Science", "Physics"], "published_date": "2025-11-06", "url": "https://arxiv.org/abs/2511.04652", "pdf_url": "https://arxiv.org/pdf/2511.04652v1", "arxiv_id": "2511.04652", "doi": "10.48550/arXiv.2511.04652", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3609}
{"id": "2b8e739be17d954bee543be8b12ee72c414cf3d7de582af80fc3fea2784ea881", "sources": ["arxiv", "semantic_scholar"], "title": "Human Mesh Modeling for Anny Body", "abstract": "Parametric body models provide the structural basis for many human-centric tasks, yet existing models often rely on costly 3D scans and learned shape spaces that are proprietary and demographically narrow. We introduce Anny, a simple, fully differentiable, and scan-free human body model grounded in anthropometric knowledge from the MakeHuman community. Anny defines a continuous, interpretable shape space, where phenotype parameters (e.g. gender, age, height, weight) control blendshapes spanning a wide range of human forms--across ages (from infants to elders), body types, and proportions. Calibrated using WHO population statistics, it provides realistic and demographically grounded human shape variation within a single unified model. Thanks to its openness and semantic control, Anny serves as a versatile foundation for 3D human modeling--supporting millimeter-accurate scan fitting, controlled synthetic data generation, and Human Mesh Recovery (HMR). We further introduce Anny-One, a collection of 800k photorealistic images generated with Anny, showing that despite its simplicity, HMR models trained with Anny can match the performance of those trained with scan-based body models. The Anny body model and its code are released under the Apache 2.0 license, making Anny an accessible foundation for human-centric 3D modeling.", "authors": ["Romain Brégier", "Guénolé Fiche", "Laura Bravo-Sánchez", "Thomas Lucas", "Matthieu Armando", "Philippe Weinzaepfel", "Grégory Rogez", "Fabien Baradel"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-05", "url": "https://arxiv.org/abs/2511.03589", "pdf_url": "https://arxiv.org/pdf/2511.03589v2", "arxiv_id": "2511.03589", "doi": "10.48550/arXiv.2511.03589", "citation_count": 7, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/naver/anny", "venue": "arXiv.org", "quality_score": 0.556}
{"id": "ede1f48e9942d46264fb042bbf0f8ee397dab10fd9e5d6584387fb2688ea5290", "sources": ["arxiv", "semantic_scholar"], "title": "Merging Continual Pretraining Models for Domain-Specialized LLMs: A Case Study in Finance", "abstract": "While LLMs excel at general tasks, they struggle in specialized domains like finance, requiring diverse skills in domain knowledge, mathematical reasoning, and multilingual processing. Merging domain-specific Continual Pre-training (CPT) \"experts\" offers a practical alternative to costly and unstable multi-skill training. However, unlike established Supervised Fine-Tuning (SFT) model-based merging, CPT model merging remains largely unexplored. We address this gap by creating financial LLMs from experts in finance, math, and Japanese. We propose a three-stage evaluation focusing on knowledge recovery, complementarity, and emergence, and assess three merging methods (Task Arithmetic, TIES, and DARE-TIES) on a comprehensive financial benchmark curated from 18 tasks across 8 established datasets. Results show that merging an expert with its base model recovers general knowledge lost during CPT, while merging experts improves performance and can yield emergent cross-domain skills. Among the methods, Task Arithmetic performs strongly but is hyperparameter-sensitive, whereas TIES is more robust. Our findings also suggest that while model similarity correlates with merging success, emergent skills depend on more complex factors. This work presents the first foundational analysis of CPT model merging, establishing a principled framework and providing clear guidance for building multi-skill LLMs from existing assets.", "authors": ["Kentaro Ueda", "François Portet", "Hirohiko Suwa", "Keiichi Yasumoto"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-04", "url": "https://arxiv.org/abs/2511.02451", "pdf_url": "https://arxiv.org/pdf/2511.02451v1", "arxiv_id": "2511.02451", "doi": "10.48550/arXiv.2511.02451", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3586}
{"id": "47ac3b07fb2499582dfff7d95e28fd2e07aeca76184b577ccddb163070e425cd", "sources": ["arxiv", "semantic_scholar"], "title": "Model Merging with Functional Dual Anchors", "abstract": "Model merging is an efficient post-training strategy for integrating knowledge from multiple finetuned checkpoints of a shared foundation model. Existing methods operate in the parameter space, combining task vectors to mitigate conflicts, but remain constrained by parameter inconsistencies. We propose Functional Dual Anchors (FDAs), a framework that instead models the input-representation space. FDAs are synthetic inputs whose induced gradients align with task vectors, capturing task-specific functional shifts relative to the pretrained model. This perspective bridges joint multi-task training and post-hoc merging, offering both robustness and flexibility. We further introduce a principled initialization scheme and show that FDAs are complementary to parameter-space model merging. Comprehensive experiments demonstrate the effectiveness of FDAs in model merging.", "authors": ["Kexuan Shi", "Yandong Wen", "Weiyang Liu"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-24", "url": "https://arxiv.org/abs/2510.21223", "pdf_url": "https://arxiv.org/pdf/2510.21223v1", "arxiv_id": "2510.21223", "doi": "10.48550/arXiv.2510.21223", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.346}
{"id": "d64abf2f883e84edbd93e6ffcbb7004e901cb32b901d2419651bb932038ea02e", "sources": ["arxiv", "semantic_scholar"], "title": "Capability Ceilings in Autoregressive Language Models: Empirical Evidence from Knowledge-Intensive Tasks", "abstract": "We document empirical capability ceilings in decoder-only autoregressive language models across knowledge-intensive tasks. Systematic evaluation of OPT and Pythia model families (70M-30B parameters, spanning 240 times scaling) reveals that knowledge retrieval tasks show negligible accuracy improvement despite smooth loss reduction. On MMLU mathematics benchmarks, accuracy remains flat at 19-20% (below 25% random chance) across all scales while cross-entropy loss decreases by 31%. In contrast, procedural tasks like arithmetic show conventional scaling where both metrics improve together. Attention intervention experiments reveal high sensitivity to perturbation: swapping attention patterns between models causes catastrophic performance collapse (complete accuracy loss) rather than graceful degradation. These measurements have immediate engineering implications: for knowledge-intensive applications using OPT and Pythia architectures, parameter scaling beyond 1-2B offers minimal accuracy gains despite continued loss improvement. Our findings quantify capability-specific scaling failures in these model families to inform resource allocation decisions. Whether these patterns reflect fundamental constraints of decoder-only architectures or implementation-specific limitations remains an open question requiring investigation across diverse architectural approaches.", "authors": ["Javier Marín"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-23", "url": "https://arxiv.org/abs/2510.21866", "pdf_url": "https://arxiv.org/pdf/2510.21866v1", "arxiv_id": "2510.21866", "doi": "10.48550/arXiv.2510.21866", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3449}
{"id": "a371c173fa7ec3b87f668b9288e7f8108ee524d52fa892cfb2a1314b6cc860b2", "sources": ["arxiv", "semantic_scholar"], "title": "Adapting Multilingual Models to Code-Mixed Tasks via Model Merging", "abstract": "We study model merging as a practical alternative to conventional adaptation strategies for code-mixed NLP. Starting from a multilingual base model, we: (i) perform continued pre-training (CPT) on unlabeled code-mixed text to obtain an adapted checkpoint, (ii) merge checkpoint with the base model, and (iii) fine-tune (FT) on the downstream task data. We evaluate our approach for sentence classification (sentiment and hate speech) task in English-Hindi (En-Hi) and English-Spanish (En-Es) using XLM-R and Llama-3.2-1B models. Our results show that merged models consistently outperform full fine-tuning and CPT->FT. We observe gains of 2--5 points in F1 over full fine-tuning and ~1-2 points over CPT->FT, indicating that unlabeled data is leveraged more effectively via merging than via CPT alone. Zero-/few-shot prompting with larger LLMs (e.g., Llama-3.3-70B) lags behind fine-tuned and merged checkpoints, underscoring limits of in-context learning for code-mixed inputs. We further test cross-pair transfer by training on En-Hi and evaluating on En-Ta and En-Ml: merged checkpoints transfer more strongly than monolingual-English baselines (e.g., TV/TIES variants reaching 0.65-0.68 F1 vs 0.61-0.63 for full fine-tuning), suggesting that code-mixed knowledge is a more reliable substrate for low-resource pairs. We conclude with adaptation recipes matched to common data regimes (labeled only; labeled+unlabeled; transfer-only) and discuss limitations and scaling considerations for broader tasks and larger models.", "authors": ["Prashant Kodali", "Vaishnavi Shivkumar", "Swarang Joshi", "Monojit Choudhary", "Ponnurangam Kumaraguru", "Manish Shrivastava"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-22", "url": "https://arxiv.org/abs/2510.19782", "pdf_url": "https://arxiv.org/pdf/2510.19782v2", "arxiv_id": "2510.19782", "doi": "10.1145/3799830.3799852", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "ACM IKDD Conference on Data Science", "quality_score": 0.3438}
{"id": "9fb80a9fc4b1c2f256df22e7efd05db1e411dd68e6ca20e8419a143f9b10ee83", "sources": ["arxiv", "semantic_scholar"], "title": "MIN-Merging: Merge the Important Neurons for Model Merging", "abstract": "Recent advances in deep learning have led to a surge of open-source models across diverse domains. While model merging offers a promising way to combine their strengths, existing approaches often suffer from parameter conflicts that degrade performance on domain-specific tasks. We propose MIN-Merging, a router-based framework that selectively merges the most important neurons to reduce such conflicts. Extensive experiments on Computer Vision(CV) and Natural Language Processing(NLP) benchmarks show that MIN-Merging achieves consistent gains on in-domain tasks while retaining the generalization ability of pretrained models on out-of-domain tasks. These results highlight its effectiveness as a practical solution to the parameter conflict problem in model merging.", "authors": ["Yunfei Liang"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-18", "url": "https://arxiv.org/abs/2510.17890", "pdf_url": "https://arxiv.org/pdf/2510.17890v2", "arxiv_id": "2510.17890", "doi": "10.48550/arXiv.2510.17890", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5242}
{"id": "61d7f5e295921aede5c87def6017b2dbbf9707fc1eb37a9d371646715d099d3f", "sources": ["arxiv", "semantic_scholar"], "title": "Six Proofs of Interpolation for the Modal Logic K", "abstract": "In this chapter, we present six different proofs of Craig interpolation for the modal logic K, each using a different set of techniques (model-theoretic, proof-theoretic, syntactic, automata-theoretic, using quasi-models, and algebraic). We compare the pros and cons of each proof technique.", "authors": ["Nick Bezhanishvili", "Balder ten Cate", "Rosalie Iemhoff"], "categories": ["cs.LO"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-18", "url": "https://arxiv.org/abs/2510.16398", "pdf_url": "https://arxiv.org/pdf/2510.16398v2", "arxiv_id": "2510.16398", "doi": "10.48550/arXiv.2510.16398", "citation_count": 3, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3392}
{"id": "ea237e51da3028808d7c6564f34fc44f384387262adc7728917615cdd4572b17", "sources": ["arxiv", "semantic_scholar"], "title": "Purifying Task Vectors in Knowledge-Aware Subspace for Model Merging", "abstract": "Model merging aims to integrate task-specific abilities from individually fine-tuned models into a single model without extra training. In recent model merging methods, task vector has become a fundamental building block, as it can encapsulate the residual information from finetuning. However, the merged model often suffers from notable performance degradation due to the conflicts caused by task-irrelevant redundancy in task vectors. Existing efforts in overcoming redundancy by randomly dropping elements in the parameter space involves randomness and lacks knowledge awareness. To address these challenges, in this study, we propose Purifying TAsk Vectors (PAVE) in knowledge-aware subspace. Concretely, we sample some training examples from each task, and feed them into their corresponding fine-tuned models to acquire the covariance matrices before linear layers. We then perform a context-oriented singular value decomposition, which accentuates the weight components most relevant to the target knowledge. As a result, we can split fine-tuned model weights into task-relevant and redundant components in the knowledge-aware subspace, and purify the task vector by pruning the redundant components. To induce fair pruning efforts across models, we further introduce a spectral rank allocation strategy by optimizing a normalized activated pruning error. The task vector purification by our method as a plug-and-play scheme is applicable across various task vector-based merging methods to improve their performance. In experiments, we demonstrate the effectiveness of PAVE across a diverse set of merging methods, tasks, and model architectures.", "authors": ["Bang An", "Yibo Yang", "Philip Torr", "Bernard Ghanem"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-16", "url": "https://arxiv.org/abs/2510.14697", "pdf_url": "https://arxiv.org/pdf/2510.14697v1", "arxiv_id": "2510.14697", "doi": "10.48550/arXiv.2510.14697", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3369}
{"id": "e93a218744a82719bfd5f42659b24f6e33d27f1db5eff5733af69c263fbfdf5e", "sources": ["arxiv", "semantic_scholar"], "title": "Weight Weaving: Parameter Pooling for Data-Free Model Merging", "abstract": "Model merging provides a cost-effective and data-efficient combination of specialized deep neural networks through parameter integration. This technique leverages expert models across downstream tasks without requiring retraining. Most model merging approaches critically depend on scaling hyper-parameters $λ$, which weight each model's contribution globally or individually. Principled approaches for setting scaling factors without accessing any data (data-free) are scarce, often leading researchers to tune $λ$ using privileged data from the evaluation set, which is obviously unfeasible in practice. To address this limitation, we introduce Weight Weaving, a plug-and-play technique that pools model weights across $λ$ values search space using user-defined pooling functions, such as averaging, random selection, or even existing model merging methods. Our method demonstrates high modularity, imposing minimal constraints on the search space. It operates orthogonally to existing model merging methods and eliminates evaluation data requirements. We validate Weight Weaving across three ViT variants in three experimental setups: vision multi-task learning, vision continual learning, and domain generalization. Our method consistently improves the performance of several model merging methods, achieving average accuracy gains of up to 15.9 percentage points in a data-free setting.", "authors": ["Levy Chaves", "Eduardo Valle", "Sandra Avila"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-15", "url": "https://arxiv.org/abs/2510.13921", "pdf_url": "https://arxiv.org/pdf/2510.13921v1", "arxiv_id": "2510.13921", "doi": "10.48550/arXiv.2510.13921", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3357}
{"id": "a051feaeee8ec3dc8cfce5862561ca5ab9ab10e9ee8eef1aef83f8a9e14bac46", "sources": ["arxiv", "semantic_scholar"], "title": "Towards Reversible Model Merging For Low-rank Weights", "abstract": "Model merging aims to combine multiple fine-tuned models into a single set of weights that performs well across all source tasks. While prior work has shown that merging can approximate the performance of individual fine-tuned models for each task, it largely overlooks scenarios where models are compressed into low-rank representations, either through low-rank adaptation (LoRA) or post-training singular value decomposition (SVD). We first demonstrate that applying conventional merging methods to low-rank weights leads to severe performance degradation in the merged model. Motivated by this phenomenon, we propose a fundamentally different approach: instead of collapsing all adapters into one set of weights, we construct a compact basis (e.g., an equivalent of holding two or more models) from which original task-specific models can be recovered via linear combination. This reframes merging as generating a reconstruction-capable model space rather than producing a single merged model. Crucially, this allows us to ``revert'' to each individual model when needed, recognizing that no merged model can consistently outperform one specialized for its task. Building on this insight, we introduce our method, Reversible Model Merging (RMM), an efficient, data-free, and flexible method that provides a closed-form solution for selecting the optimal basis of model weights and task-specific coefficients for linear combination. Extensive experiments across diverse datasets and model scales demonstrate that RMM consistently outperforms existing merging approaches, preserving the performance of low-rank compressed models by a significant margin.", "authors": ["Mohammadsajad Alipour", "Mohammad Mohammadi Amiri"], "categories": ["cs.LG", "cs.AI", "cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-15", "url": "https://arxiv.org/abs/2510.14163", "pdf_url": "https://arxiv.org/pdf/2510.14163v1", "arxiv_id": "2510.14163", "doi": "10.48550/arXiv.2510.14163", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3357}
{"id": "835188ea4527da006fce8aa6733da3907118dcab5104a73393936a9774a2740c", "sources": ["arxiv", "semantic_scholar"], "title": "Revisiting Model Interpolation for Efficient Reasoning", "abstract": "Model merging, typically on Instruct and Thinking models, has shown remarkable performance for efficient reasoning. In this paper, we systematically revisit the simplest merging method that interpolates two weights directly. Particularly, we observe that model interpolation follows a three-stage evolutionary paradigm with distinct behaviors on the reasoning trajectory. These dynamics provide a principled guide for navigating the performance-cost trade-off. Empirical results demonstrate that a strategically interpolated model surprisingly surpasses sophisticated model merging baselines on both efficiency and effectiveness. We further validate our findings with extensive ablation studies on model layers, modules, and decoding strategies. Ultimately, this work demystifies model interpolation and offers a practical framework for crafting models with precisely targeted reasoning capabilities. Code is available at \\href{https://github.com/wutaiqiang/MI}{Github}.", "authors": ["Taiqiang Wu", "Runming Yang", "Tao Liu", "Jiahao Wang", "Ngai Wong"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-13", "url": "https://arxiv.org/abs/2510.10977", "pdf_url": "https://arxiv.org/pdf/2510.10977v2", "arxiv_id": "2510.10977", "doi": "10.48550/arXiv.2510.10977", "citation_count": 8, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/wutaiqiang/MI}{Github}", "venue": "arXiv.org", "quality_score": 0.5153}
{"id": "5c3a2e7ffe28bf88ce70621a4e881e8fabf60d2ccabc3f4141273d4e1fe55bd4", "sources": ["arxiv", "semantic_scholar"], "title": "Generative World Modelling for Humanoids: 1X World Model Challenge Technical Report", "abstract": "World models are a powerful paradigm in AI and robotics, enabling agents to reason about the future by predicting visual observations or compact latent states. The 1X World Model Challenge introduces an open-source benchmark of real-world humanoid interaction, with two complementary tracks: sampling, focused on forecasting future image frames, and compression, focused on predicting future discrete latent codes. For the sampling track, we adapt the video generation foundation model Wan-2.2 TI2V-5B to video-state-conditioned future frame prediction. We condition the video generation on robot states using AdaLN-Zero, and further post-train the model using LoRA. For the compression track, we train a Spatio-Temporal Transformer model from scratch. Our models achieve 23.0 dB PSNR in the sampling task and a Top-500 CE of 6.6386 in the compression task, securing 1st place in both challenges.", "authors": ["Riccardo Mereu", "Aidan Scannell", "Yuxin Hou", "Yi Zhao", "Aditya Jitta", "Antonio Dominguez", "Luigi Acerbi", "Amos Storkey", "Paul Chang"], "categories": ["cs.LG", "cs.AI", "cs.RO"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-08", "url": "https://arxiv.org/abs/2510.07092", "pdf_url": "https://arxiv.org/pdf/2510.07092v1", "arxiv_id": "2510.07092", "doi": "10.48550/arXiv.2510.07092", "citation_count": 5, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5065}
{"id": "62633c6c252beb0b4aae766bf3ee7e91a0ca9afd882ccc08bc6fbd47d488b4cf", "sources": ["arxiv", "semantic_scholar"], "title": "Learning to Interpret Weight Differences in Language Models", "abstract": "Finetuning (pretrained) language models is a standard approach for updating their internal parametric knowledge and specializing them to new tasks and domains. However, the corresponding model weight changes (\"weight diffs\") are not generally interpretable. While inspecting the finetuning dataset can give a sense of how the model might have changed, these datasets are often not publicly available or are too large to work with directly. Towards the goal of comprehensively understanding weight diffs in natural language, we introduce Diff Interpretation Tuning (DIT), a method that trains models to describe their own finetuning-induced modifications. Our approach uses synthetic, labeled weight diffs to train a DIT-adapter, which can be applied to a compatible finetuned model to make it describe how it has changed. We demonstrate in two proof-of-concept settings (reporting hidden behaviors and summarizing finetuned knowledge) that our method enables models to describe their finetuning-induced modifications using accurate natural language descriptions.", "authors": ["Avichal Goel", "Yoon Kim", "Nir Shavit", "Tony T. Wang"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-06", "url": "https://arxiv.org/abs/2510.05092", "pdf_url": "https://arxiv.org/pdf/2510.05092v4", "arxiv_id": "2510.05092", "doi": "10.48550/arXiv.2510.05092", "citation_count": 4, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/Aviously/diff-interpretation-tuning", "venue": "arXiv.org", "quality_score": 0.5029}
{"id": "3b76ef3b115fa16c3219e74f11e59e2a8d6a1633ec81fe8e8b09af681f8731da", "sources": ["arxiv", "semantic_scholar"], "title": "How does the optimizer implicitly bias the model merging loss landscape?", "abstract": "Model merging combines independent solutions with different capabilities into a single one while maintaining the same inference cost. Two popular approaches are linear interpolation, which simply averages multiple model weights, and task arithmetic, which combines task vectors obtained by the difference between finetuned and base models. While useful in practice, what properties make merging effective are poorly understood. This paper explores how the optimization dynamics affect the loss landscape geometry and its impact on merging success. We show that a single quantity -- the effective noise scale -- unifies the impact of different optimizer components on model merging. Across architectures and datasets, merging success is a non-monotonic function of the effective noise scale, with a distinct optimum. Decomposing this quantity, we find that larger learning rates, stronger weight decay, smaller batch sizes, and data augmentation all independently modulate the effective noise scale and exhibit the same qualitative trend. Unlike prior work connecting optimizer noise to the flatness or generalization of individual minima, we show that it also affects the global loss landscape, predicting when independently trained solutions can be successfully merged. Our findings broaden the understanding of how optimization shapes the loss landscape geometry and its consequences for model merging, suggesting that training dynamics could be further manipulated to improve model merging.", "authors": ["Chenxiang Zhang", "Alexander Theus", "Damien Teney", "Antonio Orvieto", "Jun Pang", "Sjouke Mauw"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-06", "url": "https://arxiv.org/abs/2510.04686", "pdf_url": "https://arxiv.org/pdf/2510.04686v2", "arxiv_id": "2510.04686", "doi": "10.48550/arXiv.2510.04686", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3254}
{"id": "997e330af66c5a44c97b0a05f0f23917a51a2ac6948203afda211d53201b5e66", "sources": ["arxiv", "semantic_scholar"], "title": "Model Merging to Maintain Language-Only Performance in Developmentally Plausible Multimodal Models", "abstract": "State-of-the-art vision-and-language models consist of many parameters and learn from enormous datasets, surpassing the amounts of linguistic data that children are exposed to as they acquire a language. This paper presents our approach to the multimodal track of the BabyLM challenge addressing this discrepancy. We develop language-only and multimodal models in low-resource settings using developmentally plausible datasets, with our multimodal models outperforming previous BabyLM baselines. One finding in the multimodal language model literature is that these models tend to underperform in \\textit{language-only} tasks. Therefore, we focus on maintaining language-only abilities in multimodal models. To this end, we experiment with \\textit{model merging}, where we fuse the parameters of multimodal models with those of language-only models using weighted linear interpolation. Our results corroborate the findings that multimodal models underperform in language-only benchmarks that focus on grammar, and model merging with text-only models can help alleviate this problem to some extent, while maintaining multimodal performance.", "authors": ["Ece Takmaz", "Lisa Bylinina", "Jakub Dotlacil"], "categories": ["cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-02", "url": "https://arxiv.org/abs/2510.01845", "pdf_url": "https://arxiv.org/pdf/2510.01845v1", "arxiv_id": "2510.01845", "doi": "10.48550/arXiv.2510.01845", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.2042}
{"id": "ff4d8cdedd83ba2eee2343afecc168bd61ede19bd718f9bf92032736ce5de566", "sources": ["arxiv", "semantic_scholar"], "title": "Expert Merging: Model Merging with Unsupervised Expert Alignment and Importance-Guided Layer Chunking", "abstract": "Model merging, which combines multiple domain-specialized experts into a single model, offers a practical path to endow Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) with broad capabilities without the cost of joint training or serving many models. However, training-free methods rely on hand-tuned coefficients, whereas training-based methods primarily align parameters rather than downstream task behavior and typically treat all layers uniformly, ignoring inter-layer heterogeneity. We introduce Expert Merging, a training-light method that learns a small set of layer-wise coefficients using only unlabeled calibration data. The coefficients are optimized to explicitly align the merged model's hidden states and logits with those of the corresponding experts, with a coefficient regularizer for stability and task-weighted losses for controllable trade-offs. To capture inter-layer variation, Expert Merging++ augments this design with importance-guided chunking: a normalized layer-importance metric, derived from learned coefficients, task-vector magnitudes, and parameter counts, allocates more chunk-wise coefficients to high-importance layers while keeping low-importance layers lightweight. The result is a label-free, parameter-efficient, and scalable approach to multi-expert model merging across LLMs and MLLMs. Across MLLM backbones (InternVL and Qwen2-VL) and the LLM backbone (Mistral), our method surpasses strong training-free and training-based merging baselines, with Expert Merging++ delivering further gains and, in some cases, even exceeding supervised Mixture Training. The source code is available at https://github.com/Littleor/ExpertMerging.", "authors": ["Dengming Zhang", "Xiaowen Ma", "Zhenliang Ni", "Zhenkai Wu", "Han Shu", "Xin Jiang", "Xinghao Chen"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-30", "url": "https://arxiv.org/abs/2509.25712", "pdf_url": "https://arxiv.org/pdf/2509.25712v1", "arxiv_id": "2509.25712", "doi": "10.48550/arXiv.2509.25712", "citation_count": 3, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/Littleor/ExpertMerging", "venue": "arXiv.org", "quality_score": 0.4923}
{"id": "706e6d13695e4ff8967a649b854a3aa373aadaf2db59288a82ed0a30ed778c62", "sources": ["arxiv", "semantic_scholar"], "title": "Real-Aware Residual Model Merging for Deepfake Detection", "abstract": "Deepfake generators evolve quickly, making exhaustive data collection and repeated retraining impractical. We argue that model merging is a natural fit for deepfake detection: unlike generic multi-task settings with disjoint labels, deepfake specialists share the same binary decision and differ in generator-specific artifacts. Empirically, we show that simple weight averaging preserves Real representations while attenuating Fake-specific cues. Building upon these findings, we propose Real-aware Residual Model Merging (R$^2$M), a training-free parameter-space merging framework. R$^2$M estimates a shared Real component via a low-rank factorization of task vectors, decomposes each specialist into a Real-aligned part and a Fake residual, denoises residuals with layerwise rank truncation, and aggregates them with per-task norm matching to prevent any single generator from dominating. A concise rationale explains why a simple head suffices: the Real component induces a common separation direction in feature space, while truncated residuals contribute only minor off-axis variations. Across in-distribution, cross-dataset, and unseen-dataset, R$^2$M outperforms joint training and other merging baselines. Importantly, R$^2$M is also composable: when a new forgery family appears, we fine-tune one specialist and re-merge, eliminating the need for retraining.", "authors": ["Jinhee Park", "Guisik Kim", "Choongsang Cho", "Junseok Kwon"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-29", "url": "https://arxiv.org/abs/2509.24367", "pdf_url": "https://arxiv.org/pdf/2509.24367v1", "arxiv_id": "2509.24367", "doi": "10.48550/arXiv.2509.24367", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3174}
{"id": "e3f8c953a710ac55788ec40a66f9883244f849389b979cc33ba0d81b88ed1248", "sources": ["arxiv", "semantic_scholar"], "title": "Toward a Holistic Approach to Continual Model Merging", "abstract": "We present a holistic framework for Continual Model Merging (CMM) that intervenes at three critical stages: pre-merging, during merging, and post-merging-to address two fundamental challenges in continual learning. In particular, conventional approaches either maintain a growing list of per-domain task vectors, leading to scalability issues or rely solely on weight-space merging when old data is inaccessible, thereby losing crucial functional information. Our method overcomes these limitations by first fine-tuning the main model within its tangent space on domain-specific data; this linearization amplifies per-task weight disentanglement, effectively mitigating across-task interference. During merging, we leverage functional information from available optimizer states beyond mere parameter averages to avoid the need to revisit old data. Finally, a post-merging correction aligns the representation discrepancy between pre- and post-merged models, reducing bias and enhancing overall performance-all while operating under constant memory constraints without accessing historical data. Extensive experiments on standard class-incremental and domain-incremental benchmarks demonstrate that our approach not only achieves competitive performance but also provides a scalable and efficient solution to the catastrophic forgetting problem.", "authors": ["Hoang Phan", "Sungmin Cha", "Tung Lam Tran", "Qi Lei"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-28", "url": "https://arxiv.org/abs/2509.23592", "pdf_url": "https://arxiv.org/pdf/2509.23592v2", "arxiv_id": "2509.23592", "doi": "10.48550/arXiv.2509.23592", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3162}
{"id": "f02301f94db25329dd707c705833fec9269475a568d340038867a368e632983b", "sources": ["arxiv", "semantic_scholar"], "title": "Effect of Model Merging in Domain-Specific Ad-hoc Retrieval", "abstract": "In this study, we evaluate the effect of model merging in ad-hoc retrieval tasks. Model merging is a technique that combines the diverse characteristics of multiple models. We hypothesized that applying model merging to domain-specific ad-hoc retrieval tasks could improve retrieval effectiveness. To verify this hypothesis, we merged the weights of a source retrieval model and a domain-specific (non-retrieval) model using a linear interpolation approach. A key advantage of our approach is that it requires no additional fine-tuning of the models. We conducted two experiments each in the medical and Japanese domains. The first compared the merged model with the source retrieval model, and the second compared it with a LoRA fine-tuned model under both full and limited data settings for model construction. The experimental results indicate that model merging has the potential to produce more effective domain-specific retrieval models than the source retrieval model, and may serve as a practical alternative to LoRA fine-tuning, particularly when only a limited amount of data is available.", "authors": ["Taiga Sasaki", "Takehiro Yamamoto", "Hiroaki Ohshima", "Sumio Fujita"], "categories": ["cs.IR"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-26", "url": "https://arxiv.org/abs/2509.21966", "pdf_url": "https://arxiv.org/pdf/2509.21966v1", "arxiv_id": "2509.21966", "doi": "10.48550/arXiv.2509.21966", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Conference on Information and Knowledge Management", "quality_score": 0.314}
{"id": "cf2c512905980d1c3f6ad6817e9978f7029f269ed84e598d638d76d81c59a3c4", "sources": ["arxiv", "semantic_scholar"], "title": "The Thinking Spectrum: An Empirical Study of Tunable Reasoning in LLMs through Model Merging", "abstract": "The growing demand for large language models (LLMs) with tunable reasoning capabilities in many real-world applications highlights a critical need for methods that can efficiently produce a spectrum of models balancing reasoning depth and computational cost. Model merging has emerged as a promising, training-free technique to address this challenge by arithmetically combining the weights of a general-purpose model with a specialized reasoning model. While various merging techniques exist, their potential to create a spectrum of models with fine-grained control over reasoning abilities remains largely unexplored. This work presents a large-scale empirical study evaluating a range of model merging techniques across multiple reasoning benchmarks. We systematically vary merging strengths to construct accuracy-efficiency curves, providing the first comprehensive view of the tunable performance landscape. Our findings reveal that model merging offers an effective and controllable method for calibrating the trade-off between reasoning accuracy and token efficiency, even when parent models have highly divergent weight spaces. Crucially, we identify instances of Pareto Improvement, where a merged model achieves both higher accuracy and lower token consumption than one of its parents. Our study provides the first comprehensive analysis of this tunable space, offering practical guidelines for creating LLMs with specific reasoning profiles to meet diverse application demands.", "authors": ["Xiaochong Lan", "Yu Zheng", "Shiteng Cao", "Yong Li"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-26", "url": "https://arxiv.org/abs/2509.22034", "pdf_url": "https://arxiv.org/pdf/2509.22034v2", "arxiv_id": "2509.22034", "doi": "10.48550/arXiv.2509.22034", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.314}
{"id": "9d2d43618fdaed60a29ed7c9bee403a7f777028b884d1275471f3443edab2099", "sources": ["arxiv", "semantic_scholar"], "title": "Faster, Smaller, and Smarter: Task-Aware Expert Merging for Online MoE Inference", "abstract": "Sparse Mixture of Experts (SMoE) has become a preferred architecture for scaling Transformer capacity without increasing computational cost, as it activates only a small subset of experts for each input. However, deploying such an approach for \\textit{online inference} remains challenging due to the large size of a full SMoE model and the complexity of expert routing, especially in resource-constrained edge networks. Moreover, during the online inference, task information is often unavailable, making the task-level routing error-prone. In this work, we propose a novel tree-structured adaptive neural bandit router, \\texttt{Tanbr}, to enable efficient and reliable online MoE inference. Instead of relying on explicit task tags, \\texttt{Tanbr} estimates the task distribution over time from historical data and uses it to guide task-aware expert merging within a given pre-trained MoE. To handle the large continuous space of merging weights, \\texttt{Tanbr} employs a binary tree to progressively partition the space and generate finer candidate weights. It then applies a neural bandit to learn the non-linear mapping from merging weight to model performance and decides optimal expert merging. We prove that \\texttt{Tanbr} achieves a sublinear regret bound of {\\small $\\mathcal{O}(\\sqrt{T} \\log(T))$} over {\\small $T$} rounds, despite operating over a continuous decision space, matching regret bounds compared to existing methods. Extensive experiments show that \\texttt{Tanbr} reduces inference latency by at least {\\small $45\\%$} and memory usage by up to {\\small $25\\%$}, while maintaining a high accuracy compared to many state-of-the-art methods.", "authors": ["Ziyi Han", "Xutong Liu", "Ruiting Zhou", "Xiangxiang Dai", "John C. S. Lui"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-24", "url": "https://arxiv.org/abs/2509.19781", "pdf_url": "https://arxiv.org/pdf/2509.19781v2", "arxiv_id": "2509.19781", "doi": "10.48550/arXiv.2509.19781", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3117}
{"id": "8b813b03d2f7b4baa2791b0c192f2ebaecc06cc4efdb9755a05de020dee0f606", "sources": ["arxiv", "semantic_scholar"], "title": "Rethinking Pulmonary Embolism Segmentation: A Study of Current Approaches and Challenges with an Open Weight Model", "abstract": "Pulmonary Embolism (PE) is a life-threatening condition for which accurate and timely detection is critical to patient care. However, our systematic study of PE segmentation algorithms reveals concerning limitations in the current state of research. Challenges such as small and inconsistent datasets, a lack of reproducible baselines, and limited comparative evaluation across models are hindering progress in the field. In this study, we curated a densely annotated dataset comprising 490 CTPA scans, each from a unique patient (430 for training and 60 for testing). We evaluated nine widely used segmentation architectures, including both CNN- and ViT-based models, in 2D and 3D configurations, using mean Dice Similarity Coefficient (mDSC) and Average Symmetric Surface Distance (ASSD) as evaluation metrics. Furthermore, the highest-performing model was evaluated on a public dataset without fine-tuning and achieved reasonable generalization performance. Our results show that: (1) a 3D U-Net with ResNet encoding blocks remains a highly effective architecture for PE segmentation; (2) 3D models consistently outperform their 2D counterparts; (3) across all architectures, when trained and evaluated on the same datasets, model error patterns are highly consistent; and (4) distal emboli remain particularly challenging due to both task complexity and the scarcity of high-quality datasets, highlighting the need for datasets with more comprehensive and consistent distal PE coverage. To promote research reproducibility, the architecture and pretrained weights of our best-performing model are publicly available at https://github.com/mazurowski-lab/PulmonaryEmbolismSegmentation", "authors": ["Yixin Zhang", "Ryan Chamberlain", "Lawrence Ngo", "Kevin Kramer", "Maciej A. Mazurowski"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science", "Medicine"], "published_date": "2025-09-22", "url": "https://arxiv.org/abs/2509.18308", "pdf_url": "https://arxiv.org/pdf/2509.18308v3", "arxiv_id": "2509.18308", "doi": "10.1007/s10278-026-01958-4", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/mazurowski-lab/PulmonaryEmbolismSegmentation", "venue": null, "quality_score": 0.3656}
{"id": "6f2df63da9702e87c9697cf21c44e60f2e6679d16de35a1c1d5238382582760e", "sources": ["arxiv", "semantic_scholar"], "title": "Accurate and Efficient Low-Rank Model Merging in Core Space", "abstract": "In this paper, we address the challenges associated with merging low-rank adaptations of large neural networks. With the rise of parameter-efficient adaptation techniques, such as Low-Rank Adaptation (LoRA), model fine-tuning has become more accessible. While fine-tuning models with LoRA is highly efficient, existing merging methods often sacrifice this efficiency by merging fully-sized weight matrices. We propose the Core Space merging framework, which enables the merging of LoRA-adapted models within a common alignment basis, thereby preserving the efficiency of low-rank adaptation while substantially improving accuracy across tasks. We further provide a formal proof that projection into Core Space ensures no loss of information and provide a complexity analysis showing the efficiency gains. Extensive empirical results demonstrate that Core Space significantly improves existing merging techniques and achieves state-of-the-art results on both vision and language tasks while utilizing a fraction of the computational resources. Codebase is available at https://github.com/apanariello4/core-space-merging.", "authors": ["Aniello Panariello", "Daniel Marczak", "Simone Magistri", "Angelo Porrello", "Bartłomiej Twardowski", "Andrew D. Bagdanov", "Simone Calderara", "Joost van de Weijer"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-22", "url": "https://arxiv.org/abs/2509.17786", "pdf_url": "https://arxiv.org/pdf/2509.17786v4", "arxiv_id": "2509.17786", "doi": "10.48550/arXiv.2509.17786", "citation_count": 23, "influential_citation_count": 3, "has_code": true, "code_url": "https://github.com/apanariello4/core-space-merging", "venue": "arXiv.org", "quality_score": 0.4781}
{"id": "b76c8bb81b39bf0934453d616b03d8b41bf7c40686eda78ea89217baaea8184f", "sources": ["arxiv", "semantic_scholar"], "title": "Black-box Model Merging for Language-Model-as-a-Service with Massive Model Repositories", "abstract": "Model merging refers to the process of integrating multiple distinct models into a unified model that preserves and combines the strengths and capabilities of the individual models. Most existing approaches rely on task vectors to combine models, typically under the assumption that model parameters are accessible. However, for extremely large language models (LLMs) such as GPT-4, which are often provided solely as black-box services through API interfaces (Language-Model-as-a-Service), model weights are not available to end users. This presents a significant challenge, which we refer to as black-box model merging (BMM) with massive LLMs. To address this challenge, we propose a derivative-free optimization framework based on the evolutionary algorithm (Evo-Merging) that enables effective model merging using only inference-time API queries. Our method consists of two key components: (1) sparsity-based denoising, designed to identify and filter out irrelevant or redundant information across models, and (2) sign-aware scaling, which dynamically computes optimal combination weights for the relevant models based on their performance. We also provide a formal justification, along with a theoretical analysis, for our asymmetric sparsification. Extensive experimental evaluations demonstrate that our approach achieves state-of-the-art results on a range of tasks, significantly outperforming existing strong baselines.", "authors": ["Shilian Chen", "Jie Zhou", "Tianyu Huai", "Yujiang Lu", "Junsong Li", "Bihao Zhan", "Qianjun Pan", "Yutao Yang", "Xin Li", "Qin Chen", "Hang Yan", "Liang He"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-16", "url": "https://arxiv.org/abs/2509.12951", "pdf_url": "https://arxiv.org/pdf/2509.12951v1", "arxiv_id": "2509.12951", "doi": "10.48550/arXiv.2509.12951", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3025}
{"id": "183464d9b604dd0adfa6a09c92607209d7ac86b30879d6a847971e7c7f9327a5", "sources": ["arxiv", "semantic_scholar"], "title": "DivMerge: A divergence-based model merging method for multi-tasking", "abstract": "Multi-task learning (MTL) is often achieved by merging datasets before fine-tuning, but the growing availability of fine-tuned models has led to new approaches such as model merging via task arithmetic. A major challenge in this setting is task interference, which worsens as the number of tasks increases. We propose a method that merges models trained on different tasks into a single model, maintaining strong performance across all tasks. Our approach leverages Jensen-Shannon divergence to guide the merging process without requiring additional labelled data, and automatically balances task importance. Unlike existing methods, our approach remains robust as the number of tasks grows and consistently outperforms prior work.", "authors": ["Brahim Touayouch", "Loïc Fosse", "Géraldine Damnati", "Gwénolé Lecorvé"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-02", "url": "https://arxiv.org/abs/2509.02108", "pdf_url": "https://arxiv.org/pdf/2509.02108v3", "arxiv_id": "2509.02108", "doi": "10.48550/arXiv.2509.02108", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference of the European Chapter of the Association for Computational Linguistics", "quality_score": 0.2865}
{"id": "04497e0888382fce4ed84aca7e8147feef7438a339c0c1986a3244f3662ca01a", "sources": ["arxiv", "semantic_scholar"], "title": "Surrogate Benchmarks for Model Merging Optimization", "abstract": "Model merging techniques aim to integrate the abilities of multiple models into a single model. Most model merging techniques have hyperparameters, and their setting affects the performance of the merged model. Because several existing works show that tuning hyperparameters in model merging can enhance the merging outcome, developing hyperparameter optimization algorithms for model merging is a promising direction. However, its optimization process is computationally expensive, particularly in merging LLMs. In this work, we develop surrogate benchmarks for optimization of the merging hyperparameters to realize algorithm development and performance comparison at low cost. We define two search spaces and collect data samples to construct surrogate models to predict the performance of a merged model from a hyperparameter. We demonstrate that our benchmarks can predict the performance of merged models well and simulate optimization algorithm behaviors.", "authors": ["Rio Akizuki", "Yuya Kudo", "Nozomu Yoshinari", "Yoichi Hirose", "Toshiyuki Nishimoto", "Kento Uchida", "Shinichi Shirakawa"], "categories": ["cs.LG", "cs.AI", "cs.NE"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-02", "url": "https://arxiv.org/abs/2509.02555", "pdf_url": "https://arxiv.org/pdf/2509.02555v2", "arxiv_id": "2509.02555", "doi": "10.48550/arXiv.2509.02555", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/shiralab/SMM-Bench", "venue": "arXiv.org", "quality_score": 0.4427}
{"id": "d99b85035d1745f2684b8524f7268e7496036796d91ec49c110c6939bb315f96", "sources": ["arxiv", "semantic_scholar"], "title": "Rethinking Layer-wise Model Merging through Chain of Merges", "abstract": "Fine-tuning pretrained models has become a standard pathway to achieve state-of-the-art performance across a wide range of domains, leading to a proliferation of task-specific model variants. As the number of such specialized models increases, merging them into a unified model without retraining has become a critical challenge. Existing merging techniques operate at the level of individual layers, thereby overlooking the inter-layer dependencies inherent in deep networks. We show that this simplification leads to distributional mismatches, particularly in methods that rely on intermediate activations, as changes in early layers are not properly propagated to downstream layers during merging. We identify these mismatches as a form of internal covariate shift, comparable to the phenomenon encountered in the initial phases of neural networks training. To address this, we propose Chain of Merges (CoM), a layer-wise merging procedure that sequentially merges weights across layers while sequentially updating activation statistics. By explicitly accounting for inter-layer interactions, CoM mitigates covariate shift and produces a coherent merged model through a series of conditionally optimal updates. Experiments on standard benchmarks demonstrate that CoM achieves state-of-the-art performance.", "authors": ["Pietro Buzzega", "Riccardo Salami", "Angelo Porrello", "Simone Calderara"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-29", "url": "https://arxiv.org/abs/2508.21421", "pdf_url": "https://arxiv.org/pdf/2508.21421v3", "arxiv_id": "2508.21421", "doi": "10.48550/arXiv.2508.21421", "citation_count": 3, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2819}
{"id": "ef30e256e10766fe8ad7bc235618fa670297fdb4f580a76be4f4f413efe99a9b", "sources": ["arxiv", "semantic_scholar"], "title": "PSO-Merging: Merging Models Based on Particle Swarm Optimization", "abstract": "Model merging has emerged as an efficient strategy for constructing multitask models by integrating the strengths of multiple available expert models, thereby reducing the need to fine-tune a pre-trained model for all the tasks from scratch. Existing data-independent methods struggle with performance limitations due to the lack of data-driven guidance. Data-driven approaches also face key challenges: gradient-based methods are computationally expensive, limiting their practicality for merging large expert models, whereas existing gradient-free methods often fail to achieve satisfactory results within a limited number of optimization steps. To address these limitations, this paper introduces PSO-Merging, a novel data-driven merging method based on the Particle Swarm Optimization (PSO). In this approach, we initialize the particle swarm with a pre-trained model, expert models, and sparsified expert models. We then perform multiple iterations, with the final global best particle serving as the merged model. Experimental results on different language models show that PSO-Merging generally outperforms baseline merging methods, offering a more efficient and scalable solution for model merging.", "authors": ["Kehao Zhang", "Shaolei Zhang", "Yang Feng"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-27", "url": "https://arxiv.org/abs/2508.19839", "pdf_url": "https://arxiv.org/pdf/2508.19839v1", "arxiv_id": "2508.19839", "doi": "10.48550/arXiv.2508.19839", "citation_count": 1, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2796}
{"id": "781898fd0ab0221eca26b21f114aa682ec3dcc6b60351e5bbe0add3845e2c27e", "sources": ["arxiv", "semantic_scholar"], "title": "From Interpolating Formulas to Separating Languages and Back Again", "abstract": "Traditionally, research on Craig interpolation is concerned with (a) establishing the Craig interpolation property (CIP) of a logic saying that every valid implication in the logic has a Craig interpolant and (b) designing algorithms that extract Craig interpolants from proofs. Logics that lack the CIP are regarded as `pathological' and excluded from consideration. In this chapter, we survey variations and generalisations of traditional Craig interpolation. First, we consider Craig interpolants for implications in logics without the CIP, focusing on the decidability and complexity of deciding their existence. We then generalise interpolation by looking for Craig interpolants in languages L' that can be weaker than the language L of the given implication. Thus, do not only we restrict the non-logical symbols of Craig interpolants but also the logical ones. The resulting L/L'-interpolation problem generalises L/L'-definability, the question whether an L-formula is equivalent to some L'-formula. After that, we move from logical languages to formal languages where interpolation disguises itself as separation: given two disjoint languages in a class C, does there exist a separating language in a smaller class C'? This question is particularly well-studied in the case when the input languages are regular and the separating language is first-order definable. Finally, we connect the different research strands by showing how the decidability of the separation problem for regular languages can be used to prove the decidability of Craig interpolant existence for linear temporal logic LTL.", "authors": ["Agi Kurucz", "Frank Wolter", "Michael Zakharyaschev"], "categories": ["cs.LO"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-18", "url": "https://arxiv.org/abs/2508.12805", "pdf_url": "https://arxiv.org/pdf/2508.12805v2", "arxiv_id": "2508.12805", "doi": "10.48550/arXiv.2508.12805", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2693}
{"id": "51bc6b1b51a96838d8e53256d7b3cbd899aef10a1ccc0d466f5faafa5e77314e", "sources": ["arxiv", "semantic_scholar"], "title": "Interpolation in Classical Propositional Logic", "abstract": "We introduce Craig interpolation and related notions such as uniform interpolation, Beth definability, and theory decomposition in classical propositional logic. We present four approaches to computing interpolants: via quantifier elimination, from formulas in disjunctive normal form, and by extraction from resolution or tableau refutations. We close with a discussion of the size of interpolants and links to circuit complexity.", "authors": ["Patrick Koopmann", "Christoph Wernhard", "Frank Wolter"], "categories": ["cs.LO"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-15", "url": "https://arxiv.org/abs/2508.11449", "pdf_url": "https://arxiv.org/pdf/2508.11449v2", "arxiv_id": "2508.11449", "doi": "10.48550/arXiv.2508.11449", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2658}
{"id": "addf0a390f50f57377da991e64de30fc6d929a0359aa0e91c2ed8f926329e29e", "sources": ["arxiv", "semantic_scholar"], "title": "Objective Soups: Multilingual Multi-Task Modeling for Speech Processing", "abstract": "Training a single model for multilingual, multi-task speech processing (MSP) is severely hampered by conflicting objectives between tasks like speech recognition and translation. While multi-objective optimization (MOO) aims to align gradient updates, its effectiveness diminishes as the number of tasks grows, making it difficult to find a common descent direction. This raises a fundamental question: should highly conflicting objectives be optimized jointly or separated into a hierarchical structure? To address this question, this paper investigates three multi-objective MSP formulations, which we refer to as \\textbf{objective soup recipes}. These formulations apply multi-objective optimization at different optimization levels to mitigate potential conflicts among all objectives. To ensure efficiency, we introduce a lightweight layer-selection mechanism that computes the conflict-avoiding gradient using only the most problematic layers, minimizing computational and memory overhead. Extensive experiments on CoVoST v2, LibriSpeech, and AISHELL-1 reveal that a bi-level recipe separating recognition and translation tasks consistently outperforms standard flat optimization. Our work demonstrates that hierarchical MOO is a more effective and scalable approach for building state-of-the-art MSP models. Our code has been released at https://github.com/afmsaif/Objective_Soups.", "authors": ["A F M Saif", "Lisha Chen", "Xiaodong Cui", "Songtao Lu", "Brian Kingsbury", "Tianyi Chen"], "categories": ["eess.AS", "cs.LG", "math.OC", "stat.ML"], "fields_of_study": ["Engineering", "Computer Science", "Mathematics"], "published_date": "2025-08-12", "url": "https://arxiv.org/abs/2508.09228", "pdf_url": "https://arxiv.org/pdf/2508.09228v1", "arxiv_id": "2508.09228", "doi": "10.48550/arXiv.2508.09228", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/afmsaif/Objective_Soups", "venue": "arXiv.org", "quality_score": 0.4055}
{"id": "0a7a3696879748e3fe5aee5a4e2f67c0cda97eac732062f3f73cea5f1f6f8d10", "sources": ["arxiv", "semantic_scholar"], "title": "RCP-Merging: Merging Long Chain-of-Thought Models with Domain-Specific Models by Considering Reasoning Capability as Prior", "abstract": "Large Language Models (LLMs) with long chain-of-thought (CoT) capability, termed Reasoning Models, demonstrate superior intricate problem-solving abilities through multi-step long CoT reasoning. To create a dual-capability model with long CoT capability and domain-specific knowledge without substantial computational and data costs, model merging emerges as a highly resource-efficient method. However, significant challenges lie in merging domain-specific LLMs with long CoT ones since nowadays merging methods suffer from reasoning capability degradation, even gibberish output and output collapse. To overcome this, we introduce RCP-Merging: Merging Long Chain-of-Thought Models with Domain-Specific Models by Considering Reasoning Capability as Prior, a novel merging framework designed to integrate domain-specific LLMs with long CoT capability, meanwhile maintaining model performance in the original domain. Treating reasoning model weights as foundational prior, our method utilizes a reasoning capability indicator to preserve core long CoT capability model weights while selectively merging essential domain-specific weights. We conducted extensive experiments on Qwen2.5-7B, Llama3.1-8B, and Qwen2.5-1.5B models in BioMedicine and Finance domains. Our results show that RCP-Merging successfully merges a reasoning model with domain-specific ones, improving domain task performance by 9.5% and 9.2% over state-of-the-art methods, without significantly harming the original long CoT reasoning capability.", "authors": ["Junyao Yang", "Jianwei Wang", "Huiping Zhuang", "Cen Chen", "Ziqian Zeng"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-05", "url": "https://arxiv.org/abs/2508.03140", "pdf_url": "https://arxiv.org/pdf/2508.03140v2", "arxiv_id": "2508.03140", "doi": "10.48550/arXiv.2508.03140", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.2544}
{"id": "f7eff10a9d298701e2c61eff719cefb411b1bca0928a197a613867102476d11c", "sources": ["arxiv", "semantic_scholar"], "title": "RegMean++: Enhancing Effectiveness and Generalization of Regression Mean for Model Merging", "abstract": "Regression Mean (RegMean), an approach that formulates model merging as a linear regression problem, aims to find the optimal weights for each linear layer in the merged model by minimizing the discrepancy in predictions between the merged and candidate models. RegMean provides a precise closed-form solution for the merging problem; therefore, it offers explainability and computational efficiency. However, RegMean merges each linear layer independently, overlooking how the features and information in earlier layers propagate through deeper layers and influence the final predictions of the merged model. Here, we introduce RegMean++, a simple yet effective alternative to RegMean, that explicitly incorporates both intra-layer and cross-layer dependencies between merged models' layers into RegMean's objective. By accounting for these dependencies, RegMean++ better captures the behaviors of the merged model. Extensive experiments demonstrate that RegMean++ consistently outperforms RegMean across diverse settings, including in-domain (ID) and out-of-domain (OOD) generalization, sequential merging, large-scale tasks, and robustness under several types of distribution shifts. Furthermore, RegMean++ achieves competitive performance across diverse settings compared to various advanced model merging methods.", "authors": ["The-Hai Nguyen", "Dang Huu-Tien", "Takeshi Suzuki", "Le-Minh Nguyen"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-05", "url": "https://arxiv.org/abs/2508.03121", "pdf_url": "https://arxiv.org/pdf/2508.03121v3", "arxiv_id": "2508.03121", "doi": "10.48550/arXiv.2508.03121", "citation_count": 8, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2544}
{"id": "0b21d163f7dbdfd1b80a778cf4fda4124d774b8a5076ae7fad5a23fa9ff39607", "sources": ["arxiv", "semantic_scholar"], "title": "RouteMark: A Fingerprint for Intellectual Property Attribution in Routing-based Model Merging", "abstract": "Model merging via Mixture-of-Experts (MoE) has emerged as a scalable solution for consolidating multiple task-specific models into a unified sparse architecture, where each expert is derived from a model fine-tuned on a distinct task. While effective for multi-task integration, this paradigm introduces a critical yet underexplored challenge: how to attribute and protect the intellectual property (IP) of individual experts after merging. We propose RouteMark, a framework for IP protection in merged MoE models through the design of expert routing fingerprints. Our key insight is that task-specific experts exhibit stable and distinctive routing behaviors under probing inputs. To capture these patterns, we construct expert-level fingerprints using two complementary statistics: the Routing Score Fingerprint (RSF), quantifying the intensity of expert activation, and the Routing Preference Fingerprint (RPF), characterizing the input distribution that preferentially activates each expert. These fingerprints are reproducible, task-discriminative, and lightweight to construct. For attribution and tampering detection, we introduce a similarity-based matching algorithm that compares expert fingerprints between a suspect and a reference (victim) model. Extensive experiments across diverse tasks and CLIP-based MoE architectures show that RouteMark consistently yields high similarity for reused experts and clear separation from unrelated ones. Moreover, it remains robust against both structural tampering (expert replacement, addition, deletion) and parametric tampering (fine-tuning, pruning, permutation), outperforming weight- and activation-based baseliness. Our work lays the foundation for RouteMark as a practical and broadly applicable framework for IP verification in MoE-based model merging.", "authors": ["Xin He", "Junxi Shen", "Zhenheng Tang", "Xiaowen Chu", "Bo Li", "Ivor W. Tsang", "Yew-Soon Ong"], "categories": ["cs.CR", "cs.AI", "cs.ET", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-03", "url": "https://arxiv.org/abs/2508.01784", "pdf_url": "https://arxiv.org/pdf/2508.01784v1", "arxiv_id": "2508.01784", "doi": "10.48550/arXiv.2508.01784", "citation_count": 4, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2521}
{"id": "0cc404fb911e8986e5d97f551f12f36192344ee1ab65a8906c1ecd6ee67bf953", "sources": ["arxiv", "semantic_scholar"], "title": "DisTaC: Conditioning Task Vectors via Distillation for Robust Model Merging", "abstract": "Model merging has emerged as an efficient and flexible paradigm for multi-task learning, with numerous methods being proposed in recent years. However, these state-of-the-art techniques are typically evaluated on benchmark suites that are highly favorable to model merging, and their robustness in more realistic settings remains largely unexplored. In this work, we first investigate the vulnerabilities of model-merging methods and pinpoint the source-model characteristics that critically underlie them. Specifically, we identify two factors that are particularly harmful to the merging process: (1) disparities in task vector norms, and (2) the low confidence of the source models. To address this issue, we propose DisTaC (Distillation for Task vector Conditioning), a novel method that pre-conditions these problematic task vectors before the merge. DisTaC leverages knowledge distillation to adjust a task vector's norm and increase source-model confidence while preserving its essential task-specific knowledge. Our extensive experiments demonstrate that by pre-conditioning task vectors with DisTaC, state-of-the-art merging techniques can successfully integrate models exhibiting the harmful traits -- where they would otherwise fail -- achieving significant performance gains.", "authors": ["Kotaro Yoshida", "Yuji Naraki", "Takafumi Horie", "Ryotaro Shimizu", "Hiroki Naganuma"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-02", "url": "https://arxiv.org/abs/2508.01148", "pdf_url": "https://arxiv.org/pdf/2508.01148v3", "arxiv_id": "2508.01148", "doi": "10.48550/arXiv.2508.01148", "citation_count": 2, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/katoro8989/DisTaC", "venue": "arXiv.org", "quality_score": 0.3878}
{"id": "83b45e3cd611a03f51592b2c1af3c9b3d6ae794376e94baec8ab3b719f8695e1", "sources": ["arxiv", "semantic_scholar"], "title": "Accurate and Consistent Graph Model Generation from Text with Large Language Models", "abstract": "Graph model generation from natural language description is an important task with many applications in software engineering. With the rise of large language models (LLMs), there is a growing interest in using LLMs for graph model generation. Nevertheless, LLM-based graph model generation typically produces partially correct models that suffer from three main issues: (1) syntax violations: the generated model may not adhere to the syntax defined by its metamodel, (2) constraint inconsistencies: the structure of the model might not conform to some domain-specific constraints, and (3) inaccuracy: due to the inherent uncertainty in LLMs, the models can include inaccurate, hallucinated elements. While the first issue is often addressed through techniques such as constraint decoding or filtering, the latter two remain largely unaddressed. Motivated by recent self-consistency approaches in LLMs, we propose a novel abstraction-concretization framework that enhances the consistency and quality of generated graph models by considering multiple outputs from an LLM. Our approach first constructs a probabilistic partial model that aggregates all candidate outputs and then refines this partial model into the most appropriate concrete model that satisfies all constraints. We evaluate our framework on several popular open-source and closed-source LLMs using diverse datasets for model generation tasks. The results demonstrate that our approach significantly improves both the consistency and quality of the generated graph models.", "authors": ["Boqi Chen", "Ou Wei", "Bingzhou Zheng", "Gunter Mussbacher"], "categories": ["cs.SE", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-01", "url": "https://arxiv.org/abs/2508.00255", "pdf_url": "https://arxiv.org/pdf/2508.00255v1", "arxiv_id": "2508.00255", "doi": "10.1109/MODELS67397.2025.00018", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "ACM/IEEE International Conference on Model Driven Engineering Languages and Systems", "quality_score": 0.386}
{"id": "7e5290ff3fb7735a80b4c7c261d5f826ec2f369f14fd6f4d096ac6e39b2fb30b", "sources": ["arxiv", "semantic_scholar"], "title": "Evaluating Contrast Localizer for Identifying Causal Units in Social & Mathematical Tasks in Language Models", "abstract": "This work adapts a neuroscientific contrast localizer to pinpoint causally relevant units for Theory of Mind (ToM) and mathematical reasoning tasks in large language models (LLMs) and vision-language models (VLMs). Across 11 LLMs and 5 VLMs ranging in size from 3B to 90B parameters, we localize top-activated units using contrastive stimulus sets and assess their causal role via targeted ablations. We compare the effect of lesioning functionally selected units against low-activation and randomly selected units on downstream accuracy across established ToM and mathematical benchmarks. Contrary to expectations, low-activation units sometimes produced larger performance drops than the highly activated ones, and units derived from the mathematical localizer often impaired ToM performance more than those from the ToM localizer. These findings call into question the causal relevance of contrast-based localizers and highlight the need for broader stimulus sets and more accurately capture task-specific units.", "authors": ["Yassine Jamaa", "Badr AlKhamissi", "Satrajit Ghosh", "Martin Schrimpf"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-31", "url": "https://arxiv.org/abs/2508.08276", "pdf_url": "https://arxiv.org/pdf/2508.08276v3", "arxiv_id": "2508.08276", "doi": "10.48550/arXiv.2508.08276", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2486}
{"id": "b6d95981fc87e2deaa8211adf4e2c5f97db6d78bbbea4ffa97dfb627ad1caca3", "sources": ["arxiv", "semantic_scholar"], "title": "Forgetting of task-specific knowledge in model merging-based continual learning", "abstract": "This paper investigates the linear merging of models in the context of continual learning (CL). Using controlled visual cues in computer vision experiments, we demonstrate that merging largely preserves or enhances shared knowledge, while unshared task-specific knowledge rapidly degrades. We further find that merging models from an incremental training process consistently outperforms merging models trained in parallel.", "authors": ["Timm Hess", "Gido M van de Ven", "Tinne Tuytelaars"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-31", "url": "https://arxiv.org/abs/2507.23311", "pdf_url": "https://arxiv.org/pdf/2507.23311v1", "arxiv_id": "2507.23311", "doi": "10.48550/arXiv.2507.23311", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2486}
{"id": "ceb554a3dc20154386b31b8ddecfbebb2f37977d4821b584ad045c082f29e44f", "sources": ["arxiv", "semantic_scholar"], "title": "Interpolation in model theory", "abstract": "We bring an abstract model theory perspective to interpolation. We ask, what is the role of interpolation in the study of extensions of first order logic, such as infinitary logics, generalized quantifiers and higher order logics? The abstract model theory approach reveals the basic connections between various interpolation properties in isolation, on their own, as well as with respect to other model theoretic properties, such as compactness.", "authors": ["Jouko Väänänen"], "categories": ["math.LO"], "fields_of_study": ["Mathematics"], "published_date": "2025-07-25", "url": "https://arxiv.org/abs/2507.19097", "pdf_url": "https://arxiv.org/pdf/2507.19097v1", "arxiv_id": "2507.19097", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1539}
{"id": "5a77cbe83f702330e5774cc9aebc17ebb83ec6a0449af3cedb15d8e1a78f40ee", "sources": ["arxiv", "semantic_scholar"], "title": "Look the Other Way: Designing 'Positive' Molecules with Negative Data via Task Arithmetic", "abstract": "The scarcity of molecules with desirable properties (i.e., `positive' molecules) is an inherent bottleneck for generative molecule design. To sidestep such obstacle, here we propose molecular task arithmetic: training a model on diverse and abundant negative examples to learn 'property directions' - without accessing any positively labeled data - and moving models in the opposite property directions to generate positive molecules. When analyzed on 33 design experiments with distinct molecular entities (small molecules, proteins), model architectures, and scales, molecular task arithmetic generated more diverse and successful designs than models trained on positive molecules in general. Moreover, we employed molecular task arithmetic in dual-objective and few-shot design tasks. We find that molecular task arithmetic can consistently increase the diversity of designs while maintaining desirable complex design properties, such as good docking scores to a protein. With its simplicity, data efficiency, and performance, molecular task arithmetic bears the potential to become the de facto transfer learning strategy for de novo molecule design.", "authors": ["Rıza Özçelik", "Sarah de Ruiter", "Francesca Grisoni"], "categories": ["cs.LG", "physics.chem-ph", "q-bio.BM"], "fields_of_study": ["Computer Science", "Physics", "Biology"], "published_date": "2025-07-23", "url": "https://arxiv.org/abs/2507.17876", "pdf_url": "https://arxiv.org/pdf/2507.17876v2", "arxiv_id": "2507.17876", "doi": "10.48550/arXiv.2507.17876", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2395}
{"id": "522330362bb045b81e4998f0d12924a1785603d4a12d3c1b009c42f30ee36f18", "sources": ["arxiv", "semantic_scholar"], "title": "RegCL: Continual Adaptation of Segment Anything Model via Model Merging", "abstract": "To address the performance limitations of the Segment Anything Model (SAM) in specific domains, existing works primarily adopt adapter-based one-step adaptation paradigms. However, some of these methods are specific developed for specific domains. If used on other domains may lead to performance degradation. This issue of catastrophic forgetting severely limits the model's scalability. To address this issue, this paper proposes RegCL, a novel non-replay continual learning (CL) framework designed for efficient multi-domain knowledge integration through model merging. Specifically, RegCL incorporates the model merging algorithm into the continual learning paradigm by merging the parameters of SAM's adaptation modules (e.g., LoRA modules) trained on different domains. The merging process is guided by weight optimization, which minimizes prediction discrepancies between the merged model and each of the domain-specific models. RegCL effectively consolidates multi-domain knowledge while maintaining parameter efficiency, i.e., the model size remains constant regardless of the number of tasks, and no historical data storage is required. Experimental results demonstrate that RegCL achieves favorable continual learning performance across multiple downstream datasets, validating its effectiveness in dynamic scenarios.", "authors": ["Yuan-Chen Shu", "Zhiwei Lin", "Yongtao Wang"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-16", "url": "https://arxiv.org/abs/2507.12297", "pdf_url": "https://arxiv.org/pdf/2507.12297v1", "arxiv_id": "2507.12297", "doi": "10.48550/arXiv.2507.12297", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2315}
{"id": "32c7897506408bb9ab694e8fd4183dfbd7b51986d015e2f0b88fb6635b946f6f", "sources": ["arxiv", "semantic_scholar"], "title": "On the Surprising Effectiveness of a Single Global Merging in Decentralized Learning", "abstract": "Decentralized learning provides a scalable alternative to parameter-server-based training, yet its performance is often hindered by limited peer-to-peer communication. In this paper, we study how communication should be scheduled over time, including determining when and how frequently devices synchronize. Counterintuitive empirical results show that concentrating communication budgets in the later stages of decentralized training remarkably improves global test performance. Surprisingly, we uncover that fully connected communication at the final step, implemented by a single global merging, can significantly improve the performance of decentralized learning under high data heterogeneity. Our theoretical contributions, which explain these phenomena, are the first to establish that the globally merged model of decentralized SGD can match the convergence rate of parallel SGD. Technically, we reinterpret part of the discrepancy among local models, which were previously considered as detrimental noise, as constructive components essential for matching this rate. This work provides evidence that decentralized learning is able to generalize under high data heterogeneity and limited communication, while offering broad new avenues for model merging research.", "authors": ["Tongtian Zhu", "Tianyu Zhang", "Mingze Wang", "Zhanpeng Zhou", "Can Wang"], "categories": ["cs.LG", "cs.DC", "cs.MA", "stat.ML"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2025-07-09", "url": "https://arxiv.org/abs/2507.06542", "pdf_url": "https://arxiv.org/pdf/2507.06542v4", "arxiv_id": "2507.06542", "doi": null, "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "ICLR 2026 (Oral Presentation)", "quality_score": 0.2234}
{"id": "02f72d2be06681e91f07407176c622b245008a164d03f4fb01b51c521cad6138", "sources": ["arxiv", "semantic_scholar"], "title": "Semantic Frame Interpolation", "abstract": "Generating intermediate video content of varying lengths based on given first and last frames, along with text prompt information, offers significant research and application potential. However, traditional frame interpolation tasks primarily focus on scenarios with a small number of frames, no text control, and minimal differences between the first and last frames. Recent community developers have utilized large video models represented by Wan to endow frame-to-frame capabilities. However, these models can only generate a fixed number of frames and often fail to produce satisfactory results for certain frame lengths, while this setting lacks a clear official definition and a well-established benchmark. In this paper, we first propose a new practical Semantic Frame Interpolation (SFI) task from the perspective of academic definition, which covers the above two settings and supports inference at multiple frame rates. To achieve this goal, we propose a novel SemFi model building upon Wan2.1, which incorporates a Mixture-of-LoRA module to ensure the generation of high-consistency content that aligns with control conditions across various frame length limitations. Furthermore, we propose SFI-300K, the first general-purpose dataset and benchmark specifically designed for SFI. To support this, we collect and process data from the perspective of SFI, carefully designing evaluation metrics and methods to assess the model's performance across multiple dimensions, encompassing image and video, and various aspects, including consistency and diversity. Through extensive experiments on SFI-300K, we demonstrate that our method is particularly well-suited to meet the requirements of the SFI task.", "authors": ["Yijia Hong", "Jiangning Zhang", "Ran Yi", "Yuji Wang", "Weijian Cao", "Xiaobin Hu", "Zhucun Xue", "Yabiao Wang", "Chengjie Wang", "Lizhuang Ma"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-07", "url": "https://arxiv.org/abs/2507.05173", "pdf_url": "https://arxiv.org/pdf/2507.05173v1", "arxiv_id": "2507.05173", "doi": "10.48550/arXiv.2507.05173", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/hyj542682306/Semantic-Frame-Interpolation", "venue": "arXiv.org", "quality_score": 0.3418}
{"id": "497dadcc7a4cbc249933184439eae1cfbebcb66aaf089adcffee8550aec6b12f", "sources": ["arxiv", "semantic_scholar"], "title": "Interpolation with Automated First-Order Reasoning", "abstract": "We consider interpolation from the viewpoint of fully automated theorem proving in first-order logic as a general core technique for mechanized knowledge processing. For Craig interpolation, our focus is on the two-stage approach, where first an essentially propositional ground interpolant is calculated that is then lifted to a quantified first-order formula. We discuss two possibilities to obtain a ground interpolant from a proof: with clausal tableaux, and with resolution. Established preprocessing techniques for first-order proving can also be applied for Craig interpolation if they are restricted in specific ways. Equality encodings from automated reasoning justify strengthened variations of Craig interpolation. Contributions to Craig interpolation that emerged from automated reasoning include variations for logics used in databases and logic programming. As an approach to uniform interpolation we introduce second-order quantifier elimination with examples and describe the basic algorithms DLS and SCAN.", "authors": ["Christoph Wernhard"], "categories": ["cs.LO"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-02", "url": "https://arxiv.org/abs/2507.01577", "pdf_url": "https://arxiv.org/pdf/2507.01577v2", "arxiv_id": "2507.01577", "doi": "10.48550/arXiv.2507.01577", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2154}
{"id": "7151cb13702d3afbca18da7a946462ae831aae808ba97ef078e398e7090c64bb", "sources": ["arxiv", "semantic_scholar"], "title": "SE-Merging: A Self-Enhanced Approach for Dynamic Model Merging", "abstract": "Model merging has gained increasing attention due to its intriguing property: interpolating the parameters of different task-specific fine-tuned models leads to multi-task abilities. However, despite its empirical success, the underlying mechanisms of model merging remain poorly understood. In this work, we delve into the mechanism behind model merging from a representation perspective. Our analysis reveals that model merging achieves multi-task abilities through two key capabilities: i) distinguishing samples from different tasks, and ii) adapting to the corresponding expert model for each sample. These two capabilities allow the merged model to retain task-specific expertise, enabling efficient multi-task adaptation. Building on these insights, we propose \\texttt{SE-Merging}, a self-enhanced model merging framework that leverages these two characteristics to dynamically identify the corresponding task for each sample and then adaptively rescales the merging coefficients to further enhance task-specific expertise in the merged model. Notably, \\texttt{SE-Merging} achieves dynamic model merging without additional training. Extensive experiments demonstrate that \\texttt{SE-Merging} achieves significant performance improvements while remaining compatible with existing model merging techniques.", "authors": ["Zijun Chen", "Zhanpeng Zhou", "Bo Zhang", "Weinan Zhang", "Xi Sun", "Junchi Yan"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-22", "url": "https://arxiv.org/abs/2506.18135", "pdf_url": "https://arxiv.org/pdf/2506.18135v1", "arxiv_id": "2506.18135", "doi": "10.1109/IJCNN64981.2025.11229001", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "IEEE International Joint Conference on Neural Network", "quality_score": 0.204}
{"id": "2c5f0168c2da2b116c0d68ce7e1046f81dbbfc3a5db290755f9e6971f17a5f34", "sources": ["arxiv", "semantic_scholar"], "title": "Subspace-Boosted Model Merging", "abstract": "Model merging enables the combination of multiple specialized expert models into a single model capable of performing multiple tasks. However, the benefits of merging an increasing amount of specialized experts generally lead to diminishing returns and reduced overall performance gains. In this work, we empirically and theoretically analyze this limitation, proving that for Task Arithmetic-based methods, as more experts are merged, the common information dominates the task-specific information, leading to inevitable rank collapse. To mitigate this issue, we introduce Subspace Boosting, which operates on the singular value decomposed task vector space and maintains task vector ranks. Subspace Boosting raises merging efficacy for up to 20 experts by large margins of more than 10% when evaluated on both vision and language benchmarks. Moreover, we propose employing Higher-Order Generalized Singular Value Decomposition to quantify task similarity, offering a new interpretable perspective on model merging. Code and models are available at https://github.com/ronskoro/Subspace-Boosting.", "authors": ["Ronald Skorobogat", "Karsten Roth", "Mariana-Iuliana Georgescu"], "categories": ["cs.LG", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-19", "url": "https://arxiv.org/abs/2506.16506", "pdf_url": "https://arxiv.org/pdf/2506.16506v3", "arxiv_id": "2506.16506", "doi": "10.48550/arXiv.2506.16506", "citation_count": 7, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/ronskoro/Subspace-Boosting", "venue": "arXiv.org", "quality_score": 0.3099}
{"id": "29e74e5cd5f69d616ea4622a5bfeddf3d9dcea5dd2daa51c953109bd3ee4ba50", "sources": ["arxiv", "semantic_scholar"], "title": "CALM: Consensus-Aware Localized Merging for Multi-Task Learning", "abstract": "Model merging aims to integrate the strengths of multiple fine-tuned models into a unified model while preserving task-specific capabilities. Existing methods, represented by task arithmetic, are typically classified into global- and local-aware methods. However, global-aware methods inevitably cause parameter interference, while local-aware methods struggle to maintain the effectiveness of task-specific details in the merged model. To address these limitations, we propose a Consensus-Aware Localized Merging (CALM) method which incorporates localized information aligned with global task consensus, ensuring its effectiveness post-merging. CALM consists of three key components: (1) class-balanced entropy minimization sampling, providing a more flexible and reliable way to leverage unsupervised data; (2) an efficient-aware framework, selecting a small set of tasks for sequential merging with high scalability; (3) a consensus-aware mask optimization, aligning localized binary masks with global task consensus and merging them conflict-free. Experiments demonstrate the superiority and robustness of our CALM, significantly outperforming existing methods and achieving performance close to traditional MTL.", "authors": ["Kunda Yan", "Min Zhang", "Sen Cui", "Zikun Qu", "Bo Jiang", "Feng Liu", "Changshui Zhang"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-16", "url": "https://arxiv.org/abs/2506.13406", "pdf_url": "https://arxiv.org/pdf/2506.13406v1", "arxiv_id": "2506.13406", "doi": "10.48550/arXiv.2506.13406", "citation_count": 7, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.2258}
{"id": "ded0edd16b631916585bc057280f1986c01eb16072caf7cf7f9c73fef2e2fd3d", "sources": ["arxiv", "semantic_scholar"], "title": "A correlation-permutation approach for speech-music encoders model merging", "abstract": "Creating a unified speech and music model requires expensive pre-training. Model merging can instead create an unified audio model with minimal computational expense. However, direct merging is challenging when the models are not aligned in the weight space. Motivated by Git Re-Basin, we introduce a correlation-permutation approach that aligns a music encoder's internal layers with a speech encoder. We extend previous work to the case of merging transformer layers. The method computes a permutation matrix that maximizes the model's features-wise cross-correlations layer by layer, enabling effective fusion of these otherwise disjoint models. The merged model retains speech capabilities through this method while significantly enhancing music performance, achieving an improvement of 14.83 points in average score compared to linear interpolation model merging. This work allows the creation of unified audio models from independently trained encoders.", "authors": ["Fabian Ritter-Gutierrez", "Yi-Cheng Lin", "Jeremy H. M Wong", "Hung-yi Lee", "Eng Siong Chng", "Nancy F. Chen"], "categories": ["cs.SD", "cs.AI", "eess.AS"], "fields_of_study": ["Computer Science", "Engineering"], "published_date": "2025-06-13", "url": "https://arxiv.org/abs/2506.11403", "pdf_url": "https://arxiv.org/pdf/2506.11403v1", "arxiv_id": "2506.11403", "doi": "10.1109/ASRU65441.2025.11433848", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Automatic Speech Recognition & Understanding", "quality_score": 0.1936}
{"id": "672a28c2e4b567190c7b80035791968725f50a819dd4d09c6a9ddaf0451d68cf", "sources": ["arxiv", "semantic_scholar"], "title": "Pisces: An Auto-regressive Foundation Model for Image Understanding and Generation", "abstract": "Recent advances in large language models (LLMs) have enabled multimodal foundation models to tackle both image understanding and generation within a unified framework. Despite these gains, unified models often underperform compared to specialized models in either task. A key challenge in developing unified models lies in the inherent differences between the visual features needed for image understanding versus generation, as well as the distinct training processes required for each modality. In this work, we introduce Pisces, an auto-regressive multimodal foundation model that addresses this challenge through a novel decoupled visual encoding architecture and tailored training techniques optimized for multimodal generation. Combined with meticulous data curation, pretraining, and finetuning, Pisces achieves competitive performance in both image understanding and image generation. We evaluate Pisces on over 20 public benchmarks for image understanding, where it demonstrates strong performance across a wide range of tasks. Additionally, on GenEval, a widely adopted benchmark for image generation, Pisces exhibits robust generative capabilities. Our extensive analysis reveals the synergistic relationship between image understanding and generation, and the benefits of using separate visual encoders, advancing the field of unified multimodal models.", "authors": ["Zhiyang Xu", "Jiuhai Chen", "Zhaojiang Lin", "Xichen Pan", "Lifu Huang", "Tianyi Zhou", "Madian Khabsa", "Qifan Wang", "Di Jin", "Michihiro Yasunaga", "Lili Yu", "Xi Victoria Lin", "Shaoliang Nie"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-12", "url": "https://arxiv.org/abs/2506.10395", "pdf_url": "https://arxiv.org/pdf/2506.10395v2", "arxiv_id": "2506.10395", "doi": "10.48550/arXiv.2506.10395", "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1945}
{"id": "444bae612e821193ddae6a8f586ba1e1bef88d40fcf3c012aecb98829ec94575", "sources": ["arxiv", "semantic_scholar"], "title": "Merging Smarter, Generalizing Better: Enhancing Model Merging on OOD Data", "abstract": "Multi-task learning (MTL) concurrently trains a model on diverse task datasets to exploit common features, thereby improving overall performance across the tasks. Recent studies have dedicated efforts to merging multiple independent model parameters into a unified model for MTL, thus circumventing the need for training data and expanding the scope of applicable scenarios of MTL. However, current approaches to model merging predominantly concentrate on enhancing performance within in-domain (ID) datasets, often overlooking their efficacy on out-of-domain (OOD) datasets. In this work, we proposed LwPTV (Layer-wise Pruning Task Vector) by building a saliency score, measuring the redundancy of parameters in task vectors. Designed in this way ours can achieve mask vector for each task and thus perform layer-wise pruning on the task vectors, only keeping the pre-trained model parameters at the corresponding layer in merged model. Owing to its flexibility, our method can be seamlessly integrated with most of existing model merging methods to improve their performance on OOD tasks. Extensive experiments demonstrate that the application of our method results in substantial enhancements in OOD performance while preserving the ability on ID tasks.", "authors": ["Bingjie Zhang", "Hongkang Li", "Changlong Shi", "Guowei Rong", "He Zhao", "Dongsheng Wang", "Dandan Guo", "Meng Wang"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-10", "url": "https://arxiv.org/abs/2506.09093", "pdf_url": "https://arxiv.org/pdf/2506.09093v2", "arxiv_id": "2506.09093", "doi": "10.48550/arXiv.2506.09093", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1902}
{"id": "9b1fd161d283862ada01138f9ef9fb3bd417ac8374c654f90f0a7662b17318ac", "sources": ["arxiv", "semantic_scholar"], "title": "StatsMerging: Statistics-Guided Model Merging via Task-Specific Teacher Distillation", "abstract": "Model merging has emerged as a promising solution to accommodate multiple large models within constrained memory budgets. We present StatsMerging, a novel lightweight learning-based model merging method guided by weight distribution statistics without requiring ground truth labels or test samples. StatsMerging offers three key advantages: (1) It uniquely leverages singular values from singular value decomposition (SVD) to capture task-specific weight distributions, serving as a proxy for task importance to guide task coefficient prediction; (2) It employs a lightweight learner StatsMergeLearner to model the weight distributions of task-specific pre-trained models, improving generalization and enhancing adaptation to unseen samples; (3) It introduces Task-Specific Teacher Distillation for merging vision models with heterogeneous architectures, a merging learning paradigm that avoids costly ground-truth labels by task-specific teacher distillation. Notably, we present two types of knowledge distillation, (a) distilling knowledge from task-specific models to StatsMergeLearner; and (b) distilling knowledge from models with heterogeneous architectures prior to merging. Extensive experiments across eight tasks demonstrate the effectiveness of StatsMerging. Our results show that StatsMerging outperforms state-of-the-art techniques in terms of overall accuracy, generalization to unseen tasks, and robustness to image quality variations.", "authors": ["Ranjith Merugu", "Bryan Bo Cao", "Shubham Jain"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-05", "url": "https://arxiv.org/abs/2506.04567", "pdf_url": "https://arxiv.org/pdf/2506.04567v1", "arxiv_id": "2506.04567", "doi": "10.48550/arXiv.2506.04567", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1845}
{"id": "b9f18dd1962a037d523bf353152749f5c23d8ccb18589b5d3d889914e40a47ce", "sources": ["arxiv", "semantic_scholar"], "title": "AceVFI: A Comprehensive Survey of Advances in Video Frame Interpolation", "abstract": "Video Frame Interpolation (VFI) is a core low-level vision task that synthesizes intermediate frames between existing ones while ensuring spatial and temporal coherence. Over the past decades, VFI methodologies have evolved from classical motion compensation-based approach to a wide spectrum of deep learning-based approaches, including kernel-, flow-, hybrid-, phase-, GAN-, Transformer-, Mamba-, and most recently, diffusion-based models. We introduce AceVFI, a comprehensive and up-to-date review of the VFI field, covering over 250 representative papers. We systematically categorize VFI methods based on their core design principles and architectural characteristics. Further, we classify them into two major learning paradigms: Center-Time Frame Interpolation (CTFI) and Arbitrary-Time Frame Interpolation (ATFI). We analyze key challenges in VFI, including large motion, occlusion, lighting variation, and non-linear motion. In addition, we review standard datasets, loss functions, evaluation metrics. We also explore VFI applications in other domains and highlight future research directions. This survey aims to serve as a valuable reference for researchers and practitioners seeking a thorough understanding of the modern VFI landscape.", "authors": ["Dahyeon Kye", "Changhyun Roh", "Sukhun Ko", "Chanho Eom", "Jihyong Oh"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-01", "url": "https://arxiv.org/abs/2506.01061", "pdf_url": "https://arxiv.org/pdf/2506.01061v3", "arxiv_id": "2506.01061", "doi": "10.1109/TCSVT.2026.3672288", "citation_count": 12, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/CMLab-Korea/Awesome-Video-Frame-Interpolation", "venue": "IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), 2026", "quality_score": 0.2785}
{"id": "2e08087d9f5941d3cc979dbdb9d3871a33e1579e77165ed1a26fd68fa2e6b1be", "sources": ["arxiv", "semantic_scholar"], "title": "WorldGym: World Model as An Environment for Policy Evaluation", "abstract": "Evaluating robot control policies is difficult: real-world testing is costly, and handcrafted simulators require manual effort to improve in realism and generality. We propose a world-model-based policy evaluation environment (WorldGym), an autoregressive, action-conditioned video generation model which serves as a proxy to real world environments. Policies are evaluated via Monte Carlo rollouts in the world model, with a vision-language model providing rewards. We evaluate a set of VLA-based real-robot policies in the world model using only initial frames from real robots, and show that policy success rates within the world model highly correlate with real-world success rates. Moreoever, we show that WorldGym is able to preserve relative policy rankings across different policy versions, sizes, and training checkpoints. Due to requiring only a single start frame as input, the world model further enables efficient evaluation of robot policies' generalization ability on novel tasks and environments. We find that modern VLA-based robot policies still struggle to distinguish object shapes and can become distracted by adversarial facades of objects. While generating highly realistic object interaction remains challenging, WorldGym faithfully emulates robot motions and offers a practical starting point for safe and reproducible policy evaluation before deployment.", "authors": ["Julian Quevedo", "Ansh Kumar Sharma", "Yixiang Sun", "Varad Suryavanshi", "Percy Liang", "Sherry Yang"], "categories": ["cs.RO", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-31", "url": "https://arxiv.org/abs/2506.00613", "pdf_url": "https://arxiv.org/pdf/2506.00613v3", "arxiv_id": "2506.00613", "doi": null, "citation_count": 26, "influential_citation_count": 5, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.3891}
{"id": "b79fc3861283032a0c41d77a1264245432b84f130045d1eca1fecc7db6a1cf2f", "sources": ["arxiv", "semantic_scholar"], "title": "On Fairness of Task Arithmetic: The Role of Task Vectors", "abstract": "Model editing techniques, particularly task arithmetic with task vectors, offer an efficient alternative to full fine-tuning by enabling direct parameter updates through simple arithmetic operations. While this approach promises substantial computational savings, its impact on fairness has remained largely unexplored -- despite growing concern over biased outcomes in high-stakes applications such as hate speech detection. In this work, we present the first systematic study of group fairness in task arithmetic within this binary text and image classification regime, comparing it against full fine-tuning (FFT) and Low-Rank Adaptation (LoRA). We evaluate across multiple language models and datasets using standard group fairness metrics, including Demographic Parity and Equalized Odds. Our analysis shows that task vectors can be tuned to achieve competitive accuracy while reducing disparities, and that merging subgroup-specific task vectors provides a practical mechanism for steering fairness outcomes. We further provide a theoretical bound linking task vector scaling to fairness metrics, offering insight into the observed trade-offs. Together, these findings establish task arithmetic not only as a cost-efficient editing method but also as a fairness-aware alternative to existing adaptation techniques, within the standard group-fair classification setting, laying the groundwork for responsible deployment of large language models.", "authors": ["Hiroki Naganuma", "Kotaro Yoshida", "Laura Gomezjurado Gonzalez", "Takafumi Horie", "Yuji Naraki", "Ryotaro Shimizu"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-30", "url": "https://arxiv.org/abs/2505.24262", "pdf_url": "https://arxiv.org/pdf/2505.24262v2", "arxiv_id": "2505.24262", "doi": "10.48550/arXiv.2505.24262", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1776}
{"id": "8d395e489e68c94f785c3438e7295b51f0fa078b3c5370094d6dbbf98c5c6a3b", "sources": ["arxiv", "semantic_scholar"], "title": "Continual Learning in Vision-Language Models via Aligned Model Merging", "abstract": "Continual learning is conventionally tackled through sequential fine-tuning, a process that, while enabling adaptation, inherently favors plasticity over the stability needed to retain prior knowledge. While existing approaches attempt to mitigate catastrophic forgetting, a bias towards recent tasks persists as they build upon this sequential nature. In this work we present a new perspective based on model merging to maintain stability while still retaining plasticity. Rather than just sequentially updating the model weights, we propose merging newly trained task parameters with previously learned ones, promoting a better balance. To maximize the effectiveness of the merging process, we propose a simple mechanism that promotes learning aligned weights with previous ones, thereby avoiding interference when merging. We evaluate this approach on large Vision-Language Models (VLMs), and demonstrate its effectiveness in reducing forgetting, increasing robustness to various task orders and similarities, and improving generalization.", "authors": ["Ghada Sokar", "Gintare Karolina Dziugaite", "Anurag Arnab", "Ahmet Iscen", "Pablo Samuel Castro", "Cordelia Schmid"], "categories": ["cs.CV", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-30", "url": "https://arxiv.org/abs/2506.03189", "pdf_url": "https://arxiv.org/pdf/2506.03189v1", "arxiv_id": "2506.03189", "doi": "10.48550/arXiv.2506.03189", "citation_count": 5, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1945}
{"id": "a6185c0abb0fc6ffa6cc046793083f9775593024e262a5207904b57dc3c28370", "sources": ["arxiv", "semantic_scholar"], "title": "Towards Minimizing Feature Drift in Model Merging: Layer-wise Task Vector Fusion for Adaptive Knowledge Integration", "abstract": "Multi-task model merging aims to consolidate knowledge from multiple fine-tuned task-specific experts into a unified model while minimizing performance degradation. Existing methods primarily approach this by minimizing differences between task-specific experts and the unified model, either from a parameter-level or a task-loss perspective. However, parameter-level methods exhibit a significant performance gap compared to the upper bound, while task-loss approaches entail costly secondary training procedures. In contrast, we observe that performance degradation closely correlates with feature drift, i.e., differences in feature representations of the same sample caused by model merging. Motivated by this observation, we propose Layer-wise Optimal Task Vector Merging (LOT Merging), a technique that explicitly minimizes feature drift between task-specific experts and the unified model in a layer-by-layer manner. LOT Merging can be formulated as a convex quadratic optimization problem, enabling us to analytically derive closed-form solutions for the parameters of linear and normalization layers. Consequently, LOT Merging achieves efficient model consolidation through basic matrix operations. Extensive experiments across vision and vision-language benchmarks demonstrate that LOT Merging significantly outperforms baseline methods, achieving improvements of up to 4.4% (ViT-B/32) over state-of-the-art approaches. The source code is available at https://github.com/SunWenJu123/model-merging.", "authors": ["Wenju Sun", "Qingyong Li", "Wen Wang", "Yang Liu", "Yangli-ao Geng", "Boyang Li"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-29", "url": "https://arxiv.org/abs/2505.23859", "pdf_url": "https://arxiv.org/pdf/2505.23859v2", "arxiv_id": "2505.23859", "doi": "10.48550/arXiv.2505.23859", "citation_count": 8, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/SunWenJu123/model-merging", "venue": "arXiv.org", "quality_score": 0.2727}
{"id": "d46fe0e2e297448d62f9dcb82a0e01b7738d329c18efdbf45383fe74d2d08247", "sources": ["arxiv", "semantic_scholar"], "title": "Merge Hijacking: Backdoor Attacks to Model Merging of Large Language Models", "abstract": "Model merging for Large Language Models (LLMs) directly fuses the parameters of different models finetuned on various tasks, creating a unified model for multi-domain tasks. However, due to potential vulnerabilities in models available on open-source platforms, model merging is susceptible to backdoor attacks. In this paper, we propose Merge Hijacking, the first backdoor attack targeting model merging in LLMs. The attacker constructs a malicious upload model and releases it. Once a victim user merges it with any other models, the resulting merged model inherits the backdoor while maintaining utility across tasks. Merge Hijacking defines two main objectives-effectiveness and utility-and achieves them through four steps. Extensive experiments demonstrate the effectiveness of our attack across different models, merging algorithms, and tasks. Additionally, we show that the attack remains effective even when merging real-world models. Moreover, our attack demonstrates robustness against two inference-time defenses (Paraphrasing and CLEANGEN) and one training-time defense (Fine-pruning).", "authors": ["Zenghui Yuan", "Yangming Xu", "Jiawen Shi", "Pan Zhou", "Lichao Sun"], "categories": ["cs.CR"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-29", "url": "https://arxiv.org/abs/2505.23561", "pdf_url": "https://arxiv.org/pdf/2505.23561v1", "arxiv_id": "2505.23561", "doi": "10.48550/arXiv.2505.23561", "citation_count": 7, "influential_citation_count": 1, "has_code": true, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.2727}
{"id": "a31bb996138dc749a9d7545f34808a061be87735759c988898dede4b41d960a6", "sources": ["arxiv", "semantic_scholar"], "title": "Decom-Renorm-Merge: Model Merging on the Right Space Improves Multitasking", "abstract": "In the era of large-scale training, model merging has evolved into a tool for creating multitasking models efficiently. It enables the knowledge of models to be fused, without the need for heavy computation as required in traditional multitask learning. Existing merging methods often assume that entries at identical positions in weight matrices serve the same function, enabling straightforward entry-wise comparison and merging. However, this assumption overlooks the complexity of finetuned neural networks, where neurons may develop distinct feature compositions, making direct entry-wise merging problematic. We present Decom-Renorm-Merge (DRM), a simple yet effective approach that leverages Singular Value Decomposition to decompose and coordinate weight matrices into an aligned joint space, where entry-wise merging becomes possible. We showcase the effectiveness of DRM across various settings ranging from smaller encoder-based such as ViT and DeBERTa, encoder-decoder-based such as T5, and larger decoder-based such as Llama3.1-8B. Our experimental results show that DRM outperforms several state-of-the-art merging techniques across full finetuning and low-rank adaptation settings. Moreover, our analysis reveals renormalization as the crucial component for creating a robust and even joint space for merging, significantly contributing to the method's performance.", "authors": ["Yuatyong Chaichana", "Thanapat Trachu", "Peerat Limkonchotiwat", "Konpat Preechakul", "Tirasan Khandhawit", "Ekapol Chuangsuwanich"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-29", "url": "https://arxiv.org/abs/2505.23117", "pdf_url": "https://arxiv.org/pdf/2505.23117v2", "arxiv_id": "2505.23117", "doi": "10.48550/arXiv.2505.23117", "citation_count": 2, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/yophis/decom-renorm-merge", "venue": "arXiv.org", "quality_score": 0.2727}
{"id": "e2d4a10523deca5e118442fa9a191fd76d1568fb4efeb74d2db218bb220a1538", "sources": ["arxiv", "semantic_scholar"], "title": "Unraveling LoRA Interference: Orthogonal Subspaces for Robust Model Merging", "abstract": "Fine-tuning large language models (LMs) for individual tasks yields strong performance but is expensive for deployment and storage. Recent works explore model merging to combine multiple task-specific models into a single multi-task model without additional training. However, existing merging methods often fail for models fine-tuned with low-rank adaptation (LoRA), due to significant performance degradation. In this paper, we show that this issue arises from a previously overlooked interplay between model parameters and data distributions. We propose Orthogonal Subspaces for Robust model Merging (OSRM) to constrain the LoRA subspace *prior* to fine-tuning, ensuring that updates relevant to one task do not adversely shift outputs for others. Our approach can seamlessly integrate with most existing merging algorithms, reducing the unintended interference among tasks. Extensive experiments on eight datasets, tested with three widely used LMs and two large LMs, demonstrate that our method not only boosts merging performance but also preserves single-task accuracy. Furthermore, our approach exhibits greater robustness to the hyperparameters of merging. These results highlight the importance of data-parameter interaction in model merging and offer a plug-and-play solution for merging LoRA models.", "authors": ["Haobo Zhang", "Jiayu Zhou"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Medicine", "Computer Science"], "published_date": "2025-05-28", "url": "https://arxiv.org/abs/2505.22934", "pdf_url": "https://arxiv.org/pdf/2505.22934v2", "arxiv_id": "2505.22934", "doi": "10.48550/arXiv.2505.22934", "citation_count": 8, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.2386}
{"id": "d40d7d3d86fd67f192e0f3563bf9b353d5264314fe71ebcbb0a40a9f9700a8bd", "sources": ["arxiv", "semantic_scholar"], "title": "OptMerge: Unifying Multimodal LLM Capabilities and Modalities via Model Merging", "abstract": "Foundation models update slowly due to resource-intensive training, whereas domain-specific models evolve rapidly between releases. Model merging seeks to combine multiple expert models into a single, more capable model, reducing storage and serving costs while supporting decentralized development. Despite its potential, previous studies have primarily focused on merging visual classification models or Large Language Models (LLMs) for code and math tasks. Recently, Multimodal LLMs (MLLMs) that extend LLMs through large-scale multimodal training have gained traction. However, there lacks a benchmark for model merging research that clearly divides the tasks for MLLM training and evaluation. In this paper, $\\textbf{(i)}$ we introduce a model merging benchmark for MLLMs, which includes multiple tasks such as VQA, Geometry, Chart, OCR, and Grounding, studying both LoRA and full fine-tuning models. Moreover, we explore how model merging can combine different modalities (e.g., vision-language, audio-language, and video-language models), moving toward the Omni-language model. $\\textbf{(ii)}$ We implement 10 model merging algorithms on the benchmark. Furthermore, we propose a novel method that removes noise from task vectors and robustly optimizes the merged vector based on a loss defined over task vector interactions, achieving an average performance gain of 2.48%. $\\textbf{(iii)}$ We find that model merging offers a promising way for building improved MLLMs without requiring training data. Our results also demonstrate that the complementarity among multiple modalities outperforms individual modalities.", "authors": ["Yongxian Wei", "Runxi Cheng", "Weike Jin", "Enneng Yang", "Li Shen", "Lu Hou", "Sinan Du", "Chun Yuan", "Xiaochun Cao", "Dacheng Tao"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-26", "url": "https://arxiv.org/abs/2505.19892", "pdf_url": "https://arxiv.org/pdf/2505.19892v3", "arxiv_id": "2505.19892", "doi": null, "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1747}
{"id": "f878f4f4ef4bb59b4f68b2d2980e6b4c658043047a0a04a01cda5cc65e321c14", "sources": ["arxiv", "semantic_scholar"], "title": "NAN: A Training-Free Solution to Coefficient Estimation in Model Merging", "abstract": "Model merging offers a training-free alternative to multi-task learning by combining independently fine-tuned models into a unified one without access to raw data. However, existing approaches often rely on heuristics to determine the merging coefficients, limiting their scalability and generality. In this work, we revisit model merging through the lens of least-squares optimization and show that the optimal merging weights should scale with the amount of task-specific information encoded in each model. Based on this insight, we propose NAN, a simple yet effective method that estimates model merging coefficients via the inverse of parameter norm. NAN is training-free, plug-and-play, and applicable to a wide range of merging strategies. Extensive experiments on show that NAN consistently improves performance of baseline methods.", "authors": ["Chongjie Si", "Kangtao Lv", "Jingjing Jiang", "Yadao Wang", "Yongwei Wang", "Xiaokang Yang", "Wenbo Su", "Bo Zheng", "Wei Shen"], "categories": ["cs.LG", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-22", "url": "https://arxiv.org/abs/2505.16148", "pdf_url": "https://arxiv.org/pdf/2505.16148v1", "arxiv_id": "2505.16148", "doi": "10.48550/arXiv.2505.16148", "citation_count": 3, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1684}
{"id": "48cd5a042de1da296c35f5b56611cb1173909310b7fdb689380be939145e4f2b", "sources": ["arxiv", "semantic_scholar"], "title": "Merge to Mix: Mixing Datasets via Model Merging", "abstract": "Mixing datasets for fine-tuning large models (LMs) has become critical for maximizing performance on downstream tasks. However, composing effective dataset mixtures typically relies on heuristics and trial-and-error, often requiring multiple fine-tuning runs to achieve the desired outcome. We propose a novel method, $\\textit{Merge to Mix}$, that accelerates composing dataset mixtures through model merging. Model merging is a recent technique that combines the abilities of multiple individually fine-tuned LMs into a single LM by using a few simple arithmetic operations. Our key insight is that merging models individually fine-tuned on each dataset in a mixture can effectively serve as a surrogate for a model fine-tuned on the entire mixture. Merge to Mix leverages this insight to accelerate selecting dataset mixtures without requiring full fine-tuning on each candidate mixture. Our experiments demonstrate that Merge to Mix surpasses state-of-the-art methods in dataset selection for fine-tuning LMs.", "authors": ["Zhixu Silvia Tao", "Kasper Vinken", "Hao-Wei Yeh", "Avi Cooper", "Xavier Boix"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-21", "url": "https://arxiv.org/abs/2505.16066", "pdf_url": "https://arxiv.org/pdf/2505.16066v1", "arxiv_id": "2505.16066", "doi": "10.48550/arXiv.2505.16066", "citation_count": 6, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2386}
{"id": "2b5d8d7ad965899f47090c03c32be333cde2720104fc1b157ee1846c4af86dc3", "sources": ["arxiv", "semantic_scholar"], "title": "CCNU at SemEval-2025 Task 3: Leveraging Internal and External Knowledge of Large Language Models for Multilingual Hallucination Annotation", "abstract": "We present the system developed by the Central China Normal University (CCNU) team for the Mu-SHROOM shared task, which focuses on identifying hallucinations in question-answering systems across 14 different languages. Our approach leverages multiple Large Language Models (LLMs) with distinct areas of expertise, employing them in parallel to annotate hallucinations, effectively simulating a crowdsourcing annotation process. Furthermore, each LLM-based annotator integrates both internal and external knowledge related to the input during the annotation process. Using the open-source LLM DeepSeek-V3, our system achieves the top ranking (\\#1) for Hindi data and secures a Top-5 position in seven other languages. In this paper, we also discuss unsuccessful approaches explored during our development process and share key insights gained from participating in this shared task.", "authors": ["Xu Liu", "Guanyi Chen"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-17", "url": "https://arxiv.org/abs/2505.11965", "pdf_url": "https://arxiv.org/pdf/2505.11965v1", "arxiv_id": "2505.11965", "doi": "10.48550/arXiv.2505.11965", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": null, "quality_score": 0.1923}
{"id": "20dea47a1d5fafcbff7325e14ae0d35b221123f860e2c845991a0843eddb75ad", "sources": ["arxiv", "semantic_scholar"], "title": "Mergenetic: a Simple Evolutionary Model Merging Library", "abstract": "Model merging allows combining the capabilities of existing models into a new one - post hoc, without additional training. This has made it increasingly popular thanks to its low cost and the availability of libraries that support merging on consumer GPUs. Recent work shows that pairing merging with evolutionary algorithms can boost performance, but no framework currently supports flexible experimentation with such strategies in language models. We introduce Mergenetic, an open-source library for evolutionary model merging. Mergenetic enables easy composition of merging methods and evolutionary algorithms while incorporating lightweight fitness estimators to reduce evaluation costs. We describe its design and demonstrate that Mergenetic produces competitive results across tasks and languages using modest hardware.", "authors": ["Adrian Robert Minut", "Tommaso Mencattini", "Andrea Santilli", "Donato Crisostomi", "Emanuele Rodolà"], "categories": ["cs.LG", "cs.AI", "cs.NE"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-16", "url": "https://arxiv.org/abs/2505.11427", "pdf_url": "https://arxiv.org/pdf/2505.11427v1", "arxiv_id": "2505.11427", "doi": "10.18653/v1/2025.acl-demo.55", "citation_count": 4, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/tommasomncttn/mergenetic", "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.2497}
{"id": "56058ee0f7b2af9b760827d040018674f0cea830db72a55068f5eae35a78d643", "sources": ["arxiv", "semantic_scholar"], "title": "CAT Merging: A Training-Free Approach for Resolving Conflicts in Model Merging", "abstract": "Multi-task model merging offers a promising paradigm for integrating multiple expert models into a unified model without additional training. Existing state-of-the-art techniques, such as Task Arithmetic and its variants, merge models by accumulating task vectors -- the parameter differences between pretrained and finetuned models. However, task vector accumulation is often hindered by knowledge conflicts, leading to performance degradation. To address this challenge, we propose Conflict-Aware Task Merging (CAT Merging), a novel training-free framework that selectively trims conflict-prone components from the task vectors. CAT Merging introduces several parameter-specific strategies, including projection for linear weights and masking for scaling and shifting parameters in normalization layers. Extensive experiments on vision, language, and vision-language tasks demonstrate that CAT Merging effectively suppresses knowledge conflicts, achieving average accuracy improvements of up to 2.5% (ViT-B/32) and 2.0% (ViT-L/14) over state-of-the-art methods.", "authors": ["Wenju Sun", "Qingyong Li", "Yangli-ao Geng", "Boyang Li"], "categories": ["cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-11", "url": "https://arxiv.org/abs/2505.06977", "pdf_url": "https://arxiv.org/pdf/2505.06977v2", "arxiv_id": "2505.06977", "doi": "10.48550/arXiv.2505.06977", "citation_count": 17, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.3138}
{"id": "ccfd23b57c88be4e7f155479abea72b9c2196cf9f435c14d507ef40e179f4ef8", "sources": ["arxiv", "semantic_scholar"], "title": "Proof Complexity and Feasible Interpolation", "abstract": "This is a survey on propositional proof complexity aimed at introducing the basics of the field with a particular focus on a method known as feasible interpolation. This method is used to construct \"hard theorems\" for several proof systems for both classical and non-classical logics. Here, a \"hard theorem\" refers to a theorem in the logic whose shortest proofs are super-polynomially long in the length of the theorem itself. To make this survey more accessible, we only assume a basic familiarity with propositional, modal, and first-order logic, as well as a basic understanding of the key concepts in computational complexity, such as the definitions of the classes $\\mathbf{NP}$ and $\\mathbf{PSPACE}$. Any additional concepts will be introduced and explained as needed.", "authors": ["Amirhossein Akbar Tabatabai"], "categories": ["math.LO", "cs.CC"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2025-05-05", "url": "https://arxiv.org/abs/2505.03002", "pdf_url": "https://arxiv.org/pdf/2505.03002v1", "arxiv_id": "2505.03002", "doi": "10.48550/arXiv.2505.03002", "citation_count": 1, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1505}
{"id": "f3d55c530b0187b93c963f7681fc65454054e2db92825325ee1f1d03ab1b4c23", "sources": ["arxiv", "semantic_scholar"], "title": "Investigating Task Arithmetic for Zero-Shot Information Retrieval", "abstract": "Large Language Models (LLMs) have shown impressive zero-shot performance across a variety of Natural Language Processing tasks, including document re-ranking. However, their effectiveness degrades on unseen tasks and domains, largely due to shifts in vocabulary and word distributions. In this paper, we investigate Task Arithmetic, a technique that combines the weights of LLMs pre-trained on different tasks or domains via simple mathematical operations, such as addition or subtraction, to adapt retrieval models without requiring additional fine-tuning. Our method is able to synthesize diverse tasks and domain knowledge into a single model, enabling effective zero-shot adaptation in different retrieval contexts. Extensive experiments on publicly available scientific, biomedical, and multilingual datasets show that our method improves state-of-the-art re-ranking performance by up to 18% in NDCG@10 and 15% in P@10. In addition to these empirical gains, our analysis provides insights into the strengths and limitations of Task Arithmetic as a practical strategy for zero-shot learning and model adaptation. We make our code publicly available at https://github.com/DetectiveMB/Task-Arithmetic-for-ZS-IR.", "authors": ["Marco Braga", "Pranav Kasela", "Alessandro Raganato", "Gabriella Pasi"], "categories": ["cs.IR", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-01", "url": "https://arxiv.org/abs/2505.00649", "pdf_url": "https://arxiv.org/pdf/2505.00649v1", "arxiv_id": "2505.00649", "doi": "10.1145/3726302.3730216", "citation_count": 4, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/DetectiveMB/Task-Arithmetic-for-ZS-IR", "venue": "Italian Information Retrieval Workshop", "quality_score": 0.2231}
{"id": "6d76dcb05adccf34f062a77a5fc454a1d72d51fb66643964b2036f9bb5eafb1a", "sources": ["arxiv", "semantic_scholar"], "title": "A stochastic epidemic model with memory of the last infection and waning immunity", "abstract": "We adapt the article of Forien, Pang, Pardoux and Zotsa: Arxiv preprint Arxiv2210.04667(2022), on epidemic models with varying infectivity and waning immunity, to incorporate the memory of the last infection. To this end, we introduce a parametric approach and consider a piecewise deterministic Markov process modeling both the evolution of the parameter, also called the trait, and the age of infection of individuals over time. At each new infection, a new trait is randomly chosen for the infected individual according to a Markov kernel, and their age is reset to zero. In the large population limit, we derive a partial differential equation (PDE) that describes the density of traits and ages. The main goal is to study the conditions under which endemic equilibria exist for the deterministic PDE model and to establish an endemicity threshold that depends on the model parameters. The local stability of these equilibria is also analyzed. The endemicity threshold is computed for several examples, including models that incorporate a vaccination policy, and a local stability result is obtained for a memory-free SIS-type model.", "authors": ["Hélène Guérin", "Arsene Brice Zotsa-Ngoufack"], "categories": ["math.PR", "q-bio.PE"], "fields_of_study": ["Mathematics", "Biology"], "published_date": "2025-05-01", "url": "https://arxiv.org/abs/2505.00601", "pdf_url": "https://arxiv.org/pdf/2505.00601v1", "arxiv_id": "2505.00601", "doi": null, "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1193}
{"id": "19f8b0f64027c5d56e9b024859b14768723a28864e15bf12d199b461a87710e2", "sources": ["arxiv", "semantic_scholar"], "title": "UniBiomed: A Universal Foundation Model for Grounded Biomedical Image Interpretation", "abstract": "The integration of AI-assisted biomedical image analysis into clinical practice demands AI-generated findings that are not only accurate but also interpretable to clinicians. However, existing biomedical AI models generally lack the ability to simultaneously generate diagnostic findings and localize corresponding biomedical objects. This limitation makes it challenging for clinicians to correlate AI-generated findings with visual evidence (e.g., tiny lesions) in images and interpret the results of AI models. To address this challenge, we introduce UniBiomed, the first universal foundation model for grounded biomedical image interpretation, which is capable of generating accurate diagnostic findings and simultaneously segmenting the corresponding biomedical targets. UniBiomed is based on a novel integration of Multi-modal Large Language Model and Segment Anything Model, which can effectively unify diverse biomedical tasks in universal training for advancing grounded interpretation. To develop UniBiomed, we curate a large-scale dataset comprising over 27 million triplets of images, region annotations, and text descriptions across ten biomedical imaging modalities. Extensive validation on 70 internal and 14 external datasets demonstrated the state-of-the-art performance of UniBiomed in diverse biomedical tasks, including image segmentation, disease recognition, region-aware diagnosis, vision question answering, and report generation. In summary, UniBiomed is a powerful and versatile biomedical foundation model, unlocking the untapped grounded interpretation capability for optimizing AI-assisted biomedical image analysis.", "authors": ["Linshan Wu", "Yuxiang Nie", "Sunan He", "Jiaxin Zhuang", "Luyang Luo", "Tao Li", "Zhuoyao Xie", "Dexuan Chen", "Yinghua Zhao", "Neeraj Mahboobani", "Varut Vardhanabhuti", "Ronald Cheong Kin Chan", "Yifan Peng", "Pranav Rajpurkar", "Hao Chen"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-30", "url": "https://arxiv.org/abs/2504.21336", "pdf_url": "https://arxiv.org/pdf/2504.21336v3", "arxiv_id": "2504.21336", "doi": "10.48550/arXiv.2504.21336", "citation_count": 9, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.25}
{"id": "cfa5576cd488ae2a561daf5828b199b05b9fe21efe45cc30777854dd64992265", "sources": ["arxiv", "semantic_scholar"], "title": "Dynamic Fisher-weighted Model Merging via Bayesian Optimization", "abstract": "The fine-tuning of pre-trained language models has resulted in the widespread availability of task-specific models. Model merging offers an efficient way to create multi-task models by combining these fine-tuned models at the parameter level, without the need for training data or joint training on multiple datasets. Existing merging approaches typically involve scaling the parameters model-wise or integrating parameter importance parameter-wise. Both approaches exhibit their own weaknesses, leading to a notable performance gap compared to multi-task fine-tuning. In this paper, we unify these seemingly distinct strategies into a more general merging framework, and introduce Dynamic Fisher-weighted Merging (DF-Merge). Specifically, candidate models are associated with a set of coefficients that linearly scale their fine-tuned parameters. Bayesian optimization is applied to dynamically adjust these coefficients, aiming to maximize overall performance on validation sets. Each iteration of this process integrates parameter importance based on the Fisher information conditioned by the coefficients. Experimental results show that DF-Merge outperforms strong baselines across models of different sizes and a variety of tasks. Our analysis shows that the effectiveness of DF-Merge arises from the unified view of merging and that near-optimal performance is achievable in a few iterations, even with minimal validation data.", "authors": ["Sanwoo Lee", "Jiahao Liu", "Qifan Wang", "Jingang Wang", "Xunliang Cai", "Yunfang Wu"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-26", "url": "https://arxiv.org/abs/2504.18992", "pdf_url": "https://arxiv.org/pdf/2504.18992v1", "arxiv_id": "2504.18992", "doi": "10.48550/arXiv.2504.18992", "citation_count": 13, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "North American Chapter of the Association for Computational Linguistics", "quality_score": 0.2865}
{"id": "4302b7cce4166202a1e2648cdd75635dcc126cbd8c3a0b8088ae3cf312102dee", "sources": ["arxiv", "semantic_scholar"], "title": "Parameter-Efficient Checkpoint Merging via Metrics-Weighted Averaging", "abstract": "Checkpoint merging is a technique for combining multiple model snapshots into a single superior model, potentially reducing training time for large language models. This paper explores checkpoint merging in the context of parameter-efficient fine-tuning (PEFT), where only small adapter modules (e.g. LoRA) are trained. We propose Metrics-Weighted Averaging (MWA), a simple yet effective method to merge model checkpoints by weighting their parameters according to performance metrics. In particular, we investigate weighting by training loss and by training steps, under the intuition that lower-loss or later-step checkpoints are more valuable. We introduce a formula with a penalty factor to adjust weight distribution, requiring only one hyperparameter regardless of the number of checkpoints. Experiments on three fine-tuning tasks (mathematical reasoning, preference alignment, and general instruction tuning) show that MWA consistently produces merged models that outperform the naive uniform average of checkpoints. Notably, loss-weighted merging often yields the best results, delivering up to 5% higher task accuracy than the baseline uniform merge and even surpassing the final individual checkpoint's performance. These findings validate checkpoint merging for PEFT and demonstrate that a metric-driven weighting heuristic can efficiently boost model performance with minimal computational overhead.", "authors": ["Shi Jie Yu", "Sehyun Choi"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-23", "url": "https://arxiv.org/abs/2504.18580", "pdf_url": "https://arxiv.org/pdf/2504.18580v1", "arxiv_id": "2504.18580", "doi": "10.48550/arXiv.2504.18580", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1352}
{"id": "18e61d5b46216e059f1fe0b8a99dc0f0395c71d8c33a8d92e44c38cd7607fb8c", "sources": ["arxiv", "semantic_scholar"], "title": "Mitigating Parameter Interference in Model Merging via Sharpness-Aware Fine-Tuning", "abstract": "Large-scale deep learning models with a pretraining-finetuning paradigm have led to a surge of numerous task-specific models fine-tuned from a common pre-trained model. Recently, several research efforts have been made on merging these large models into a single multi-task model, particularly with simple arithmetic on parameters. Such merging methodology faces a central challenge: interference between model parameters fine-tuned on different tasks. Few recent works have focused on designing a new fine-tuning scheme that can lead to small parameter interference, however at the cost of the performance of each task-specific fine-tuned model and thereby limiting that of a merged model. To improve the performance of a merged model, we note that a fine-tuning scheme should aim for (1) smaller parameter interference and (2) better performance of each fine-tuned model on the corresponding task. In this work, we aim to design a new fine-tuning objective function to work towards these two goals. In the course of this process, we find such objective function to be strikingly similar to sharpness-aware minimization (SAM) objective function, which aims to achieve generalization by finding flat minima. Drawing upon our observation, we propose to fine-tune pre-trained models via sharpness-aware minimization. The experimental and theoretical results showcase the effectiveness and orthogonality of our proposed approach, improving performance upon various merging and fine-tuning methods. Our code is available at https://github.com/baiklab/SAFT-Merge.", "authors": ["Yeoreum Lee", "Jinwook Jung", "Sungyong Baik"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-20", "url": "https://arxiv.org/abs/2504.14662", "pdf_url": "https://arxiv.org/pdf/2504.14662v1", "arxiv_id": "2504.14662", "doi": "10.48550/arXiv.2504.14662", "citation_count": 9, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/baiklab/SAFT-Merge", "venue": "International Conference on Learning Representations", "quality_score": 0.25}
{"id": "83164da5f201c2bc04e008676675bab1f39ada3965fbba4fa0f4dd9a5980ba9a", "sources": ["arxiv", "semantic_scholar"], "title": "DMM: Building a Versatile Image Generation Model via Distillation-Based Model Merging", "abstract": "The success of text-to-image (T2I) generation models has spurred a proliferation of numerous model checkpoints fine-tuned from the same base model on various specialized datasets. This overwhelming specialized model production introduces new challenges for high parameter redundancy and huge storage cost, thereby necessitating the development of effective methods to consolidate and unify the capabilities of diverse powerful models into a single one. A common practice in model merging adopts static linear interpolation in the parameter space to achieve the goal of style mixing. However, it neglects the features of T2I generation task that numerous distinct models cover sundry styles which may lead to incompatibility and confusion in the merged model. To address this issue, we introduce a style-promptable image generation pipeline which can accurately generate arbitrary-style images under the control of style vectors. Based on this design, we propose the score distillation based model merging paradigm (DMM), compressing multiple models into a single versatile T2I model. Moreover, we rethink and reformulate the model merging task in the context of T2I generation, by presenting new merging goals and evaluation protocols. Our experiments demonstrate that DMM can compactly reorganize the knowledge from multiple teacher models and achieve controllable arbitrary-style generation.", "authors": ["Tianhui Song", "Weixin Feng", "Shuai Wang", "Xubin Li", "Tiezheng Ge", "Bo Zheng", "Limin Wang"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-16", "url": "https://arxiv.org/abs/2504.12364", "pdf_url": "https://arxiv.org/pdf/2504.12364v1", "arxiv_id": "2504.12364", "doi": "10.48550/arXiv.2504.12364", "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1945}
{"id": "736838d80219024166d1949c353e3c0671fbea4ef5062f45fe7e149455943991", "sources": ["arxiv", "semantic_scholar"], "title": "Single-Input Multi-Output Model Merging: Leveraging Foundation Models for Dense Multi-Task Learning", "abstract": "Model merging is a flexible and computationally tractable approach to merge single-task checkpoints into a multi-task model. Prior work has solely focused on constrained multi-task settings where there is a one-to-one mapping between a sample and a task, overlooking the paradigm where multiple tasks may operate on the same sample, e.g., scene understanding. In this paper, we focus on the multi-task setting with single-input-multiple-outputs (SIMO) and show that it qualitatively differs from the single-input-single-output model merging settings studied in the literature due to the existence of task-specific decoders and diverse loss objectives. We identify that existing model merging methods lead to significant performance degradation, primarily due to representation misalignment between the merged encoder and task-specific decoders. We propose two simple and efficient fixes for the SIMO setting to re-align the feature representation after merging. Compared to joint fine-tuning, our approach is computationally effective and flexible, and sheds light into identifying task relationships in an offline manner. Experiments on NYUv2, Cityscapes, and a subset of the Taskonomy dataset demonstrate: (1) task arithmetic suffices to enable multi-task capabilities; however, the representations generated by the merged encoder has to be re-aligned with the task-specific heads; (2) the proposed architecture rivals traditional multi-task learning in performance but requires fewer samples and training steps by leveraging the existence of task-specific models.", "authors": ["Juan Garcia Giraldo", "Nikolaos Dimitriadis", "Ke Wang", "Pascal Frossard"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-15", "url": "https://arxiv.org/abs/2504.11268", "pdf_url": "https://arxiv.org/pdf/2504.11268v1", "arxiv_id": "2504.11268", "doi": "10.48550/arXiv.2504.11268", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.126}
{"id": "3f869b04217f2d2b4e5fb38c2c1bf9e94a5413d69328993f7c44e494fdd70bb1", "sources": ["arxiv", "semantic_scholar"], "title": "Leveraging Submodule Linearity Enhances Task Arithmetic Performance in LLMs", "abstract": "Task arithmetic is a straightforward yet highly effective strategy for model merging, enabling the resultant model to exhibit multi-task capabilities. Recent research indicates that models demonstrating linearity enhance the performance of task arithmetic. In contrast to existing methods that rely on the global linearization of the model, we argue that this linearity already exists within the model's submodules. In particular, we present a statistical analysis and show that submodules (e.g., layers, self-attentions, and MLPs) exhibit significantly higher linearity than the overall model. Based on these findings, we propose an innovative model merging strategy that independently merges these submodules. Especially, we derive a closed-form solution for optimal merging weights grounded in the linear properties of these submodules. Experimental results demonstrate that our method consistently outperforms the standard task arithmetic approach and other established baselines across different model scales and various tasks. This result highlights the benefits of leveraging the linearity of submodules and provides a new perspective for exploring solutions for effective and practical multi-task model merging.", "authors": ["Rui Dai", "Sile Hu", "Xu Shen", "Yonggang Zhang", "Xinmei Tian", "Jieping Ye"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-15", "url": "https://arxiv.org/abs/2504.10902", "pdf_url": "https://arxiv.org/pdf/2504.10902v1", "arxiv_id": "2504.10902", "doi": "10.48550/arXiv.2504.10902", "citation_count": 10, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Conference on Learning Representations", "quality_score": 0.2603}
{"id": "86d7f7327da1e67bcd5a85edd96c31a68f1955920778729b8c1e51f826797b4c", "sources": ["arxiv", "semantic_scholar"], "title": "PALACE v1.0: Paranal Airglow Line And Continuum Emission model", "abstract": "Below about 2.3 $μ$m, the nighttime emission of the Earth's atmosphere is dominated by non-thermal radiation from the mesosphere and thermosphere. As this airglow can even outshine scattered moonlight in the near-infrared regime, the understanding of the Earth's night-sky brightness requires good knowledge of the complex airglow emission spectrum and its variability. As airglow modelling is very challenging, the comprehensive characterisation of airglow emission requires large data sets of empirical data. For fixed locations, this can be best achieved by archived spectra of large astronomical telescopes with a wide wavelength coverage, high spectral resolving power, and good temporal sampling. Using 10 years of data from the X-shooter echelle spectrograph in the wavelength range from 0.3 to 2.5 $μ$m and additional data from the Ultraviolet and Visual Echelle Spectrograph at the Very Large Telescope at Cerro Paranal in Chile, we have succeeded to build a comprehensive spectroscopic airglow model for this low-latitude site under consideration of theoretical data from the HITRAN database for molecules and from different sources for atoms. The Paranal Airglow Line And Continuum Emission (PALACE) model comprises 9 chemical species, 26,541 emission lines, and 3 unresolved continuum components. Moreover, there are climatologies of relative intensity, solar cycle effect, and residual variability with respect to local time and day of year for 23 variability classes. Spectra can be calculated with a stand-alone code for different conditions, also including optional atmospheric absorption and scattering. In comparison to the observed X-shooter spectra, PALACE shows convincing agreement and is significantly better than the previous, widely used airglow model for Cerro Paranal.", "authors": ["Stefan Noll", "Carsten Schmidt", "Patrick Hannawald", "Wolfgang Kausch", "Stefan Kimeswenger"], "categories": ["physics.ao-ph", "astro-ph.EP", "astro-ph.IM"], "fields_of_study": ["Physics"], "published_date": "2025-04-14", "url": "https://arxiv.org/abs/2504.10683", "pdf_url": "https://arxiv.org/pdf/2504.10683v1", "arxiv_id": "2504.10683", "doi": "10.5194/gmd-18-4353-2025", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Geoscientific Model Development", "quality_score": 0.1249}
{"id": "206c0baf338cd2251d94e0c05978086ae8cd32cf892e987ef2d65d555bc0985b", "sources": ["arxiv", "semantic_scholar"], "title": "FedMerge: Federated Personalization via Model Merging", "abstract": "One global model in federated learning (FL) might not be sufficient to serve many clients with non-IID tasks and distributions. While there has been advances in FL to train multiple global models for better personalization, they only provide limited choices to clients so local finetuning is still indispensable. In this paper, we propose a novel ``FedMerge'' approach that can create a personalized model per client by simply merging multiple global models with automatically optimized and customized weights. In FedMerge, a few global models can serve many non-IID clients, even without further local finetuning. We formulate this problem as a joint optimization of global models and the merging weights for each client. Unlike existing FL approaches where the server broadcasts one or multiple global models to all clients, the server only needs to send a customized, merged model to each client. Moreover, instead of periodically interrupting the local training and re-initializing it to a global model, the merged model aligns better with each client's task and data distribution, smoothening the local-global gap between consecutive rounds caused by client drift. We evaluate FedMerge on three different non-IID settings applied to different domains with diverse tasks and data types, in which FedMerge consistently outperforms existing FL approaches, including clustering-based and mixture-of-experts (MoE) based methods.", "authors": ["Shutong Chen", "Tianyi Zhou", "Guodong Long", "Jing Jiang", "Chengqi Zhang"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-09", "url": "https://arxiv.org/abs/2504.06768", "pdf_url": "https://arxiv.org/pdf/2504.06768v3", "arxiv_id": "2504.06768", "doi": "10.48550/arXiv.2504.06768", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.1193}
{"id": "12c5ea21581bb85fdfb5ff0050b052585163f69b2af0c1bfc7c3a8783072aa1f", "sources": ["arxiv", "semantic_scholar"], "title": "Efficient Model Editing with Task-Localized Sparse Fine-tuning", "abstract": "Task arithmetic has emerged as a promising approach for editing models by representing task-specific knowledge as composable task vectors. However, existing methods rely on network linearization to derive task vectors, leading to computational bottlenecks during training and inference. Moreover, linearization alone does not ensure weight disentanglement, the key property that enables conflict-free composition of task vectors. To address this, we propose TaLoS which allows to build sparse task vectors with minimal interference without requiring explicit linearization and sharing information across tasks. We find that pre-trained models contain a subset of parameters with consistently low gradient sensitivity across tasks, and that sparsely updating only these parameters allows for promoting weight disentanglement during fine-tuning. Our experiments prove that TaLoS improves training and inference efficiency while outperforming current methods in task addition and negation. By enabling modular parameter editing, our approach fosters practical deployment of adaptable foundation models in real-world applications.", "authors": ["Leonardo Iurada", "Marco Ciccone", "Tatiana Tommasi"], "categories": ["cs.LG", "cs.AI", "cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-03", "url": "https://arxiv.org/abs/2504.02620", "pdf_url": "https://arxiv.org/pdf/2504.02620v1", "arxiv_id": "2504.02620", "doi": "10.48550/arXiv.2504.02620", "citation_count": 13, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/iurada/talos-task-arithmetic", "venue": "International Conference on Learning Representations", "quality_score": 0.2865}
{"id": "c05f328ad855864399b1eccd282d45840dd25baec0a834d25987d81c47c190b0", "sources": ["arxiv", "semantic_scholar"], "title": "AdaMMS: Model Merging for Heterogeneous Multimodal Large Language Models with Unsupervised Coefficient Optimization", "abstract": "Recently, model merging methods have demonstrated powerful strengths in combining abilities on various tasks from multiple Large Language Models (LLMs). While previous model merging methods mainly focus on merging homogeneous models with identical architecture, they meet challenges when dealing with Multimodal Large Language Models (MLLMs) with inherent heterogeneous property, including differences in model architecture and the asymmetry in the parameter space. In this work, we propose AdaMMS, a novel model merging method tailored for heterogeneous MLLMs. Our method tackles the challenges in three steps: mapping, merging and searching. Specifically, we first design mapping function between models to apply model merging on MLLMs with different architecture. Then we apply linear interpolation on model weights to actively adapt the asymmetry in the heterogeneous MLLMs. Finally in the hyper-parameter searching step, we propose an unsupervised hyper-parameter selection method for model merging. As the first model merging method capable of merging heterogeneous MLLMs without labeled data, extensive experiments on various model combinations demonstrated that AdaMMS outperforms previous model merging methods on various vision-language benchmarks.", "authors": ["Yiyang Du", "Xiaochen Wang", "Chi Chen", "Jiabo Ye", "Yiru Wang", "Peng Li", "Ming Yan", "Ji Zhang", "Fei Huang", "Zhifang Sui", "Maosong Sun", "Yang Liu"], "categories": ["cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-31", "url": "https://arxiv.org/abs/2503.23733", "pdf_url": "https://arxiv.org/pdf/2503.23733v1", "arxiv_id": "2503.23733", "doi": "10.1109/CVPR52734.2025.00879", "citation_count": 15, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Computer Vision and Pattern Recognition", "quality_score": 0.301}
{"id": "be319889b04822ac2299242a054dd954dce4fc6df152502f8e972649119fe350", "sources": ["arxiv", "semantic_scholar"], "title": "Breaking Language Barriers in Visual Language Models via Multilingual Textual Regularization", "abstract": "Rapid advancements in Visual Language Models (VLMs) have transformed multimodal understanding but are often constrained by generating English responses regardless of the input language. This phenomenon has been termed as Image-induced Fidelity Loss (IFL) and stems from limited multimodal multilingual training data. To address this, we propose a continuous multilingual integration strategy that injects text-only multilingual data during visual instruction tuning, preserving the language model's original multilingual capabilities. Extensive evaluations demonstrate that our approach significantly improves linguistic fidelity across languages without degradation in visual performance. We also explore model merging, which improves language fidelity but comes at the cost of visual performance. In contrast, our core method achieves robust multilingual alignment without trade-offs, offering a scalable and effective path to mitigating IFL for global VLM adoption.", "authors": ["Iñigo Pikabea", "Iñaki Lacunza", "Oriol Pareras", "Carlos Escolano", "Aitor Gonzalez-Agirre", "Javier Hernando", "Marta Villegas"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-28", "url": "https://arxiv.org/abs/2503.22577", "pdf_url": "https://arxiv.org/pdf/2503.22577v2", "arxiv_id": "2503.22577", "doi": "10.48550/arXiv.2503.22577", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1505}
{"id": "b81b9301137fac8c40c4a93928b4cfa9d0790a210b6aef15fe3720e7ff28a0c4", "sources": ["arxiv", "semantic_scholar"], "title": "AdaRank: Adaptive Rank Pruning for Enhanced Model Merging", "abstract": "Model merging has emerged as a promising approach for unifying independently fine-tuned models into an integrated framework, significantly enhancing computational efficiency in multi-task learning. Recently, several SVD-based techniques have been introduced to exploit low-rank structures for enhanced merging, but their reliance on such manually designed rank selection often leads to cross-task interference and suboptimal performance. In this paper, we propose AdaRank, a novel model merging framework that adaptively selects the most beneficial singular directions of task vectors to merge multiple models. We empirically show that the dominant singular components of task vectors can cause critical interference with other tasks, and that naive truncation across tasks and layers degrades performance. In contrast, AdaRank dynamically prunes the singular components that cause interference and offers an optimal amount of information to each task vector by learning to prune ranks during test-time via entropy minimization. Our analysis demonstrates that such method mitigates detrimental overlaps among tasks, while empirical results show that AdaRank consistently achieves state-of-the-art performance with various backbones and number of tasks, reducing the performance gap between fine-tuned models to nearly 1%.", "authors": ["Chanhyuk Lee", "Jiho Choi", "Chanryeol Lee", "Donggyun Kim", "Seunghoon Hong"], "categories": ["cs.LG", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-28", "url": "https://arxiv.org/abs/2503.22178", "pdf_url": "https://arxiv.org/pdf/2503.22178v3", "arxiv_id": "2503.22178", "doi": "10.48550/arXiv.2503.22178", "citation_count": 7, "influential_citation_count": 1, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2258}
{"id": "41a290eae1f935f5613cfe2cb909e945c241413d957c278bb78d0aa1e827eb10", "sources": ["arxiv", "semantic_scholar"], "title": "ZJUKLAB at SemEval-2025 Task 4: Unlearning via Model Merging", "abstract": "This paper presents the ZJUKLAB team's submission for SemEval-2025 Task 4: Unlearning Sensitive Content from Large Language Models. This task aims to selectively erase sensitive knowledge from large language models, avoiding both over-forgetting and under-forgetting issues. We propose an unlearning system that leverages Model Merging (specifically TIES-Merging), combining two specialized models into a more balanced unlearned model. Our system achieves competitive results, ranking second among 26 teams, with an online score of 0.944 for Task Aggregate and 0.487 for overall Aggregate. In this paper, we also conduct local experiments and perform a comprehensive analysis of the unlearning process, examining performance trajectories, loss dynamics, and weight perspectives, along with several supplementary experiments, to understand the effectiveness of our method. Furthermore, we analyze the shortcomings of our method and evaluation metrics, emphasizing that MIA scores and ROUGE-based metrics alone are insufficient to fully evaluate successful unlearning. Finally, we emphasize the need for more comprehensive evaluation methodologies and rethinking of unlearning objectives in future research. Code is available at https://github.com/zjunlp/unlearn/tree/main/semeval25.", "authors": ["Haoming Xu", "Shuxun Wang", "Yanqiu Zhao", "Yi Zhong", "Ziyan Jiang", "Ningyuan Zhao", "Shumin Deng", "Huajun Chen", "Ningyu Zhang"], "categories": ["cs.CL", "cs.AI", "cs.CV", "cs.LG", "cs.MM"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-27", "url": "https://arxiv.org/abs/2503.21088", "pdf_url": "https://arxiv.org/pdf/2503.21088v2", "arxiv_id": "2503.21088", "doi": "10.48550/arXiv.2503.21088", "citation_count": 3, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/zjunlp/unlearn/tree/main/semeval25", "venue": null, "quality_score": 0.1505}
{"id": "7b7a7fd7ec9e2089194ea829fbc63a35a860763886394ce23097f643bff2fd88", "sources": ["arxiv", "semantic_scholar"], "title": "Reinforced Model Merging", "abstract": "The success of large language models has garnered widespread attention for model merging techniques, especially training-free methods which combine model capabilities within the parameter space. However, two challenges remain: (1) uniform treatment of all parameters leads to performance degradation; (2) search-based algorithms are often inefficient. In this paper, we present an innovative framework termed Reinforced Model Merging (RMM), which encompasses an environment and agent tailored for merging tasks. These components interact to execute layer-wise merging actions, aiming to search the optimal merging architecture. Notably, RMM operates without any gradient computations on the original models, rendering it feasible for edge devices. Furthermore, by utilizing data subsets during the evaluation process, we addressed the bottleneck in the reward feedback phase, thereby accelerating RMM by up to 100 times. Extensive experiments demonstrate that RMM achieves state-of-the-art performance across various vision and NLP datasets and effectively overcomes the limitations of the existing baseline methods. Our code is available at https://github.com/WuDiHJQ/Reinforced-Model-Merging.", "authors": ["Jiaqi Han", "Jingwen Ye", "Shunyu Liu", "Haofei Zhang", "Jie Song", "Zunlei Feng", "Mingli Song"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-27", "url": "https://arxiv.org/abs/2503.21272", "pdf_url": "https://arxiv.org/pdf/2503.21272v1", "arxiv_id": "2503.21272", "doi": "10.1109/ICME59968.2025.11210035", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/WuDiHJQ/Reinforced-Model-Merging", "venue": "IEEE International Conference on Multimedia and Expo", "quality_score": 0.1611}
{"id": "7fdf0344ce426195621bdfc8fca0ebeb65ee0f508d764eace7ef8f0429e91a3c", "sources": ["arxiv", "semantic_scholar"], "title": "Model Assembly Learning with Heterogeneous Layer Weight Merging", "abstract": "Model merging acquires general capabilities without extra data or training by combining multiple models' parameters. Previous approaches achieve linear mode connectivity by aligning parameters into the same loss basin using permutation invariance. In this paper, we introduce Model Assembly Learning (MAL), a novel paradigm for model merging that iteratively integrates parameters from diverse models in an open-ended model zoo to enhance the base model's capabilities. Unlike previous works that require identical architectures, MAL allows the merging of heterogeneous architectures and selective parameters across layers. Specifically, the base model can incorporate parameters from different layers of multiple pre-trained models. We systematically investigate the conditions and fundamental settings of heterogeneous parameter merging, addressing all possible mismatches in layer widths between the base and target models. Furthermore, we establish key laws and provide practical guidelines for effectively implementing MAL.", "authors": ["Yi-Kai Zhang", "Jin Wang", "Xu-Xiang Zhong", "De-Chuan Zhan", "Han-Jia Ye"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-27", "url": "https://arxiv.org/abs/2503.21657", "pdf_url": "https://arxiv.org/pdf/2503.21657v1", "arxiv_id": "2503.21657", "doi": "10.48550/arXiv.2503.21657", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1505}
{"id": "87c45d304248e35cfe3b7b1c97d55b41d0361f1c2b6b8a0b48e7a26e72791290", "sources": ["arxiv", "semantic_scholar"], "title": "Unlocking Efficient Long-to-Short LLM Reasoning with Model Merging", "abstract": "The transition from System 1 to System 2 reasoning in large language models (LLMs) has marked significant advancements in handling complex tasks through deliberate, iterative thinking. However, this progress often comes at the cost of efficiency, as models tend to overthink, generating redundant reasoning steps without proportional improvements in output quality. Long-to-Short (L2S) reasoning has emerged as a promising solution to this challenge, aiming to balance reasoning depth with practical efficiency. While existing approaches, such as supervised fine-tuning (SFT), reinforcement learning (RL), and prompt engineering, have shown potential, they are either computationally expensive or unstable. Model merging, on the other hand, offers a cost-effective and robust alternative by integrating the quick-thinking capabilities of System 1 models with the methodical reasoning of System 2 models. In this work, we present a comprehensive empirical study on model merging for L2S reasoning, exploring diverse methodologies, including task-vector-based, SVD-based, and activation-informed merging. Our experiments reveal that model merging can reduce average response length by up to 55% while preserving or even improving baseline performance. We also identify a strong correlation between model scale and merging efficacy with extensive evaluations on 1.5B/7B/14B/32B models. Furthermore, we investigate the merged model's ability to self-critique and self-correct, as well as its adaptive response length based on task complexity. Our findings highlight model merging as a highly efficient and effective paradigm for L2S reasoning, offering a practical solution to the overthinking problem while maintaining the robustness of System 2 reasoning. This work can be found on Github https://github.com/hahahawu/Long-to-Short-via-Model-Merging.", "authors": ["Han Wu", "Yuxuan Yao", "Shuqi Liu", "Zehua Liu", "Xiaojin Fu", "Xiongwei Han", "Xing Li", "Hui-Ling Zhen", "Tao Zhong", "Mingxuan Yuan"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-26", "url": "https://arxiv.org/abs/2503.20641", "pdf_url": "https://arxiv.org/pdf/2503.20641v2", "arxiv_id": "2503.20641", "doi": "10.48550/arXiv.2503.20641", "citation_count": 55, "influential_citation_count": 6, "has_code": true, "code_url": "https://github.com/hahahawu/Long-to-Short-via-Model-Merging", "venue": "arXiv.org", "quality_score": 0.437}
{"id": "ca379ba94b4b9f61139dfc0365218b816dd9d0723d608bc357b9d6d5b8591f4f", "sources": ["arxiv", "semantic_scholar"], "title": "On Symmetries in Convolutional Weights", "abstract": "We explore the symmetry of the mean k x k weight kernel in each layer of various convolutional neural networks. Unlike individual neurons, the mean kernels in internal layers tend to be symmetric about their centers instead of favoring specific directions. We investigate why this symmetry emerges in various datasets and models, and how it is impacted by certain architectural choices. We show how symmetry correlates with desirable properties such as shift and flip consistency, and might constitute an inherent inductive bias in convolutional neural networks.", "authors": ["Bilal Alsallakh", "Timothy Wroge", "Vivek Miglani", "Narine Kokhlikyan"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-24", "url": "https://arxiv.org/abs/2503.19215", "pdf_url": "https://arxiv.org/pdf/2503.19215v2", "arxiv_id": "2503.19215", "doi": "10.48550/arXiv.2503.19215", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1008}
{"id": "bf6e58b5837d04185ad09b81c747cc3f7d37baf2fae428763398a4b7d6089a0e", "sources": ["arxiv", "semantic_scholar"], "title": "MaTVLM: Hybrid Mamba-Transformer for Efficient Vision-Language Modeling", "abstract": "With the advancement of RNN models with linear complexity, the quadratic complexity challenge of transformers has the potential to be overcome. Notably, the emerging Mamba-2 has demonstrated competitive performance, bridging the gap between RNN models and transformers. However, due to sequential processing and vanishing gradients, RNN models struggle to capture long-range dependencies, limiting contextual understanding. This results in slow convergence, high resource demands, and poor performance on downstream understanding and complex reasoning tasks. In this work, we present a hybrid model MaTVLM by substituting a portion of the transformer decoder layers in a pre-trained VLM with Mamba-2 layers. Leveraging the inherent relationship between attention and Mamba-2, we initialize Mamba-2 with corresponding attention weights to accelerate convergence. Subsequently, we employ a single-stage distillation process, using the pre-trained VLM as the teacher model to transfer knowledge to the MaTVLM, further enhancing convergence speed and performance. Furthermore, we investigate the impact of differential distillation loss within our training framework. We evaluate the MaTVLM on multiple benchmarks, demonstrating competitive performance against the teacher model and existing VLMs while surpassing both Mamba-based VLMs and models of comparable parameter scales. Remarkably, the MaTVLM achieves up to 3.6x faster inference than the teacher model while reducing GPU memory consumption by 27.5%, all without compromising performance. Code and models are released at http://github.com/hustvl/MaTVLM.", "authors": ["Yingyue Li", "Bencheng Liao", "Wenyu Liu", "Xinggang Wang"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-17", "url": "https://arxiv.org/abs/2503.13440", "pdf_url": "https://arxiv.org/pdf/2503.13440v2", "arxiv_id": "2503.13440", "doi": "10.1109/ICCV51701.2025.01941", "citation_count": 7, "influential_citation_count": 0, "has_code": true, "code_url": "http://github.com/hustvl/MaTVLM", "venue": "IEEE International Conference on Computer Vision", "quality_score": 0.2258}
{"id": "04935d2a65a45edfd8ecda46daf7181a464f6db457c67778985452cb3fc315f0", "sources": ["arxiv", "semantic_scholar"], "title": "FW-Merging: Scaling Model Merging with Frank-Wolfe Optimization", "abstract": "Model merging has emerged as a promising approach for multi-task learning (MTL), offering a data-efficient alternative to conventional fine-tuning. However, with the rapid development of the open-source AI ecosystem and the increasing availability of fine-tuned foundation models, existing model merging methods face two key limitations: (i) They are primarily designed for in-house fine-tuned models, making them less adaptable to diverse model sources with partially unknown model and task information, (ii) They struggle to scale effectively when merging numerous model checkpoints. To address these challenges, we formulate model merging as a constrained optimization problem and introduce a novel approach: Frank-Wolfe Merging (FW-Merging). Inspired by Frank-Wolfe optimization, our approach iteratively selects the most relevant model in the pool to minimize a linear approximation of the objective function and then executes a local merging similar to the Frank-Wolfe update. The objective function is designed to capture the desired behavior of the target-merged model, while the fine-tuned candidate models define the constraint set. More importantly, FW-Merging serves as an orthogonal technique for existing merging methods, seamlessly integrating with them to further enhance accuracy performance. Our experiments show that FW-Merging scales across diverse model sources, remaining stable with 16 irrelevant models and improving by 15.3% with 16 relevant models on 20 CV tasks, while maintaining constant memory overhead, unlike the linear overhead of data-informed merging methods. Compared with the state-of-the-art approaches, FW-Merging surpasses the data-free merging method by 32.8% and outperforms the data-informed Adamerging by 8.39% when merging 20 ViT models. Our code is open-sourced at github.com/hmarkc/FW-Merging.", "authors": ["Hao Mark Chen", "Shell Xu Hu", "Wayne Luk", "Timothy Hospedales", "Hongxiang Fan"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-16", "url": "https://arxiv.org/abs/2503.12649", "pdf_url": "https://arxiv.org/pdf/2503.12649v3", "arxiv_id": "2503.12649", "doi": "10.1109/ICCV51701.2025.00324", "citation_count": 6, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "IEEE International Conference on Computer Vision", "quality_score": 0.2113}
{"id": "55c7ddb1b8cb91039f0893caa681b7026276d841efc6983f4e39c834a1b2c2d3", "sources": ["arxiv", "semantic_scholar"], "title": "We Should Chart an Atlas of All the World's Models", "abstract": "Public model repositories now contain millions of models, yet most models remain undocumented and effectively lost. In this position paper, we advocate for charting the world's model population in a unified structure we call the Model Atlas: a graph that captures models, their attributes, and the weight transformations that connect them. The Model Atlas enables applications in model forensics, meta-ML research, and model discovery, challenging tasks given today's unstructured model repositories. However, because most models lack documentation, large atlas regions remain uncharted. Addressing this gap motivates new machine learning methods that treat models themselves as data, inferring properties such as functionality, performance, and lineage directly from their weights. We argue that a scalable path forward is to bypass the unique parameter symmetries that plague model weights. Charting all the world's models will require a community effort, and we hope its broad utility will rally researchers toward this goal.", "authors": ["Eliahu Horwitz", "Nitzan Kurer", "Jonathan Kahana", "Liel Amar", "Yedid Hoshen"], "categories": ["cs.LG", "cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-13", "url": "https://arxiv.org/abs/2503.10633", "pdf_url": "https://arxiv.org/pdf/2503.10633v2", "arxiv_id": "2503.10633", "doi": null, "citation_count": 20, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.3306}
{"id": "6f978cfdb1420133d942fcdc6264c73ed8c228ad5cbf4627c069cf233f85a760", "sources": ["arxiv", "semantic_scholar"], "title": "From Task-Specific Models to Unified Systems: A Review of Model Merging Approaches", "abstract": "Model merging has achieved significant success, with numerous innovative methods proposed to enhance capabilities by combining multiple models. However, challenges persist due to the lack of a unified framework for classification and systematic comparative analysis, leading to inconsistencies in terminologies and categorizations. Meanwhile, as an increasing number of fine-tuned models are publicly available, their original training data often remain inaccessible due to privacy concerns or intellectual property restrictions. This makes traditional multi-task learning based on shared training data impractical. In scenarios where direct access to training data is infeasible, merging model parameters to create a unified model with broad generalization across multiple domains becomes crucial, further underscoring the importance of model merging techniques. Despite the rapid progress in this field, a comprehensive taxonomy and survey summarizing recent advances and predicting future directions are still lacking. This paper addresses these gaps by establishing a new taxonomy of model merging methods, systematically comparing different approaches, and providing an overview of key developments. By offering a structured perspective on this evolving area, we aim to help newcomers quickly grasp the field's landscape and inspire further innovations.", "authors": ["Wei Ruan", "Tianze Yang", "Yifan Zhou", "Tianming Liu", "Jin Lu"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-12", "url": "https://arxiv.org/abs/2503.08998", "pdf_url": "https://arxiv.org/pdf/2503.08998v1", "arxiv_id": "2503.08998", "doi": "10.48550/arXiv.2503.08998", "citation_count": 11, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2698}
{"id": "299b09d182a58a23e29bb4201c32a9a0e1402680e46c9d4f14266f3181f5d06c", "sources": ["arxiv", "semantic_scholar"], "title": "Efficient Multi-Task Inferencing: Model Merging with Gromov-Wasserstein Feature Alignment", "abstract": "Automatic scoring of student responses enhances efficiency in education, but deploying a separate neural network for each task increases storage demands, maintenance efforts, and redundant computations. To address these challenges, this paper introduces the Gromov-Wasserstein Scoring Model Merging (GW-SMM) method, which merges models based on feature distribution similarities measured via the Gromov-Wasserstein distance. Our approach begins by extracting features from student responses using individual models, capturing both item-specific context and unique learned representations. The Gromov-Wasserstein distance then quantifies the similarity between these feature distributions, identifying the most compatible models for merging. Models exhibiting the smallest pairwise distances, typically in pairs or trios, are merged by combining only the shared layers preceding the classification head. This strategy results in a unified feature extractor while preserving separate classification heads for item-specific scoring. We validated our approach against human expert knowledge and a GPT-o1-based merging method. GW-SMM consistently outperformed both, achieving a higher micro F1 score, macro F1 score, exact match accuracy, and per-label accuracy. The improvements in micro F1 and per-label accuracy were statistically significant compared to GPT-o1-based merging (p=0.04, p=0.01). Additionally, GW-SMM reduced storage requirements by half without compromising much accuracy, demonstrating its computational efficiency alongside reliable scoring performance.", "authors": ["Luyang Fang", "Ehsan Latif", "Haoran Lu", "Yifan Zhou", "Ping Ma", "Xiaoming Zhai"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-12", "url": "https://arxiv.org/abs/2503.09774", "pdf_url": "https://arxiv.org/pdf/2503.09774v1", "arxiv_id": "2503.09774", "doi": "10.48550/arXiv.2503.09774", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.0753}
{"id": "7adf5551a3178ef3bad221ec190070523b8a37383ea0edd54441a66bf9942162", "sources": ["arxiv", "semantic_scholar"], "title": "Whoever Started the Interference Should End It: Guiding Data-Free Model Merging via Task Vectors", "abstract": "Model merging seeks to integrate task-specific expert models into a unified architecture while preserving multi-task generalization capabilities, yet parameter interference between constituent models frequently induces performance degradation. Although prior work has explored many merging strategies, resolving interference without additional data for retraining or test-time computation remains challenging. In this paper, we theoretically demonstrate that the task vectors of the linear layer constitute an approximate linear subspace for its corresponding input. Therefore, we can minimize interference under the guidance of task vectors. Based on this insight, we propose \\textbf{WUDI-Merging} (\\textbf{W}hoever started the interference sho\\textbf{U}ld en\\textbf{D} \\textbf{I}t), a simple yet effective model merging method that eliminates interference without any additional data or rescaling coefficients. Comprehensive empirical evaluations across vision and language benchmarks demonstrate our method's superiority, achieving state-of-the-art performance in data-free model merging scenarios (average 10.9\\% improvement versus baseline methods) while even outperforming mainstream test-time adaptation approaches by 3.3\\%, and only very few computing resources are required. The code will be publicly available soon.", "authors": ["Runxi Cheng", "Feng Xiong", "Yongxian Wei", "Wanyun Zhu", "Chun Yuan"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-11", "url": "https://arxiv.org/abs/2503.08099", "pdf_url": "https://arxiv.org/pdf/2503.08099v2", "arxiv_id": "2503.08099", "doi": "10.48550/arXiv.2503.08099", "citation_count": 43, "influential_citation_count": 15, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.6021}
{"id": "58fa5ff87df138bfc7c022c629eed136165f158598c532c345d84c3e8833e5fa", "sources": ["arxiv", "semantic_scholar"], "title": "Task Vector Quantization for Memory-Efficient Model Merging", "abstract": "Model merging enables efficient multi-task models by combining task-specific fine-tuned checkpoints. However, storing multiple task-specific checkpoints requires significant memory, limiting scalability and restricting model merging to larger models and diverse tasks. In this paper, we propose quantizing task vectors (i.e., the difference between pre-trained and fine-tuned checkpoints) instead of quantizing fine-tuned checkpoints. We observe that task vectors exhibit a narrow weight range, enabling low precision quantization (e.g., 4 bit) within existing task vector merging frameworks. To further mitigate quantization errors within ultra-low bit precision (e.g., 2 bit), we introduce Residual Task Vector Quantization, which decomposes the task vector into a base vector and offset component. We allocate bits based on quantization sensitivity, ensuring precision while minimizing error within a memory budget. Experiments on image classification and dense prediction show our method maintains or improves model merging performance while using only 8% of the memory required for full-precision checkpoints.", "authors": ["Youngeun Kim", "Seunghwan Lee", "Aecheon Jung", "Bogon Ryu", "Sungeun Hong"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-10", "url": "https://arxiv.org/abs/2503.06921", "pdf_url": "https://arxiv.org/pdf/2503.06921v2", "arxiv_id": "2503.06921", "doi": "10.1109/ICCV51701.2025.01870", "citation_count": 6, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "IEEE International Conference on Computer Vision", "quality_score": 0.2113}
{"id": "98b91f41089a56b6ec6b92186450b710ffec8d9faf4f2b43c2dde4c08810ab33", "sources": ["arxiv", "semantic_scholar"], "title": "To See a World in a Spark of Neuron: Disentangling Multi-task Interference for Training-free Model Merging", "abstract": "Fine-tuning pre-trained models on targeted datasets enhances task-specific performance but often comes at the expense of generalization. Model merging techniques, which integrate multiple fine-tuned models into a single multi-task model through task arithmetic, offer a promising solution. However, task interference remains a fundamental challenge, leading to performance degradation and suboptimal merged models. Existing approaches largely overlooked the fundamental roles of neurons, their connectivity, and activation, resulting in a merging process and a merged model that does not consider how neurons relay and process information. In this work, we present the first study that relies on neuronal mechanisms for model merging. Specifically, we decomposed task-specific representations into two complementary neuronal subspaces that regulate input sensitivity and task adaptability. Leveraging this decomposition, we introduced NeuroMerging, a novel merging framework developed to mitigate task interference within neuronal subspaces, enabling training-free model fusion across diverse tasks. Through extensive experiments, we demonstrated that NeuroMerging achieved superior performance compared to existing methods on multi-task benchmarks across both natural language and vision domains. Our findings highlighted the importance of aligning neuronal mechanisms in model merging, offering new insights into mitigating task interference and improving knowledge fusion. Our project is available at https://ZzzitaoFang.github.io/projects/NeuroMerging/.", "authors": ["Zitao Fang", "Guodong DU", "Shuyang Yu", "Yifei Guo", "Yiwei Zhang", "Yiyao Cao", "Jing Li", "Ho-Kin Tang", "Sim Kuan Goh"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-07", "url": "https://arxiv.org/abs/2503.05320", "pdf_url": "https://arxiv.org/pdf/2503.05320v4", "arxiv_id": "2503.05320", "doi": "10.18653/v1/2025.emnlp-main.793", "citation_count": 4, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.1747}
{"id": "fb477b5ed70621cc2376935f6c1db057250764b525814ffb8330f5c4bd73b4a1", "sources": ["arxiv", "semantic_scholar"], "title": "Can We Optimize Deep RL Policy Weights as Trajectory Modeling?", "abstract": "Learning the optimal policy from a random network initialization is the theme of deep Reinforcement Learning (RL). As the scale of DRL training increases, treating DRL policy network weights as a new data modality and exploring the potential becomes appealing and possible. In this work, we focus on the policy learning path in deep RL, represented by the trajectory of network weights of historical policies, which reflects the evolvement of the policy learning process. Taking the idea of trajectory modeling with Transformer, we propose Transformer as Implicit Policy Learner (TIPL), which processes policy network weights in an autoregressive manner. We collect the policy learning path data by running independent RL training trials, with which we then train our TIPL model. In the experiments, we demonstrate that TIPL is able to fit the implicit dynamics of policy learning and perform the optimization of policy network by inference.", "authors": ["Hongyao Tang"], "categories": ["cs.LG", "cs.AI", "cs.NE"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-06", "url": "https://arxiv.org/abs/2503.04074", "pdf_url": "https://arxiv.org/pdf/2503.04074v1", "arxiv_id": "2503.04074", "doi": "10.48550/arXiv.2503.04074", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0802}
{"id": "2a498a28148ec00f2ac33cb6607c1497bb6509e16e5ae0241072960e60f0c2a0", "sources": ["arxiv", "semantic_scholar"], "title": "LEWIS (LayEr WIse Sparsity) -- A Training Free Guided Model Merging Approach", "abstract": "As specialized large language models (LLMs) become increasingly prevalent, model merging methods are being used to combine them to create a single multi-task model without requiring any additional data or training. However, these approaches fall short when the objective of merging is to increase the downstream model's performance on a particular task-specific benchmark. In this work, we propose LEWIS (Layer Wise Sparsity), a guided model-merging framework that uses activation-based layer importance to dynamically adjust layer-wise task-vector sparsity required for the merge process. LEWIS uses a calibration dataset to prioritize critical layers during the task-vector pruning process required for model merging. This approach guides existing merging methods by preserving essential layer-wise task-specific knowledge while ensuring the merged model performs the best at benchmarks resembling the calibration dataset. Our experiments demonstrate the effectiveness of LEWIS with performance improvements of code instruction-following and math-solving models created through model merging up to 4 percent and 11.3 percent, respectively, outperforming unguided data-less model merging approaches that use uniform-sparsity.", "authors": ["Hetarth Chopra", "Vidhi Rambhia", "Vikram Adve"], "categories": ["cs.LG", "cs.CL", "stat.ML"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2025-03-05", "url": "https://arxiv.org/abs/2503.03874", "pdf_url": "https://arxiv.org/pdf/2503.03874v2", "arxiv_id": "2503.03874", "doi": "10.48550/arXiv.2503.03874", "citation_count": 2, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1505}
{"id": "abba09f006047a6ad279aa8a7fbba410b6c06d7843f1416ab9d2a9748551c526", "sources": ["arxiv", "semantic_scholar"], "title": "Multi-Level Collaboration in Model Merging", "abstract": "Parameter-level model merging is an emerging paradigm in multi-task learning with significant promise. Previous research has explored its connections with prediction-level model ensembling-commonly viewed as the upper bound for merging-to reveal the potential of achieving performance consistency between the two. However, this observation relies on certain preconditions, such as being limited to two models, using ViT-based models, and all models are fine-tuned from the same pre-trained checkpoint. To further understand the intrinsic connections between model merging and model ensembling, this paper explores an interesting possibility: If these restrictions are removed, can performance consistency still be achieved between merging and ensembling? To answer this question, we first theoretically establish a performance correlation between merging and ensembling. We find that even when previous restrictions are not met, there is still a way for model merging to attain a near-identical and superior performance similar to that of ensembling. To verify whether our findings are practical, we introduce a validation framework termed Neural Ligand (NeuLig). The learning process of NeuLig is meticulously designed with a specialized loss function supported by theoretical foundations. Experimental results demonstrate the robust resilience of NeuLig in terms of both model scale and the number of collaborating models. For instance, for the case involving 5 CLIP-ViT-B/32 models, parameter-level merging achieves the same performance as prediction-level ensembling (merging: 95.44% vs. ensembling: 95.46%).", "authors": ["Qi Li", "Runpeng Yu", "Xinchao Wang"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-03", "url": "https://arxiv.org/abs/2503.01268", "pdf_url": "https://arxiv.org/pdf/2503.01268v1", "arxiv_id": "2503.01268", "doi": "10.48550/arXiv.2503.01268", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0768}
{"id": "69aa1232cf475aada58b3d44dc5c7445303bff97d213bc68ba0213c15d3d9172", "sources": ["arxiv", "semantic_scholar"], "title": "Layer-Aware Task Arithmetic: Disentangling Task-Specific and Instruction-Following Knowledge", "abstract": "Large language models (LLMs) demonstrate strong task-specific capabilities through fine-tuning, but merging multiple fine-tuned models often leads to degraded performance due to overlapping instruction-following components. Task Arithmetic (TA), which combines task vectors derived from fine-tuning, enables multi-task learning and task forgetting but struggles to isolate task-specific knowledge from general instruction-following behavior. To address this, we propose Layer-Aware Task Arithmetic (LATA), a novel approach that assigns layer-specific weights to task vectors based on their alignment with instruction-following or task-specific components. By amplifying task-relevant layers and attenuating instruction-following layers, LATA improves task learning and forgetting performance while preserving overall model utility. Experiments on multiple benchmarks, including WikiText-2, GSM8K, and HumanEval, demonstrate that LATA outperforms existing methods in both multi-task learning and selective task forgetting, achieving higher task accuracy and alignment with minimal degradation in output quality. Our findings highlight the importance of layer-wise analysis in disentangling task-specific and general-purpose knowledge, offering a robust framework for efficient model merging and editing.", "authors": ["Yan-Lun Chen", "Yi-Ru Wei", "Chia-Yi Hsu", "Chia-Mu Yu", "Chun-Ying Huang", "Ying-Dar Lin", "Yu-Sung Wu", "Wei-Bin Lee"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-27", "url": "https://arxiv.org/abs/2502.20186", "pdf_url": "https://arxiv.org/pdf/2502.20186v1", "arxiv_id": "2502.20186", "doi": "10.48550/arXiv.2502.20186", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.1193}
{"id": "e1b9c9c30b169caa7631c02b9b65e1d093a20d89da22f1841403e0aafd1d0389", "sources": ["arxiv", "semantic_scholar"], "title": "In-Model Merging for Enhancing the Robustness of Medical Imaging Classification Models", "abstract": "Model merging is an effective strategy to merge multiple models for enhancing model performances, and more efficient than ensemble learning as it will not introduce extra computation into inference. However, limited research explores if the merging process can occur within one model and enhance the model's robustness, which is particularly critical in the medical image domain. In the paper, we are the first to propose in-model merging (InMerge), a novel approach that enhances the model's robustness by selectively merging similar convolutional kernels in the deep layers of a single convolutional neural network (CNN) during the training process for classification. We also analytically reveal important characteristics that affect how in-model merging should be performed, serving as an insightful reference for the community. We demonstrate the feasibility and effectiveness of this technique for different CNN architectures on 4 prevalent datasets. The proposed InMerge-trained model surpasses the typically-trained model by a substantial margin. The code will be made public.", "authors": ["Hu Wang", "Ibrahim Almakky", "Congbo Ma", "Numan Saeed", "Mohammad Yaqub"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-27", "url": "https://arxiv.org/abs/2502.20516", "pdf_url": "https://arxiv.org/pdf/2502.20516v2", "arxiv_id": "2502.20516", "doi": "10.48550/arXiv.2502.20516", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1193}
{"id": "048bae30639f2d76a1740661b58a65ad68911c1b26052e37167c851708d4ea1d", "sources": ["arxiv", "semantic_scholar"], "title": "CABS: Conflict-Aware and Balanced Sparsification for Enhancing Model Merging", "abstract": "Model merging based on task vectors, i.e., the parameter differences between fine-tuned models and a shared base model, provides an efficient way to integrate multiple task-specific models into a multitask model without retraining. Recent works have endeavored to address the conflicts between task vectors, one of the significant challenges faced by model merging, through sparsification; however, two issues significantly limit their performance: high parameter overlap and unbalanced weight distribution. To address these issues, we propose a simple, yet effective framework called CABS (Conflict-Aware and Balanced Sparsification), consisting of Conflict-Aware Sparsification (CA) and Balanced Sparsification (BS). CA can reduce parameter overlap by applying masks during sequential pruning, ensuring that each task vector retains distinct, non-overlapping parameters. BS leverages $n$: $m$ pruning to preserve critical weights while maintaining an even distribution across layers. Our comprehensive experiments demonstrate that CABS outperforms state-of-the-art methods across diverse tasks and model sizes.", "authors": ["Zongzhen Yang", "Binhang Qi", "Hailong Sun", "Wenrui Long", "Ruobing Zhao", "Xiang Gao"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-26", "url": "https://arxiv.org/abs/2503.01874", "pdf_url": "https://arxiv.org/pdf/2503.01874v1", "arxiv_id": "2503.01874", "doi": "10.48550/arXiv.2503.01874", "citation_count": 9, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.25}
{"id": "d27393970753e4c523fb9694e3516d3a20461878c789564cd17b8d620c028afc", "sources": ["arxiv", "semantic_scholar"], "title": "DICEPTION: A Generalist Diffusion Model for Visual Perceptual Tasks", "abstract": "This paper's primary objective is to develop a robust generalist perception model capable of addressing multiple tasks under constraints of computational resources and limited training data. We leverage text-to-image diffusion models pre-trained on billions of images and successfully introduce our DICEPTION, a visual generalist model. Exhaustive evaluations demonstrate that DICEPTION effectively tackles diverse perception tasks, even achieving performance comparable to SOTA single-task specialist models. Specifically, we achieve results on par with SAM-vit-h using only 0.06% of their data (e.g., 600K vs.\\ 1B pixel-level annotated images). We designed comprehensive experiments on architectures and input paradigms, demonstrating that the key to successfully re-purposing a single diffusion model for multiple perception tasks lies in maximizing the preservation of the pre-trained model's prior knowledge. Consequently, DICEPTION can be trained with substantially lower computational costs than conventional models requiring training from scratch. Furthermore, adapting DICEPTION to novel tasks is highly efficient, necessitating fine-tuning on as few as 50 images and approximately 1% of its parameters. Finally, we demonstrate that a subtle application of classifier-free guidance can improve the model's performance on depth and normal estimation. We also show that pixel-aligned training, as is characteristic of perception tasks, significantly enhances the model's ability to preserve fine details. DICEPTION offers valuable insights and presents a promising direction for the development of advanced diffusion-based visual generalist models. Code and Model: https://github.com/aim-uofa/Diception", "authors": ["Canyu Zhao", "Yanlong Sun", "Mingyu Liu", "Huanyi Zheng", "Muzhi Zhu", "Zhiyue Zhao", "Hao Chen", "Tong He", "Chunhua Shen"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-24", "url": "https://arxiv.org/abs/2502.17157", "pdf_url": "https://arxiv.org/pdf/2502.17157v3", "arxiv_id": "2502.17157", "doi": "10.48550/arXiv.2502.17157", "citation_count": 34, "influential_citation_count": 6, "has_code": true, "code_url": "https://github.com/aim-uofa/Diception", "venue": "arXiv.org", "quality_score": 0.4225}
{"id": "d92e5bfcdeb496788c26ffa524b4cfdb23c3381597bfcb47a8482429b5fd67f8", "sources": ["arxiv", "semantic_scholar"], "title": "RobustMerge: Parameter-Efficient Model Merging for MLLMs with Direction Robustness", "abstract": "Fine-tuning pre-trained models with custom data leads to numerous expert models on specific tasks. Merging models into one universal model to empower multi-task ability refraining from data leakage has gained popularity. With the expansion in data and model size, parameter-efficient tuning becomes the common practice for obtaining task-specific models efficiently. However, few methods are dedicated to efficient merging, and existing methods designed for full fine-tuning merging fail under efficient merging. To address the issue, we analyze from low-rank decomposition and reveal that direction robustness during merging is crucial for merging efficient modules. We furthermore uncover that compensating for the gap between stark singular values contributes to direction robustness. Therefore, we propose RobustMerge, a training-free parameter-efficient merging method with complementary parameter adaptation to maintain direction robustness. Specifically, we (1) prune parameters and scale coefficients from inter-parameter relation for singular values to maintain direction stability away from task interference, and (2) perform cross-task normalization to enhance unseen task generalization. We establish a benchmark consisting of diverse multimodal tasks, on which we conduct experiments to certify the outstanding performance and generalizability of our method. Additional studies and extensive analyses further showcase the effectiveness. Code is available at https://github.com/AuroraZengfh/RobustMerge.", "authors": ["Fanhu Zeng", "Haiyang Guo", "Fei Zhu", "Li Shen", "Hao Tang"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-24", "url": "https://arxiv.org/abs/2502.17159", "pdf_url": "https://arxiv.org/pdf/2502.17159v6", "arxiv_id": "2502.17159", "doi": null, "citation_count": 14, "influential_citation_count": 3, "has_code": true, "code_url": "https://github.com/AuroraZengfh/RobustMerge", "venue": null, "quality_score": 0.301}
{"id": "ca8814157ff1354d02a8af6186b8d6fb2d07637ea028dc63a9142f0ce94762ae", "sources": ["arxiv", "semantic_scholar"], "title": "Mixup Model Merge: Enhancing Model Merging Performance through Randomized Linear Interpolation", "abstract": "Model merging aims to integrate multiple task-specific models into a unified model that inherits the capabilities of the task-specific models, without additional training. Existing model merging methods often lack consideration of the varying contribution ratios of different task-specific models to the final merged model. In this paper, we propose Mixup Model Merge (M3), a simple yet effective method inspired by the randomized linear interpolation strategy from the Mixup data augmentation technique. M3 performs randomized linear interpolation in parameter space between two task-specific LLMs, where interpolation coefficients are sampled from a Beta distribution to explore diverse contribution ratios. This controllable randomness allows M3 to outperform standard equal-ratio merging by discovering better contribution ratio combinations. Extensive experiments show that M3 significantly (1) improves merged LLM performance across tasks, (2) enhances out-of-distribution and adversarial robustness, (3) outperforms the positive effects of the sparsification method DARE on model merging and can be further combined with DARE to achieve superior results, and (4) balances exploration efficiency and diversity in contribution ratios by tuning the Beta distribution's shape parameters. The code is provided in the supplementary materials.", "authors": ["Yue Zhou", "Yi Chang", "Yuan Wu"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-21", "url": "https://arxiv.org/abs/2502.15434", "pdf_url": "https://arxiv.org/pdf/2502.15434v3", "arxiv_id": "2502.15434", "doi": "10.48550/arXiv.2502.15434", "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1945}
{"id": "2996f1d045eef7cac771e1b013346dd411cb14f90b78d22e5bb28dd5f720db95", "sources": ["arxiv", "semantic_scholar"], "title": "Speech-FT: Merging Pre-trained And Fine-Tuned Speech Representation Models For Cross-Task Generalization", "abstract": "Fine-tuning speech representation models can enhance performance on specific tasks but often compromises their cross-task generalization ability. This degradation is often caused by excessive changes in the representations, making it difficult to retain information learned during pre-training. Existing approaches, such as regularizing weight changes during fine-tuning, may fail to maintain sufficiently high feature similarity with the pre-trained model, and thus could possibly lose cross-task generalization. To address this issue, we propose Speech-FT, a novel two-stage fine-tuning framework designed to maintain cross-task generalization while benefiting from fine-tuning. Speech-FT first applies fine-tuning specifically designed to reduce representational drift, followed by weight-space interpolation with the pre-trained model to restore cross-task generalization. Extensive experiments on HuBERT, wav2vec 2.0, DeCoAR 2.0, and WavLM Base+ demonstrate that Speech-FT consistently improves performance across a wide range of supervised, unsupervised, and multitask fine-tuning scenarios. Moreover, Speech-FT achieves superior cross-task generalization compared to fine-tuning baselines that explicitly constrain weight changes, such as weight-space regularization and LoRA fine-tuning. Our analysis reveals that Speech-FT maintains higher feature similarity to the pre-trained model compared to alternative strategies, despite allowing larger weight-space updates. Notably, Speech-FT achieves significant improvements on the SUPERB benchmark. For example, when fine-tuning HuBERT on automatic speech recognition, Speech-FT is able to reduce phone error rate from 5.17% to 3.94%, lower word error rate from 6.38% to 5.75%, and increase speaker identification accuracy from 81.86% to 84.11%. Speech-FT provides a simple yet powerful solution for further refining speech representation models after pre-training.", "authors": ["Tzu-Quan Lin", "Wei-Ping Huang", "Hao Tang", "Hung-yi Lee"], "categories": ["cs.CL", "cs.AI", "cs.SD"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-18", "url": "https://arxiv.org/abs/2502.12672", "pdf_url": "https://arxiv.org/pdf/2502.12672v4", "arxiv_id": "2502.12672", "doi": "10.1109/TASLPRO.2025.3635827", "citation_count": 6, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/nervjack2/Speech-FT", "venue": "IEEE Transactions on Audio, Speech, and Language Processing", "quality_score": 0.2113}
{"id": "85eb0624a22ae2d09c5570020b7ef57b9b91e454fec71a5526f9e36ec1c76b12", "sources": ["arxiv", "semantic_scholar"], "title": "Sens-Merging: Sensitivity-Guided Parameter Balancing for Merging Large Language Models", "abstract": "Recent advances in large language models have led to numerous task-specialized fine-tuned variants, creating a need for efficient model merging techniques that preserve specialized capabilities while avoiding costly retraining. While existing task vector-based merging methods show promise, they typically apply uniform coefficients across all parameters, overlooking varying parameter importance both within and across tasks. We present Sens-Merging, a sensitivity-guided coefficient adjustment method that enhances existing model merging techniques by operating at both task-specific and cross-task levels. Our method analyzes parameter sensitivity within individual tasks and evaluates cross-task transferability to determine optimal merging coefficients. Extensive experiments on Mistral 7B and LLaMA2-7B/13B models demonstrate that Sens-Merging significantly improves performance across general knowledge, mathematical reasoning, and code generation tasks. Notably, when combined with existing merging techniques, our method enables merged models to outperform specialized fine-tuned models, particularly in code generation tasks. Our findings reveal important trade-offs between task-specific and cross-task scalings, providing insights for future model merging strategies.", "authors": ["Shuqi Liu", "Han Wu", "Bowei He", "Xiongwei Han", "Mingxuan Yuan", "Linqi Song"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-18", "url": "https://arxiv.org/abs/2502.12420", "pdf_url": "https://arxiv.org/pdf/2502.12420v2", "arxiv_id": "2502.12420", "doi": "10.48550/arXiv.2502.12420", "citation_count": 14, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.294}
{"id": "55b3fdeefd4d2e736e6785bd7b576c2d94d4ef8ff63f8b6ed92048c0b0b03471", "sources": ["arxiv", "semantic_scholar"], "title": "Scalable Model Merging with Progressive Layer-wise Distillation", "abstract": "Model merging offers an effective way to integrate the capabilities of multiple fine-tuned models. However, the performance degradation of the merged model remains a challenge, particularly when none or few data are available. This paper first highlights the necessity of domain-specific data for model merging by proving that data-agnostic algorithms can have arbitrarily bad worst-case performance. Building on this theoretical insight, we explore the relationship between model merging and distillation, introducing a novel few-shot merging algorithm, ProDistill (Progressive Layer-wise Distillation). Unlike common belief that layer wise training hurts performance, we show that layer-wise teacher-student distillation not only enhances the scalability but also improves model merging performance. We conduct extensive experiments to show that compared to existing few-shot merging methods, ProDistill achieves state-of-the-art performance, with up to 6.14% and 6.61% improvements in vision and NLU tasks. Furthermore, we extend the experiments to models with over 10B parameters, showcasing the exceptional scalability of ProDistill.", "authors": ["Jing Xu", "Jiazheng Li", "Jingzhao Zhang"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-18", "url": "https://arxiv.org/abs/2502.12706", "pdf_url": "https://arxiv.org/pdf/2502.12706v2", "arxiv_id": "2502.12706", "doi": "10.48550/arXiv.2502.12706", "citation_count": 10, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.2603}
{"id": "231aa27f42e888ed320402895d630453b04adee4cac900c3b25356a8ea3504dd", "sources": ["arxiv", "semantic_scholar"], "title": "Bone Soups: A Seek-and-Soup Model Merging Approach for Controllable Multi-Objective Generation", "abstract": "User information needs are often highly diverse and varied. A key challenge in current research is how to achieve controllable multi-objective generation while enabling rapid adaptation to accommodate diverse user demands during test time. Existing solutions, such as Rewarded Soup, focus on merging language models individually tuned on single objectives. While easy to implement and widely used, these approaches face limitations in achieving optimal performance due to their disregard for the impacts of competing objectives on model tuning. To address this issue, we propose Bone Soup, a novel model merging approach that first seeks a series of backbone models by considering the impacts of multiple objectives and then makes the soup (i.e., merge the backbone models). Specifically, Bone Soup begins by training multiple backbone models for different objectives using multi-objective reinforcement learning. Each backbone model is guided by a combination of backbone reward signals. To ensure that these models are optimal for the Pareto front, the backbone rewards are crafted by combining standard reward functions into basis vectors, which can then be modified through a rule-based construction method. Bone Soup leverages a symmetric circulant matrix mapping to generate the merging coefficients, which are used to merge the backbone models according to user preferences. Extensive experimental results demonstrate that Bone Soup exhibits strong controllability and Pareto optimality in controllable multi-objective generation, providing a more effective and efficient approach to addressing diverse user needs at test time.", "authors": ["Guofu Xie", "Xiao Zhang", "Ting Yao", "Yunsheng Shi"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-15", "url": "https://arxiv.org/abs/2502.10762", "pdf_url": "https://arxiv.org/pdf/2502.10762v2", "arxiv_id": "2502.10762", "doi": "10.48550/arXiv.2502.10762", "citation_count": 9, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.25}
{"id": "2b4d3bfdca75ce0b21665753d88591c8ef0e5cb0903a6167328f69a7eaf02cbb", "sources": ["arxiv", "semantic_scholar"], "title": "Superpose Task-specific Features for Model Merging", "abstract": "Model merging enables powerful capabilities in neural networks without requiring additional training. In this paper, we introduce a novel perspective on model merging by leveraging the fundamental mechanisms of neural network representation. Our approach is motivated by the linear representation hypothesis, which states that neural networks encode information through linear combinations of feature vectors. We propose a method that superposes task-specific features from individual models into a merged model. Our approach specifically targets linear transformation matrices, which are crucial for feature activation and extraction in deep networks. By formulating the merging process as a linear system, we can preserve task-specific features from individual models and create merged models that effectively maintain multi-task capabilities compared to existing methods. Extensive experiments across diverse benchmarks and models demonstrate that our method outperforms existing techniques. Code is available at https://github.com/LARS-research/STF.", "authors": ["Haiquan Qiu", "You Wu", "Dong Li", "Jianmin Guo", "Quanming Yao"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-15", "url": "https://arxiv.org/abs/2502.10698", "pdf_url": "https://arxiv.org/pdf/2502.10698v2", "arxiv_id": "2502.10698", "doi": "10.18653/v1/2025.emnlp-main.210", "citation_count": 4, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/LARS-research/STF", "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.1747}
{"id": "34a9812394295e5882d875afa8163119bc0bec17c0716e5a06e456532a9e91c1", "sources": ["arxiv", "semantic_scholar"], "title": "LoRE-Merging: Exploring Low-Rank Estimation For Large Language Model Merging", "abstract": "While most current approaches rely on further training techniques, such as fine-tuning or reinforcement learning, to enhance model capacities, model merging stands out for its ability of improving models without requiring any additional training. In this paper, we propose a unified framework for model merging based on low-rank estimation of task vectors without the need for access to the base model, named \\textsc{LoRE-Merging}. Our approach is motivated by the observation that task vectors from fine-tuned models frequently exhibit a limited number of dominant singular values, making low-rank estimations less prone to interference. We implement the method by formulating the merging problem as an optimization problem. Extensive empirical experiments demonstrate the effectiveness of our framework in mitigating interference and preserving task-specific information, thereby advancing the state-of-the-art performance in model merging techniques.", "authors": ["Zehua Liu", "Han Wu", "Yuxuan Yao", "Ruifeng She", "Xiongwei Han", "Tao Zhong", "Mingxuan Yuan"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-15", "url": "https://arxiv.org/abs/2502.10749", "pdf_url": "https://arxiv.org/pdf/2502.10749v2", "arxiv_id": "2502.10749", "doi": "10.48550/arXiv.2502.10749", "citation_count": 8, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.2386}
{"id": "5fb540a1b6b172fcdc1e4a4aa0fa1e403bc26528a34a6d9af024142a18666b3e", "sources": ["arxiv", "semantic_scholar"], "title": "1bit-Merging: Dynamic Quantized Merging for Large Language Models", "abstract": "Recent advances in large language models have led to specialized models excelling in specific domains, creating a need for efficient model merging techniques. While traditional merging approaches combine parameters into a single static model, they often compromise task-specific performance. However, task-specific routing methods maintain accuracy but introduce substantial storage overhead. We present \\texttt{1bit}-Merging, a novel framework that integrates task-specific routing with 1-bit quantized task vectors to balance performance and storage efficiency. Our approach leverages the observation that different task-specific models store knowledge in distinct layers-chat models primarily in attention layers and math/code models in MLP layers, enabling targeted compression strategies. Through extensive experiments with LLaMA2 and Mistral model families across chat, mathematical reasoning, and code generation tasks, we demonstrate that 1bit-Merging achieves comparable or superior performance to existing methods while significantly reducing storage requirements. Our framework offers a practical solution for combining specialized models while maintaining their individual strengths and addressing the storage challenges of current approaches.", "authors": ["Shuqi Liu", "Yuxuan Yao", "Bowei He", "Zehua Liu", "Xiongwei Han", "Mingxuan Yuan", "Han Wu", "Linqi Song"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-15", "url": "https://arxiv.org/abs/2502.10743", "pdf_url": "https://arxiv.org/pdf/2502.10743v2", "arxiv_id": "2502.10743", "doi": "10.48550/arXiv.2502.10743", "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1945}
{"id": "1be2ad94c3aa1eff876f2d370ee8d6d4c83755e086b09421c5be53e67f0e1090", "sources": ["arxiv", "semantic_scholar"], "title": "STAR: Spectral Truncation and Rescale for Model Merging", "abstract": "Model merging is an efficient way of obtaining a multi-task model from several pretrained models without further fine-tuning, and it has gained attention in various domains, including natural language processing (NLP). Despite the efficiency, a key challenge in model merging is the seemingly inevitable decrease in task performance as the number of models increases. In this paper, we propose $\\mathbf{S}$pectral $\\mathbf{T}$runcation $\\mathbf{A}$nd $\\mathbf{R}$escale (STAR) that aims at mitigating ``merging conflicts'' by truncating small components in the respective spectral spaces, which is followed by an automatic parameter rescaling scheme to retain the nuclear norm of the original matrix. STAR requires no additional inference on original training data and is robust to hyperparamater choice. We demonstrate the effectiveness of STAR through extensive model merging cases on diverse NLP tasks. Specifically, STAR works robustly across varying model sizes, and can outperform baselines by 4.2$\\%$ when merging 12 models on Flan-T5. Our code is publicly available at https://github.com/IBM/STAR.", "authors": ["Yu-Ang Lee", "Ching-Yun Ko", "Tejaswini Pedapati", "I-Hsin Chung", "Mi-Yen Yeh", "Pin-Yu Chen"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-14", "url": "https://arxiv.org/abs/2502.10339", "pdf_url": "https://arxiv.org/pdf/2502.10339v1", "arxiv_id": "2502.10339", "doi": "10.48550/arXiv.2502.10339", "citation_count": 7, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/IBM/STAR", "venue": "North American Chapter of the Association for Computational Linguistics", "quality_score": 0.2386}
{"id": "bec8821be267ce3c1c0663439c228eb6409b9e958460f0b536f1948203a2b879", "sources": ["arxiv", "semantic_scholar"], "title": "Mix Data or Merge Models? Balancing the Helpfulness, Honesty, and Harmlessness of Large Language Model via Model Merging", "abstract": "Achieving balanced alignment of large language models (LLMs) in terms of Helpfulness, Honesty, and Harmlessness (3H optimization) constitutes a cornerstone of responsible AI. Existing methods like data mixture strategies face limitations, including heavy reliance on expert knowledge and conflicting optimization signals. While model merging offers parameter-level conflict-resolution strategies through integrating specialized models' parameters, its potential for 3H optimization remains underexplored. This paper systematically compares the effectiveness of model merging and data mixture methods in constructing 3H-aligned LLMs for the first time, revealing previously overlooked collaborative and conflict relationships among the 3H dimensions and discussing the advantages and drawbacks of data mixture (\\textit{data-level}) and model merging (\\textit{parameter-level}) methods in mitigating the conflict for balanced 3H optimization. Specially, we propose a novel \\textbf{R}eweighting \\textbf{E}nhanced task \\textbf{S}ingular \\textbf{M}erging method, \\textbf{RESM}, through outlier weighting and sparsity-aware rank selection strategies to address the challenges of preference noise accumulation and layer sparsity adaptation inherent in 3H-aligned LLM merging. Extensive evaluations can verify the effectiveness and robustness of RESM compared to previous data mixture (2\\%-5\\% gain) and model merging (1\\%-3\\% gain) methods in achieving balanced LLM alignment. We release our models through \\href{https://huggingface.co/Jinluan}{3H\\_Merging} for further investigations.", "authors": ["Jinluan Yang", "Dingnan Jin", "Anke Tang", "Li Shen", "Didi Zhu", "Zhengyu Chen", "Ziyu Zhao", "Daixin Wang", "Qing Cui", "Zhiqiang Zhang", "Jun Zhou", "Fei Wu", "Kun Kuang"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-08", "url": "https://arxiv.org/abs/2502.06876", "pdf_url": "https://arxiv.org/pdf/2502.06876v4", "arxiv_id": "2502.06876", "doi": "10.48550/arXiv.2502.06876", "citation_count": 20, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3306}
{"id": "d1e9332693cf07b4d231112b6a9be51ad40359e42ddf7636b9723c21e0b4ea53", "sources": ["arxiv", "semantic_scholar"], "title": "No Task Left Behind: Isotropic Model Merging with Common and Task-Specific Subspaces", "abstract": "Model merging integrates the weights of multiple task-specific models into a single multi-task model. Despite recent interest in the problem, a significant performance gap between the combined and single-task models remains. In this paper, we investigate the key characteristics of task matrices -- weight update matrices applied to a pre-trained model -- that enable effective merging. We show that alignment between singular components of task-specific and merged matrices strongly correlates with performance improvement over the pre-trained model. Based on this, we propose an isotropic merging framework that flattens the singular value spectrum of task matrices, enhances alignment, and reduces the performance gap. Additionally, we incorporate both common and task-specific subspaces to further improve alignment and performance. Our proposed approach achieves state-of-the-art performance on vision and language tasks across various sets of tasks and model scales. This work advances the understanding of model merging dynamics, offering an effective methodology to merge models without requiring additional training. Code is available at https://github.com/danielm1405/iso-merging .", "authors": ["Daniel Marczak", "Simone Magistri", "Sebastian Cygert", "Bartłomiej Twardowski", "Andrew D. Bagdanov", "Joost van de Weijer"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-07", "url": "https://arxiv.org/abs/2502.04959", "pdf_url": "https://arxiv.org/pdf/2502.04959v3", "arxiv_id": "2502.04959", "doi": "10.48550/arXiv.2502.04959", "citation_count": 73, "influential_citation_count": 15, "has_code": true, "code_url": "https://github.com/danielm1405/iso-merging", "venue": "International Conference on Machine Learning", "quality_score": 0.6021}
{"id": "97160a65ca30c1728715d9cf80baf854a00b8ac163ef561156e098b708e2c006", "sources": ["arxiv", "semantic_scholar"], "title": "Fine, I'll Merge It Myself: A Multi-Fidelity Framework for Automated Model Merging", "abstract": "Reasoning capabilities represent a critical frontier for large language models (LLMs), but developing them requires extensive proprietary datasets and computational resources. One way to efficiently supplement capabilities with is by model merging, which offers a promising alternative by combining multiple models without retraining. However, current merging approaches rely on manually-designed strategies for merging hyperparameters, limiting the exploration of potential model combinations and requiring significant human effort. We propose an Automated Model Merging Framework that enables fine-grained exploration of merging strategies while reducing costs through multi-fidelity approximations. We support both single and multi-objective optimization and introduce two novel search spaces: layerwise fusion (LFS) and depth-wise integration (DIS). Evaluating across a number of benchmarks, we find that the search autonomously finds 1) Merges that further boost single-objective performance, even on tasks the model has already been finetuned on, and 2) Merges that optimize multi-objective frontiers across tasks. Effective merges are found with limited compute, e.g. within less than 500 search steps.", "authors": ["Guinan Su", "Jonas Geiping"], "categories": ["cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-06", "url": "https://arxiv.org/abs/2502.04030", "pdf_url": "https://arxiv.org/pdf/2502.04030v2", "arxiv_id": "2502.04030", "doi": "10.48550/arXiv.2502.04030", "citation_count": 4, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "4204a9d45e5bfb8eee881ba5f76190d37e4afc65d55a2b4c20e0dc6d6095e014", "sources": ["arxiv", "semantic_scholar"], "title": "Activation-Informed Merging of Large Language Models", "abstract": "Model merging, a method that combines the parameters and embeddings of multiple fine-tuned large language models (LLMs), offers a promising approach to enhance model performance across various tasks while maintaining computational efficiency. This paper introduces Activation-Informed Merging (AIM), a technique that integrates the information from the activation space of LLMs into the merging process to improve performance and robustness. AIM is designed as a flexible, complementary solution that is applicable to any existing merging method. It aims to preserve critical weights from the base model, drawing on principles from continual learning (CL) and model compression. Utilizing a task-agnostic calibration set, AIM selectively prioritizes essential weights during merging. We empirically demonstrate that AIM significantly enhances the performance of merged models across multiple benchmarks. Our findings suggest that considering the activation-space information can provide substantial advancements in the model merging strategies for LLMs, with up to a 40% increase in benchmark performance.", "authors": ["Amin Heyrani Nobari", "Kaveh Alim", "Ali ArjomandBigdeli", "Akash Srivastava", "Faez Ahmed", "Navid Azizan"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-04", "url": "https://arxiv.org/abs/2502.02421", "pdf_url": "https://arxiv.org/pdf/2502.02421v3", "arxiv_id": "2502.02421", "doi": "10.48550/arXiv.2502.02421", "citation_count": 21, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3356}
{"id": "83b55dec490056686b9cb22f696edbee3ba608f4a6909e782e8a5df84dfacbc3", "sources": ["arxiv", "semantic_scholar"], "title": "Soup-of-Experts: Pretraining Specialist Models via Parameters Averaging", "abstract": "Machine learning models are routinely trained on a mixture of different data domains. Different domain weights yield very different downstream performances. We propose the Soup-of-Experts, a novel architecture that can instantiate a model at test time for any domain weights with minimal computational cost and without re-training the model. Our architecture consists of a bank of expert parameters, which are linearly combined to instantiate one model. We learn the linear combination coefficients as a function of the input domain weights. To train this architecture, we sample random domain weights, instantiate the corresponding model, and backprop through one batch of data sampled with these domain weights. We demonstrate how our approach obtains small specialized models on several language modeling tasks quickly. Soup-of-Experts are particularly appealing when one needs to ship many different specialist models quickly under a model size constraint.", "authors": ["Pierre Ablin", "Angelos Katharopoulos", "Skyler Seto", "David Grangier"], "categories": ["cs.LG", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-03", "url": "https://arxiv.org/abs/2502.01804", "pdf_url": "https://arxiv.org/pdf/2502.01804v1", "arxiv_id": "2502.01804", "doi": "10.48550/arXiv.2502.01804", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.1505}
{"id": "f4fab507cb05b21d1dca97ab210b13fb6662ce1b19fc356b8e7a985971d14394", "sources": ["arxiv", "semantic_scholar"], "title": "MergeME: Model Merging Techniques for Homogeneous and Heterogeneous MoEs", "abstract": "The recent success of specialized Large Language Models (LLMs) in domains such as mathematical reasoning and coding has led to growing interest in methods for merging these expert LLMs into a unified Mixture-of-Experts (MoE) model, with the goal of enhancing performance in each domain while retaining effectiveness on general tasks. However, the effective merging of expert models remains an open challenge, especially for models with highly divergent weight parameters or different architectures. State-of-the-art MoE merging methods only work with homogeneous model architectures and rely on simple unweighted averaging to merge expert layers, which does not address parameter interference and requires extensive fine-tuning of the merged MoE to restore performance. To address these limitations, this paper introduces new MoE merging techniques, including strategies to mitigate parameter interference, routing heuristics to reduce the need for MoE fine-tuning, and a novel method for merging experts with different architectures. Extensive experiments across multiple domains demonstrate the effectiveness of our proposed methods, reducing fine-tuning costs, improving performance over state-of-the-art methods, and expanding the applicability of MoE merging.", "authors": ["Yuhang Zhou", "Giannis Karamanolakis", "Victor Soto", "Anna Rumshisky", "Mayank Kulkarni", "Furong Huang", "Wei Ai", "Jianhua Lu"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-03", "url": "https://arxiv.org/abs/2502.00997", "pdf_url": "https://arxiv.org/pdf/2502.00997v3", "arxiv_id": "2502.00997", "doi": "10.48550/arXiv.2502.00997", "citation_count": 12, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "North American Chapter of the Association for Computational Linguistics", "quality_score": 0.2785}
{"id": "bae2473a51e2276d8c85e64d5a33b2fbcca143377de367a419084ffe381c068e", "sources": ["arxiv", "semantic_scholar"], "title": "Task Vector Bases: A Unified and Scalable Framework for Compressed Task Arithmetic", "abstract": "Task arithmetic, representing downstream tasks through linear operations on task vectors, has emerged as a simple yet powerful paradigm for transferring knowledge across diverse settings. However, maintaining a large collection of task vectors introduces scalability challenges in both storage and computation. We propose Task Vector Bases, a framework compressing $T$ task vectors into $M < T$ basis vectors while preserving the functionality of task arithmetic. By representing each task vector as a structured linear combination of basis atoms, our approach supports standard operations such as addition, negation, as well as more advanced arithmetic ones. The framework is orthogonal to other efficiency-oriented improvements in task arithmetic and can be used in combination with them. We provide theoretical analysis showing that basis compression retains addition generalization guarantees and enables principled unlearning, with error bounds depending on reconstruction quality. Empirically, our proposed basis construction methods consistently outperform heuristic basis construction baselines and, in some cases, even surpass the performance of full task vector collections across diverse downstream applications while reducing storage and computational requirements. The code is available at https://github.com/uiuctml/TaskVectorBasis.", "authors": ["Siqi Zeng", "Yifei He", "Meitong Liu", "Weiqiu You", "Yifan Hao", "Yao-Hung Hubert Tsai", "Makoto Yamada", "Han Zhao"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-03", "url": "https://arxiv.org/abs/2502.01015", "pdf_url": "https://arxiv.org/pdf/2502.01015v4", "arxiv_id": "2502.01015", "doi": null, "citation_count": 2, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/uiuctml/TaskVectorBasis", "venue": null, "quality_score": 0.1193}
{"id": "6d419be8f92c1484fa5978d5f7e56ea97e468ef2fd6b3bb8f950476959ca22bc", "sources": ["arxiv", "semantic_scholar"], "title": "Task Arithmetic in Trust Region: A Training-Free Model Merging Approach to Navigate Knowledge Conflicts", "abstract": "Multi-task model merging offers an efficient solution for integrating knowledge from multiple fine-tuned models, mitigating the significant computational and storage demands associated with multi-task training. As a key technique in this field, Task Arithmetic (TA) defines task vectors by subtracting the pre-trained model ($θ_{\\text{pre}}$) from the fine-tuned task models in parameter space, then adjusting the weight between these task vectors and $θ_{\\text{pre}}$ to balance task-generalized and task-specific knowledge. Despite the promising performance of TA, conflicts can arise among the task vectors, particularly when different tasks require distinct model adaptations. In this paper, we formally define this issue as knowledge conflicts, characterized by the performance degradation of one task after merging with a model fine-tuned for another task. Through in-depth analysis, we show that these conflicts stem primarily from the components of task vectors that align with the gradient of task-specific losses at $θ_{\\text{pre}}$. To address this, we propose Task Arithmetic in Trust Region (TATR), which defines the trust region as dimensions in the model parameter space that cause only small changes (corresponding to the task vector components with gradient orthogonal direction) in the task-specific losses. Restricting parameter merging within this trust region, TATR can effectively alleviate knowledge conflicts. Moreover, TATR serves as both an independent approach and a plug-and-play module compatible with a wide range of TA-based methods. Extensive empirical evaluations on eight distinct datasets robustly demonstrate that TATR improves the multi-task performance of several TA-based model merging methods by an observable margin.", "authors": ["Wenju Sun", "Qingyong Li", "Wen Wang", "Yangli-ao Geng", "Boyang Li"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-01-25", "url": "https://arxiv.org/abs/2501.15065", "pdf_url": "https://arxiv.org/pdf/2501.15065v1", "arxiv_id": "2501.15065", "doi": "10.1145/3746027.3755789", "citation_count": 12, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "ACM Multimedia", "quality_score": 0.2785}
{"id": "6678d44fbfdc115a789d4f7953df06c35e29c42ed82ce47ad4ce18c6a4f9bd20", "sources": ["arxiv", "semantic_scholar"], "title": "Funzac at CoMeDi Shared Task: Modeling Annotator Disagreement from Word-In-Context Perspectives", "abstract": "In this work, we evaluate annotator disagreement in Word-in-Context (WiC) tasks exploring the relationship between contextual meaning and disagreement as part of the CoMeDi shared task competition. While prior studies have modeled disagreement by analyzing annotator attributes with single-sentence inputs, this shared task incorporates WiC to bridge the gap between sentence-level semantic representation and annotator judgment variability. We describe three different methods that we developed for the shared task, including a feature enrichment approach that combines concatenation, element-wise differences, products, and cosine similarity, Euclidean and Manhattan distances to extend contextual embedding representations, a transformation by Adapter blocks to obtain task-specific representations of contextual embeddings, and classifiers of varying complexities, including ensembles. The comparison of our methods demonstrates improved performance for methods that include enriched and task-specfic features. While the performance of our method falls short in comparison to the best system in subtask 1 (OGWiC), it is competitive to the official evaluation results in subtask 2 (DisWiC).", "authors": ["Olufunke O. Sarumi", "Charles Welch", "Lucie Flek", "Jörg Schlötterer"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-01-24", "url": "https://arxiv.org/abs/2501.14617", "pdf_url": "https://arxiv.org/pdf/2501.14617v1", "arxiv_id": "2501.14617", "doi": "10.48550/arXiv.2501.14617", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1193}
{"id": "5a73af290a882ab42203443df323cd62ecb21e382e43c6cffb7b9537d12674b2", "sources": ["arxiv", "semantic_scholar"], "title": "PPO-Based Vehicle Control for Ramp Merging Scheme Assisted by Enhanced C-V2X", "abstract": "On-ramp merging presents a critical challenge in autonomous driving, as vehicles from merging lanes need to dynamically adjust their positions and speeds while monitoring traffic on the main road to prevent collisions. To address this challenge, we propose a novel merging control scheme based on reinforcement learning, which integrates lateral control mechanisms. This approach ensures the smooth integration of vehicles from the merging lane onto the main road, optimizing both fuel efficiency and passenger comfort. Furthermore, we recognize the impact of vehicle-to-vehicle (V2V) communication on control strategies and introduce an enhanced protocol leveraging Cellular Vehicle-to-Everything (C-V2X) Mode 4. This protocol aims to reduce the Age of Information (AoI) and improve communication reliability. In our simulations, we employ two AoI-based metrics to rigorously assess the protocol's effectiveness in autonomous driving scenarios. By combining the NS3 network simulator with Python, we simulate V2V communication and vehicle control simultaneously. The results demonstrate that the enhanced C-V2X Mode 4 outperforms the standard version, while the proposed control scheme ensures safe and reliable vehicle operation during on-ramp merging.", "authors": ["Qiong Wu", "Maoxin Ji", "Pingyi Fan", "Kezhi Wang", "Nan Cheng", "Wen Chen", "Khaled B. Letaief"], "categories": ["cs.NI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-01-22", "url": "https://arxiv.org/abs/2501.12656", "pdf_url": "https://arxiv.org/pdf/2501.12656v2", "arxiv_id": "2501.12656", "doi": "10.48550/arXiv.2501.12656", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/qiongwu86/PPO-Based-Vehicle-Control-for-Ramp-Merging-Scheme-Assisted-by-Enhanced-C-V2X", "venue": "arXiv.org", "quality_score": 0.0753}
{"id": "ff88520ff88a7d437fbe412c1ec6e482ef87f8ef7e860ed5789ea0d6ad491ec3", "sources": ["arxiv", "semantic_scholar"], "title": "Merging Models on the Fly Without Retraining: A Sequential Approach to Scalable Continual Model Merging", "abstract": "Deep model merging represents an emerging research direction that combines multiple fine-tuned models to harness their specialized capabilities across different tasks and domains. Current model merging techniques focus on merging all available models simultaneously, with weight interpolation-based methods being the predominant approaches. However, these conventional approaches are not well-suited for scenarios where models become available sequentially, and they often suffer from high memory requirements and potential interference between tasks. In this study, we propose a training-free projection-based continual merging method that processes models sequentially through orthogonal projections of weight matrices and adaptive scaling mechanisms. Our method operates by projecting new parameter updates onto subspaces orthogonal to existing merged parameter updates while using an adaptive scaling mechanism to maintain stable parameter distances, enabling efficient sequential integration of task-specific knowledge. Our approach maintains constant memory complexity to the number of models, minimizes interference between tasks through orthogonal projections, and retains the performance of previously merged models through adaptive task vector scaling. Extensive experiments on CLIP-ViT models demonstrate that our method achieves a 5-8% average accuracy improvement while maintaining robust performance in different task orderings.", "authors": ["Anke Tang", "Enneng Yang", "Li Shen", "Yong Luo", "Han Hu", "Bo Du", "Dacheng Tao"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-01-16", "url": "https://arxiv.org/abs/2501.09522", "pdf_url": "https://arxiv.org/pdf/2501.09522v1", "arxiv_id": "2501.09522", "doi": "10.48550/arXiv.2501.09522", "citation_count": 29, "influential_citation_count": 9, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5}
{"id": "9c9ff1394d8c8d4fdee1cdf97df4e71809d67e00e0680b2655efe5b325f421b1", "sources": ["arxiv", "semantic_scholar"], "title": "Modeling Multi-Task Model Merging as Adaptive Projective Gradient Descent", "abstract": "Merging multiple expert models offers a promising approach for performing multi-task learning without accessing their original data. Existing methods attempt to alleviate task conflicts by sparsifying task vectors or promoting orthogonality among them. However, they overlook the fundamental target of model merging: the merged model performs as closely as possible to task-specific models on respective tasks. We find these methods inevitably discard task-specific information that, while causing conflicts, is crucial for performance. Based on our findings, we frame model merging as a constrained optimization problem ($\\textit{i.e.}$, minimizing the gap between the merged model and individual models, subject to the constraint of retaining shared knowledge) and solve it via adaptive projective gradient descent. Specifically, we align the merged model with individual models by decomposing and reconstituting the loss function, alleviating conflicts through $\\textit{data-free}$ optimization of task vectors. To retain shared knowledge, we optimize this objective by projecting gradients within a $\\textit{shared subspace}$ spanning all tasks. Moreover, we view merging coefficients as adaptive learning rates and propose a task-aware, training-free strategy. Experiments show that our plug-and-play approach consistently outperforms previous methods, achieving state-of-the-art results across diverse architectures and tasks in both vision and NLP domains.", "authors": ["Yongxian Wei", "Anke Tang", "Li Shen", "Zixuan Hu", "Chun Yuan", "Xiaochun Cao"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-01-02", "url": "https://arxiv.org/abs/2501.01230", "pdf_url": "https://arxiv.org/pdf/2501.01230v3", "arxiv_id": "2501.01230", "doi": "10.48550/arXiv.2501.01230", "citation_count": 40, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.4032}
{"id": "ac0617ae1f5f7c1321638a90a12b14348b73019dfff5692911d41e305da15c6b", "sources": ["arxiv", "semantic_scholar"], "title": "A Multi-task Supervised Compression Model for Split Computing", "abstract": "Split computing ($\\neq$ split learning) is a promising approach to deep learning models for resource-constrained edge computing systems, where weak sensor (mobile) devices are wirelessly connected to stronger edge servers through channels with limited communication capacity. State-of-theart work on split computing presents methods for single tasks such as image classification, object detection, or semantic segmentation. The application of existing methods to multitask problems degrades model accuracy and/or significantly increase runtime latency. In this study, we propose Ladon, the first multi-task-head supervised compression model for multi-task split computing. Experimental results show that the multi-task supervised compression model either outperformed or rivaled strong lightweight baseline models in terms of predictive performance for ILSVRC 2012, COCO 2017, and PASCAL VOC 2012 datasets while learning compressed representations at its early layers. Furthermore, our models reduced end-to-end latency (by up to 95.4%) and energy consumption of mobile devices (by up to 88.2%) in multi-task split computing scenarios.", "authors": ["Yoshitomo Matsubara", "Matteo Mendula", "Marco Levorato"], "categories": ["cs.CV", "cs.LG", "eess.IV"], "fields_of_study": ["Computer Science", "Engineering"], "published_date": "2025-01-02", "url": "https://arxiv.org/abs/2501.01420", "pdf_url": "https://arxiv.org/pdf/2501.01420v2", "arxiv_id": "2501.01420", "doi": "10.1109/WACV61041.2025.00481", "citation_count": 5, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/yoshitomo-matsubara/ladon-multi-task-sc2", "venue": "IEEE Workshop/Winter Conference on Applications of Computer Vision", "quality_score": 0.1945}
{"id": "526787cb3d5c119b75bc6933327b5064c79998e06dd38df5b6fe9e9285f90a8b", "sources": ["arxiv", "semantic_scholar"], "title": "Training-free Heterogeneous Model Merging", "abstract": "Model merging has attracted significant attention as a powerful paradigm for model reuse, facilitating the integration of task-specific models into a singular, versatile framework endowed with multifarious capabilities. Previous studies, predominantly utilizing methods such as Weight Average (WA), have shown that model merging can effectively leverage pretrained models without the need for laborious retraining. However, the inherent heterogeneity among models poses a substantial constraint on its applicability, particularly when confronted with discrepancies in model architectures. To overcome this challenge, we propose an innovative model merging framework designed for heterogeneous models, encompassing both depth and width heterogeneity. To address depth heterogeneity, we introduce a layer alignment strategy that harmonizes model layers by segmenting deeper models, treating consecutive layers with similar representations as a cohesive segment, thus enabling the seamless merging of models with differing layer depths. For width heterogeneity, we propose a novel elastic neuron zipping algorithm that projects the weights from models of varying widths onto a common dimensional space, eliminating the need for identical widths. Extensive experiments validate the efficacy of these proposed methods, demonstrating that the merging of structurally heterogeneous models can achieve performance levels comparable to those of homogeneous merging, across both vision and NLP tasks. Our code is publicly available at https://github.com/zju-vipa/training_free_heterogeneous_model_merging.", "authors": ["Zhengqi Xu", "Han Zheng", "Jie Song", "Li Sun", "Mingli Song"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-12-29", "url": "https://arxiv.org/abs/2501.00061", "pdf_url": "https://arxiv.org/pdf/2501.00061v1", "arxiv_id": "2501.00061", "doi": "10.48550/arXiv.2501.00061", "citation_count": 4, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/zju-vipa/training_free_heterogeneous_model_merging", "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "9c672e37f946669490bf2a2cc3f20a03590eb961726f9710dd4eb60203a30d8f", "sources": ["arxiv", "semantic_scholar"], "title": "Parameter-Efficient Interventions for Enhanced Model Merging", "abstract": "Model merging combines knowledge from task-specific models into a unified multi-task model to avoid joint training on all task data. However, current methods face challenges due to representation bias, which can interfere with tasks performance. As a remedy, we propose IntervMerge, a novel approach to multi-task model merging that effectively mitigates representation bias across the model using taskspecific interventions. To further enhance its efficiency, we introduce mini-interventions, which modify only part of the representation, thereby reducing the additional parameters without compromising performance. Experimental results demonstrate that IntervMerge consistently outperforms the state-of-the-art approaches using fewer parameters.", "authors": ["Marcin Osial", "Daniel Marczak", "Bartosz Zieliński"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-12-22", "url": "https://arxiv.org/abs/2412.17023", "pdf_url": "https://arxiv.org/pdf/2412.17023v1", "arxiv_id": "2412.17023", "doi": "10.48550/arXiv.2412.17023", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1505}
{"id": "51beee1f6d0847d14264d741843dcc2935b1a7e3406dae6b74e5394f7745b3a2", "sources": ["arxiv", "semantic_scholar"], "title": "Bias Vector: Mitigating Biases in Language Models with Task Arithmetic Approach", "abstract": "The use of language models (LMs) has increased considerably in recent years, and the biases and stereotypes in training data that are reflected in the LM outputs are causing social problems. In this paper, inspired by the task arithmetic, we propose the ``Bias Vector'' method for the mitigation of these LM biases. The Bias Vector method does not require manually created debiasing data. The three main steps of our approach involve: (1) continual training the pre-trained LMs on biased data using masked language modeling; (2) constructing the Bias Vector as the difference between the weights of the biased LMs and those of pre-trained LMs; and (3) subtracting the Bias Vector from the weights of the pre-trained LMs for debiasing. We evaluated the Bias Vector method on the SEAT across three LMs and confirmed an average improvement of 0.177 points. We demonstrated that the Bias Vector method does not degrade the LM performance on downstream tasks in the GLUE benchmark. In addition, we examined the impact of scaling factors, which control the magnitudes of Bias Vectors, with effect sizes on the SEAT and conducted a comprehensive evaluation of our debiased LMs across both the SEAT and GLUE benchmarks.", "authors": ["Daiki Shirafuji", "Makoto Takenaka", "Shinya Taguchi"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-12-16", "url": "https://arxiv.org/abs/2412.11679", "pdf_url": "https://arxiv.org/pdf/2412.11679v1", "arxiv_id": "2412.11679", "doi": "10.48550/arXiv.2412.11679", "citation_count": 14, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "International Conference on Computational Linguistics", "quality_score": 0.294}
{"id": "127703b701021b6160080fcfa2c44af4600050faf41a5cc128229d063a2088af", "sources": ["arxiv", "semantic_scholar"], "title": "Targeted Angular Reversal of Weights (TARS) for Knowledge Removal in Large Language Models", "abstract": "The sheer scale of data required to train modern large language models (LLMs) poses significant risks, as models are likely to gain knowledge of sensitive topics such as bio-security, as well the ability to replicate copyrighted works. Methods designed to remove such knowledge must do so from all prompt directions, in a multi-lingual capacity and without degrading general model performance. To this end, we introduce the targeted angular reversal (TARS) method of knowledge removal from LLMs. The TARS method firstly leverages the LLM in combination with a detailed prompt to aggregate information about a selected concept in the internal representation space of the LLM. It then refines this approximate concept vector to trigger the concept token with high probability, by perturbing the approximate concept vector with noise and transforming it into token scores with the language model head. The feedforward weight vectors in the LLM which operate directly on the internal representation space, and have the highest cosine similarity with this targeting vector, are then replaced by a reversed targeting vector, thus limiting the ability of the concept to propagate through the model. The modularity of the TARS method allows for a sequential removal of concepts from Llama 3.1 8B, such as the famous literary detective Sherlock Holmes, and the planet Saturn. It is demonstrated that the probability of triggering target concepts can be reduced to 0.00 with as few as 1 TARS edit, whilst simultaneously removing the knowledge bi-directionally. Moreover, knowledge is shown to be removed across all languages despite only being targeted in English. Importantly, TARS has minimal impact on the general model capabilities, as after removing 5 diverse concepts in a modular fashion, there is minimal KL divergence in the next token probabilities of the LLM on large corpora of Wikipedia text (median of 0.0015).", "authors": ["Harry J. Davies", "Giorgos Iacovides", "Danilo P. Mandic"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-12-13", "url": "https://arxiv.org/abs/2412.10257", "pdf_url": "https://arxiv.org/pdf/2412.10257v2", "arxiv_id": "2412.10257", "doi": "10.48550/arXiv.2412.10257", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0}
{"id": "938ea924f7ec5fa34843c904cd595453cc8e20f7b60dc281b55904fd0d25e559", "sources": ["arxiv", "semantic_scholar"], "title": "Revisiting Weight Averaging for Model Merging", "abstract": "Model merging aims to build a multi-task learner by combining the parameters of individually fine-tuned models without additional training. While a straightforward approach is to average model parameters across tasks, this often results in suboptimal performance due to interference among parameters across tasks. In this paper, we present intriguing results that weight averaging implicitly induces task vectors centered around the weight averaging itself and that applying a low-rank approximation to these centered task vectors significantly improves merging performance. Our analysis shows that centering the task vectors effectively reduces task interference and most of task-specific knowledge is concentrated in the top singular vectors. Our method demonstrates robust and scalable performance on vision benchmarks across varying numbers of tasks and model sizes. Furthermore, we observe that our approach is applicable to natural language processing tasks with competitive performance.", "authors": ["Jiho Choi", "Donggyun Kim", "Chanhyuk Lee", "Seunghoon Hong"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-12-11", "url": "https://arxiv.org/abs/2412.12153", "pdf_url": "https://arxiv.org/pdf/2412.12153v2", "arxiv_id": "2412.12153", "doi": "10.48550/arXiv.2412.12153", "citation_count": 23, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3451}
{"id": "25a08161a3ad5be1fe7c4875b0defc0772f2d00e8fd4cd67ac4b42ab1989b2e4", "sources": ["arxiv", "semantic_scholar"], "title": "How to Weight Multitask Finetuning? Fast Previews via Bayesian Model-Merging", "abstract": "When finetuning multiple tasks altogether, it is important to carefully weigh them to get a good performance, but searching for good weights can be difficult and costly. Here, we propose to aid the search with fast previews to quickly get a rough idea of different reweighting options. We use model merging to create previews by simply reusing and averaging parameters of models trained on each task separately (no retraining required). To improve the quality of previews, we propose a Bayesian approach to design new merging strategies by using more flexible posteriors. We validate our findings on vision and natural-language transformers. Our work shows the benefits of model merging via Bayes to improve multitask finetuning.", "authors": ["Hugo Monzón Maldonado", "Thomas Möllenhoff", "Nico Daheim", "Iryna Gurevych", "Mohammad Emtiyaz Khan"], "categories": ["cs.LG", "cs.AI", "stat.ML"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2024-12-11", "url": "https://arxiv.org/abs/2412.08147", "pdf_url": "https://arxiv.org/pdf/2412.08147v1", "arxiv_id": "2412.08147", "doi": "10.48550/arXiv.2412.08147", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0753}
{"id": "4448c81ceeb75deb04216e08cafb39b1684a3f6e73c23ca50c45d8487f338c5d", "sources": ["arxiv", "semantic_scholar"], "title": "How to Merge Your Multimodal Models Over Time?", "abstract": "Model merging combines multiple expert models - finetuned from a base foundation model on diverse tasks and domains - into a single, more capable model. However, most existing model merging approaches assume that all experts are available simultaneously. In reality, new tasks and domains emerge progressively over time, requiring strategies to integrate the knowledge of expert models as they become available: a process we call temporal model merging. The temporal dimension introduces unique challenges not addressed in prior work, raising new questions such as: when training for a new task, should the expert model start from the merged past experts or from the original base model? Should we merge all models at each time step? Which merging techniques are best suited for temporal merging? Should different strategies be used to initialize the training and deploy the model? To answer these questions, we propose a unified framework called TIME - Temporal Integration of Model Expertise - which defines temporal model merging across three axes: (1) Initialization Phase, (2) Deployment Phase, and (3) Merging Technique. Using TIME, we study temporal model merging across model sizes, compute budgets, and learning horizons on the FoMo-in-Flux benchmark. Our comprehensive suite of experiments across TIME allows us to uncover key insights for temporal model merging, offering a better understanding of current challenges and best practices for effective temporal model merging.", "authors": ["Sebastian Dziadzio", "Vishaal Udandarao", "Karsten Roth", "Ameya Prabhu", "Zeynep Akata", "Samuel Albanie", "Matthias Bethge"], "categories": ["cs.LG", "cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-12-09", "url": "https://arxiv.org/abs/2412.06712", "pdf_url": "https://arxiv.org/pdf/2412.06712v1", "arxiv_id": "2412.06712", "doi": "10.1109/CVPR52734.2025.01907", "citation_count": 22, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/ExplainableML/fomo_in_flux", "venue": "Computer Vision and Pattern Recognition", "quality_score": 0.3404}
{"id": "a843b1091100d65dc1539fd1f20d3ad08f3f9114d03c49da44637256d4938711", "sources": ["arxiv", "semantic_scholar"], "title": "SuperMerge: An Approach For Gradient-Based Model Merging", "abstract": "Large language models, such as ChatGPT, Claude, or LLaMA, are gigantic, monolithic, and possess the superpower to simultaneously support thousands of tasks. However, high-throughput applications often prefer smaller task-specific models because of their lower latency and cost. One challenge of using task-specific models is the incremental need for solving newer tasks after the model is already deployed for existing tasks. A straightforward solution requires fine-tuning the model again for both existing and new tasks, which is computationally expensive and time-consuming. To address this issue, we propose a model merging based approach called SUPERMERGE. SUPERMERGE is a gradient-based method to systematically merge several fine-tuned models trained on existing and new tasks. SUPERMERGE is designed to be lightweight and fast, and the merged model achieves similar performance to fully fine-tuned models on all tasks. Furthermore, we proposed a hierarchical model merging strategy to reduce the peak space requirement without sacrificing the performance of the merged model. We experimentally demonstrate that SUPERMERGE outperforms existing model merging methods on common natural language processing and computer vision tasks.", "authors": ["Haoyu Yang", "Zheng Zhang", "Saket Sathe"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-12-09", "url": "https://arxiv.org/abs/2412.10416", "pdf_url": "https://arxiv.org/pdf/2412.10416v2", "arxiv_id": "2412.10416", "doi": "10.48550/arXiv.2412.10416", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0}
{"id": "7bc1309940427579042c8cddefcfd3bbb28847bc174e3036029b970395c6016b", "sources": ["arxiv", "semantic_scholar"], "title": "Large Language Models Merging for Enhancing the Link Stealing Attack on Graph Neural Networks", "abstract": "Graph Neural Networks (GNNs), specifically designed to process the graph data, have achieved remarkable success in various applications. Link stealing attacks on graph data pose a significant privacy threat, as attackers aim to extract sensitive relationships between nodes (entities), potentially leading to academic misconduct, fraudulent transactions, or other malicious activities. Previous studies have primarily focused on single datasets and did not explore cross-dataset attacks, let alone attacks that leverage the combined knowledge of multiple attackers. However, we find that an attacker can combine the data knowledge of multiple attackers to create a more effective attack model, which can be referred to cross-dataset attacks. Moreover, if knowledge can be extracted with the help of Large Language Models (LLMs), the attack capability will be more significant. In this paper, we propose a novel link stealing attack method that takes advantage of cross-dataset and Large Language Models (LLMs). The LLM is applied to process datasets with different data structures in cross-dataset attacks. Each attacker fine-tunes the LLM on their specific dataset to generate a tailored attack model. We then introduce a novel model merging method to integrate the parameters of these attacker-specific models effectively. The result is a merged attack model with superior generalization capabilities, enabling effective attacks not only on the attackers' datasets but also on previously unseen (out-of-domain) datasets. We conducted extensive experiments in four datasets to demonstrate the effectiveness of our method. Additional experiments with three different GNN and LLM architectures further illustrate the generality of our approach.", "authors": ["Faqian Guan", "Tianqing Zhu", "Wenhan Chang", "Wei Ren", "Wanlei Zhou"], "categories": ["cs.CR", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-12-08", "url": "https://arxiv.org/abs/2412.05830", "pdf_url": "https://arxiv.org/pdf/2412.05830v1", "arxiv_id": "2412.05830", "doi": "10.48550/arXiv.2412.05830", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0}
{"id": "02f06d917ca57bb7deb6cd452e134751c851f332bb3d420c40acd4fdb2ebfb89", "sources": ["arxiv", "semantic_scholar"], "title": "Multi-Task Model Merging via Adaptive Weight Disentanglement", "abstract": "Model merging has recently gained attention as an economical and scalable approach to incorporate task-specific weights from various tasks into a unified multi-task model. For example, in Task Arithmetic (TA), adding the fine-tuned weights of different tasks can enhance the model's performance on those tasks, while subtracting them leads to task forgetting. Although TA is highly effective, interference among task still hampers the performance of the merged model. Existing methods for handling conflicts between task generally rely on empirical selection, resulting in suboptimal performance. In this paper, we introduce an Adaptive Weight Disentanglement method. We begin by theoretically proving that task vectors employed in model merging should be orthogonal to minimize interference among tasks. Guided by this insight, we initialize redundant vectors such that, when subtracted from the original task vectors, the resulting vectors exhibit increased orthogonality. Additionally, we impose an norm constraint on the redundant vectors to preserve the performance of the task-specific models. Experimental results demonstrate the effectiveness of our proposed technique: it successfully extracts redundant vectors, and after their subtraction, the task vectors not only retain robust performance but also achieve superior fusion outcomes. Our code is available at \\href{https://github.com/FarisXiong/AWD.git}{https://github.com/FarisXiong/AWD.git}.", "authors": ["Feng Xiong", "Runxi Cheng", "Wang Chen", "Zhanqiu Zhang", "Yiwen Guo", "Chun Yuan", "Ruifeng Xu"], "categories": ["cs.LG", "cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-11-27", "url": "https://arxiv.org/abs/2411.18729", "pdf_url": "https://arxiv.org/pdf/2411.18729v2", "arxiv_id": "2411.18729", "doi": "10.48550/arXiv.2411.18729", "citation_count": 18, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/FarisXiong/AWD.git}{https://github.com/FarisXiong/AWD.git}", "venue": "arXiv.org", "quality_score": 0.3197}
{"id": "23f75b210dbcceaa0ce81652d866ca7692fad32e88c84335024760dd542bc11d", "sources": ["arxiv", "semantic_scholar"], "title": "Task Arithmetic Through The Lens Of One-Shot Federated Learning", "abstract": "Task Arithmetic is a model merging technique that enables the combination of multiple models' capabilities into a single model through simple arithmetic in the weight space, without the need for additional fine-tuning or access to the original training data. However, the factors that determine the success of Task Arithmetic remain unclear. In this paper, we examine Task Arithmetic for multi-task learning by framing it as a one-shot Federated Learning problem. We demonstrate that Task Arithmetic is mathematically equivalent to the commonly used algorithm in Federated Learning, called Federated Averaging (FedAvg). By leveraging well-established theoretical results from FedAvg, we identify two key factors that impact the performance of Task Arithmetic: data heterogeneity and training heterogeneity. To mitigate these challenges, we adapt several algorithms from Federated Learning to improve the effectiveness of Task Arithmetic. Our experiments demonstrate that applying these algorithms can often significantly boost performance of the merged model compared to the original Task Arithmetic approach. This work bridges Task Arithmetic and Federated Learning, offering new theoretical perspectives on Task Arithmetic and improved practical methodologies for model merging.", "authors": ["Zhixu Silvia Tao", "Ian Mason", "Sanjeev Kulkarni", "Xavier Boix"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-11-27", "url": "https://arxiv.org/abs/2411.18607", "pdf_url": "https://arxiv.org/pdf/2411.18607v2", "arxiv_id": "2411.18607", "doi": "10.48550/arXiv.2411.18607", "citation_count": 12, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.2785}
{"id": "03fea48836148067318f1a2d457e77774e74326359a578b77b348839042efcae", "sources": ["arxiv", "semantic_scholar"], "title": "Task Singular Vectors: Reducing Task Interference in Model Merging", "abstract": "Task Arithmetic has emerged as a simple yet effective method to merge models without additional training. However, by treating entire networks as flat parameter vectors, it overlooks key structural information and is susceptible to task interference. In this paper, we study task vectors at the layer level, focusing on task layer matrices and their singular value decomposition. In particular, we concentrate on the resulting singular vectors, which we refer to as Task Singular Vectors (TSV). Recognizing that layer task matrices are often low-rank, we propose TSV-Compress (TSV-C), a simple procedure that compresses them to 10% of their original size while retaining 99% of accuracy. We further leverage this low-rank space to define a new measure of task interference based on the interaction of singular vectors from different tasks. Building on these findings, we introduce TSV-Merge (TSV-M), a novel model merging approach that combines compression with interference reduction, significantly outperforming existing methods.", "authors": ["Antonio Andrea Gargiulo", "Donato Crisostomi", "Maria Sofia Bucarelli", "Simone Scardapane", "Fabrizio Silvestri", "Emanuele Rodolà"], "categories": ["cs.LG", "stat.ML"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2024-11-26", "url": "https://arxiv.org/abs/2412.00081", "pdf_url": "https://arxiv.org/pdf/2412.00081v3", "arxiv_id": "2412.00081", "doi": "10.1109/CVPR52734.2025.01742", "citation_count": 113, "influential_citation_count": 35, "has_code": false, "code_url": null, "venue": "Computer Vision and Pattern Recognition", "quality_score": 0.7782}
{"id": "9f2782216df8bc726baea14a2648e478daa2647c7d1611ca8b14de10fd24679f", "sources": ["arxiv", "semantic_scholar"], "title": "FREE-Merging: Fourier Transform for Efficient Model Merging", "abstract": "With the rapid growth of deep learning, there is an increasing availability of open-source models for various tasks. However, single fine-tuned models often fall short of meeting the diverse needs of users. Model merging has thus emerged as an efficient method to integrate the capabilities of existing models into a unified model. Nevertheless, existing model merging methods face challenging trade-offs between performance and deployment costs, primarily due to task interference. For the first time, we reveal that task interference is evident in the frequency domain of model parameters, yet current efforts only focus on spatial domain solutions, which are largely ineffective in addressing frequency domain interference. To mitigate the impact of frequency domain interference, we propose FR-Merging, an innovative method that effectively filters harmful frequency domain interference on the backbone with minimal computational overhead. Since performance loss is inevitable with cost-free methods, we propose a lightweight task-specific expert module that dynamically compensates for information loss during merging. This proposed framework, FREE-Merging (FR-Merging with experts), strikes a balanced trade-off between training cost, inference latency, storage requirements, and performance. We demonstrate the effectiveness of both FR-Merging and FREE-Merging on multiple tasks across CV, NLP, and Multi-Modal domains and show that they can be flexibly adapted to specific needs.", "authors": ["Shenghe Zheng", "Hongzhi Wang"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-11-25", "url": "https://arxiv.org/abs/2411.16815", "pdf_url": "https://arxiv.org/pdf/2411.16815v3", "arxiv_id": "2411.16815", "doi": "10.1109/ICCV51701.2025.00368", "citation_count": 11, "influential_citation_count": 2, "has_code": true, "code_url": null, "venue": "IEEE International Conference on Computer Vision", "quality_score": 0.2698}
{"id": "d2bc33ebb11d976a49b1d5d2d8ddce26ad3e4aeebc0249378a2fdc26c1018fe1", "sources": ["arxiv"], "title": "Less is More: Efficient Model Merging with Binary Task Switch", "abstract": "As an effective approach to equip models with multi-task capabilities without additional training, model merging has garnered significant attention. However, existing methods face challenges of redundant parameter conflicts and the excessive storage burden of parameters. In this work, through controlled experiments, we reveal that for task vectors, only those parameters with magnitudes above a certain threshold contribute positively to the task, exhibiting a pulse-like characteristic. We then attempt leveraging this characteristic to binarize the task vectors and reduce storage overhead. Further controlled experiments show that the binarized task vectors incur almost no decrease in fine-tuning and merging performance, and even exhibit stronger performance improvements as the proportion of redundant parameters increases. Based on these insights, we propose Task Switch (T-Switch), which decomposes task vectors into three components: 1) an activation switch instantiated by a binarized mask vector, 2) a polarity switch instantiated by a binarized sign vector, and 3) a scaling knob instantiated by a scalar coefficient. By storing task vectors in a binarized form, T-Switch alleviates parameter conflicts while ensuring efficient task parameter storage. Furthermore, to enable automated switch combination in T-Switch, we further introduce Auto-Switch, which enables training-free switch combination via retrieval from a small query set. Experiments indicate that our methods achieve significant performance improvements over existing baselines, requiring only 1-3% of the storage space of full-precision parameters.", "authors": ["Biqing Qi", "Fangyuan Li", "Zhen Wang", "Junqi Gao", "Dong Li", "Peng Ye", "Bowen Zhou"], "categories": ["cs.LG"], "fields_of_study": [], "published_date": "2024-11-24", "url": "https://arxiv.org/abs/2412.00054", "pdf_url": "https://arxiv.org/pdf/2412.00054v1", "arxiv_id": "2412.00054", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.0}
{"id": "603b26065fe441b8934357cbcd7deddc4eb0088146ab86fc6744394fc3d599bd", "sources": ["arxiv", "semantic_scholar"], "title": "Multi LoRA Meets Vision: Merging multiple adapters to create a multi task model", "abstract": "Parameter efficient finetuning (PEFT) methods are widely used in LLMs and generative models in computer vision. Especially one can use multiple of these during inference to change the behavior of the base model. In this paper we investigated whether multiple LoRA adapters trained on computer vision tasks can be merged together and used during inference without loss in performance. By achieving this, multitask models can be created just by merging different LoRAs. Merging these will reduce inference time and it will not require any additional retraining. We have trained adapters on six different tasks and evaluated their performance when they are merged together. For comparison we used a model with a frozen backbone and finetuned its head. Our results show that even with simple merging techniques creating a multitask model by merging adapters is achievable by slightly loosing performance in some cases. In our experiments we merged up to three adapters together. Depending on the task and the similarity of the data adapters were trained on, merges can outperform head finetuning. We have observed that LoRAs trained with dissimilar datasets tend to perform better compared to model trained on similar datasets.", "authors": ["Ege Kesim", "Selahattin Serdar Helli"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-11-21", "url": "https://arxiv.org/abs/2411.14064", "pdf_url": "https://arxiv.org/pdf/2411.14064v1", "arxiv_id": "2411.14064", "doi": "10.48550/arXiv.2411.14064", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "c49cdfa9e0380a75ce563eabd507b348846f1dc9b0250788b024e648176e655b", "sources": ["arxiv", "semantic_scholar"], "title": "Rethinking Weight-Averaged Model-merging", "abstract": "Model merging, particularly through weight averaging, has shown surprising effectiveness in saving computations and improving model performance without any additional training. However, the interpretability of why and how this technique works remains unclear. In this work, we reinterpret weight-averaged model merging through the lens of interpretability and provide empirical insights into the underlying mechanisms that govern its behavior. We approach the problem from three perspectives: (1) we analyze the learned weight structures and demonstrate that model weights encode structured representations that help explain the compatibility of weight averaging; (2) we compare averaging in weight space and feature space across diverse model architectures (CNNs and ViTs) and datasets, aiming to expose under which circumstances what combination paradigm will work more effectively; (3) we study the effect of parameter scaling on prediction stability, highlighting how weight averaging acts as a form of regularization that contributes to robustness. By framing these analyses in an interpretability context, our work contributes to a more transparent and systematic understanding of model merging for stakeholders interested in the safety and reliability of untrained model combination methods. The code is available at https://github.com/billhhh/Rethink-Merge.", "authors": ["Hu Wang", "Congbo Ma", "Ibrahim Almakky", "Ian Reid", "Gustavo Carneiro", "Mohammad Yaqub"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-11-14", "url": "https://arxiv.org/abs/2411.09263", "pdf_url": "https://arxiv.org/pdf/2411.09263v5", "arxiv_id": "2411.09263", "doi": "10.48550/arXiv.2411.09263", "citation_count": 6, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/billhhh/Rethink-Merge", "venue": "arXiv.org", "quality_score": 0.2113}
{"id": "c1580c35b55f57b9be6cc9373fbf39f08a47355b10d99c68987a351c8ed4b4ea", "sources": ["arxiv", "semantic_scholar"], "title": "Large Wireless Model (LWM): A Foundation Model for Wireless Channels", "abstract": "This paper presents Large Wireless Model (LWM) -- the world's first foundation model for wireless channels. Designed as a task-agnostic model, LWM generates universal, rich, contextualized channel embeddings (features) that potentially enhance performance across a wide range of downstream tasks in wireless communication and sensing systems. Towards this objective, LWM, which has a transformer-based architecture, was pre-trained in a self-supervised manner on large-scale wireless channel datasets. Our results show consistent improvements in downstream tasks when using the LWM embeddings compared to raw channel representations, especially in scenarios with high-complexity machine learning tasks and limited training datasets. This LWM's ability to learn from large-scale wireless data opens a promising direction for intelligent systems that can efficiently adapt to diverse tasks with limited data, paving the way for addressing key challenges in wireless communication and sensing systems.", "authors": ["Sadjad Alikhani", "Gouranga Charan", "Ahmed Alkhateeb"], "categories": ["cs.IT", "eess.SP"], "fields_of_study": ["Computer Science", "Engineering", "Mathematics"], "published_date": "2024-11-13", "url": "https://arxiv.org/abs/2411.08872", "pdf_url": "https://arxiv.org/pdf/2411.08872v2", "arxiv_id": "2411.08872", "doi": "10.48550/arXiv.2411.08872", "citation_count": 79, "influential_citation_count": 14, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.588}
{"id": "4845f5e7134503404efb4d68e77af61709e28b3ade81a6914fc9689dfd0c4255", "sources": ["arxiv", "semantic_scholar"], "title": "ATM: Improving Model Merging by Alternating Tuning and Merging", "abstract": "Model merging has emerged as a cost-efficient approximation to multitask learning. Among merging strategies, task arithmetic is notable for its simplicity and effectiveness. In this work, we provide a theoretical motivation for task vectors by highlighting that, under single-epoch full-batch gradient descent, they are equivalent to multitask gradients. This insight leads us to reinterpret model merging as a single step in an iterative procedure that Alternates between Tuning and Merging (ATM). We propose two applications of ATM: (1) as an alternative to multitask learning in scenarios where data sharing is restricted (e.g., federated settings), and (2) as a lightweight refinement step to improve existing model merging methods using a small validation set. Experiments across diverse vision tasks demonstrate the effectiveness of ATM.", "authors": ["Luca Zhou", "Daniele Solombrino", "Donato Crisostomi", "Maria Sofia Bucarelli", "Fabrizio Silvestri", "Emanuele Rodolà"], "categories": ["cs.LG", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-11-05", "url": "https://arxiv.org/abs/2411.03055", "pdf_url": "https://arxiv.org/pdf/2411.03055v4", "arxiv_id": "2411.03055", "doi": "10.48550/arXiv.2411.03055", "citation_count": 8, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Conference on Image Analysis and Processing", "quality_score": 0.2386}
{"id": "6594c4d68ef33e6c61f15d58490eb8c7d90e1eeab1f04ef1ac8a653a6685f566", "sources": ["arxiv", "semantic_scholar"], "title": "Local Superior Soups: A Catalyst for Model Merging in Cross-Silo Federated Learning", "abstract": "Federated learning (FL) is a learning paradigm that enables collaborative training of models using decentralized data. Recently, the utilization of pre-trained weight initialization in FL has been demonstrated to effectively improve model performance. However, the evolving complexity of current pre-trained models, characterized by a substantial increase in parameters, markedly intensifies the challenges associated with communication rounds required for their adaptation to FL. To address these communication cost issues and increase the performance of pre-trained model adaptation in FL, we propose an innovative model interpolation-based local training technique called ``Local Superior Soups.'' Our method enhances local training across different clients, encouraging the exploration of a connected low-loss basin within a few communication rounds through regularized model interpolation. This approach acts as a catalyst for the seamless adaptation of pre-trained models in in FL. We demonstrated its effectiveness and efficiency across diverse widely-used FL datasets. Our code is available at \\href{https://github.com/ubc-tea/Local-Superior-Soups}{https://github.com/ubc-tea/Local-Superior-Soups}.", "authors": ["Minghui Chen", "Meirui Jiang", "Xin Zhang", "Qi Dou", "Zehua Wang", "Xiaoxiao Li"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-31", "url": "https://arxiv.org/abs/2410.23660", "pdf_url": "https://arxiv.org/pdf/2410.23660v1", "arxiv_id": "2410.23660", "doi": "10.48550/arXiv.2410.23660", "citation_count": 8, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/ubc-tea/Local-Superior-Soups}{https://github.com/ubc-tea/Local-Superior-Soups}", "venue": "Neural Information Processing Systems", "quality_score": 0.2386}
{"id": "381a3f983bd798062d2c403f4b6daf229655f6b51250d99794ed8621d437fa75", "sources": ["arxiv", "semantic_scholar"], "title": "Efficient and Effective Weight-Ensembling Mixture of Experts for Multi-Task Model Merging", "abstract": "Multi-task learning (MTL) leverages a shared model to accomplish multiple tasks and facilitate knowledge transfer. Recent research on task arithmetic-based MTL demonstrates that merging the parameters of independently fine-tuned models can effectively achieve MTL. However, existing merging methods primarily seek a static optimal solution within the original model parameter space, which often results in performance degradation due to the inherent diversity among tasks and potential interferences. To address this challenge, in this paper, we propose a Weight-Ensembling Mixture of Experts (WEMoE) method for multi-task model merging. Specifically, we first identify critical (or sensitive) modules by analyzing parameter variations in core modules of Transformer-based models before and after finetuning. Then, our WEMoE statically merges non-critical modules while transforming critical modules into a mixture-of-experts (MoE) structure. During inference, expert modules in the MoE are dynamically merged based on input samples, enabling a more flexible and adaptive merging approach. Building on WEMoE, we further introduce an efficient-and-effective WEMoE (E-WEMoE) method, whose core mechanism involves eliminating non-essential elements in the critical modules of WEMoE and implementing shared routing across multiple MoE modules, thereby significantly reducing both the trainable parameters, the overall parameter count, and computational overhead of the merged model by WEMoE. Experimental results across various architectures and tasks demonstrate that both WEMoE and E-WEMoE outperform state-of-the-art (SOTA) model merging methods in terms of MTL performance, generalization, and robustness.", "authors": ["Li Shen", "Anke Tang", "Enneng Yang", "Guibing Guo", "Yong Luo", "Lefei Zhang", "Xiaochun Cao", "Bo Du", "Dacheng Tao"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science", "Medicine"], "published_date": "2024-10-29", "url": "https://arxiv.org/abs/2410.21804", "pdf_url": "https://arxiv.org/pdf/2410.21804v1", "arxiv_id": "2410.21804", "doi": "10.1109/TPAMI.2025.3629605", "citation_count": 23, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "IEEE Transactions on Pattern Analysis and Machine Intelligence", "quality_score": 0.3451}
{"id": "c98fc1c82ce997f1e1827c300f64048d8cb1a540042478db63f91740b49e891c", "sources": ["arxiv", "semantic_scholar"], "title": "Robust Estimation and Model Selection for the Controlled Directed Effect with Unmeasured Mediator-Outcome Confounders", "abstract": "Controlled Direct Effect (CDE) is one of the causal estimands used to evaluate both exposure and mediation effects on an outcome. When there are unmeasured confounders existing between the mediator and the outcome, the ordinary identification assumption does not work. In this manuscript, we consider an identification condition to identify CDE in the presence of unmeasured confounders. The key assumptions are: 1) the random allocation of the exposure, and 2) the existence of instrumental variables directly related to the mediator. Under these conditions, we propose a novel doubly robust estimation method, which work well if either the propensity score model or the baseline outcome model is correctly specified. Additionally, we propose a Generalized Information Criterion (GIC)-based model selection criterion for CDE that ensures model selection consistency. Our proposed procedure and related methods are applied to both simulation and real datasets to confirm the performance of these methods. Our proposed method can select the correct model with high probability and accurately estimate CDE.", "authors": ["Shunichiro Orihara", "Shinpei Imori", "Kosuke Morikawa", "Atsushi Goto", "Masataka Taguri"], "categories": ["stat.ME"], "fields_of_study": ["Mathematics"], "published_date": "2024-10-29", "url": "https://arxiv.org/abs/2410.21832", "pdf_url": "https://arxiv.org/pdf/2410.21832v1", "arxiv_id": "2410.21832", "doi": "10.1007/s10463-026-00985-w", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Annals of the Institute of Statistical Mathematics", "quality_score": 0.0}
{"id": "90e026b4117230cf0c763d9c2462225005cc672b1a396a601ab64fdddd86f128", "sources": ["arxiv", "semantic_scholar"], "title": "Model merging with SVD to tie the Knots", "abstract": "Recent model merging methods demonstrate that the parameters of fully-finetuned models specializing in distinct tasks can be combined into one model capable of solving all tasks without retraining. Yet, this success does not transfer well when merging LoRA finetuned models. We study this phenomenon and observe that the weights of LoRA finetuned models showcase a lower degree of alignment compared to their fully-finetuned counterparts. We hypothesize that improving this alignment is key to obtaining better LoRA model merges, and propose KnOTS to address this problem. KnOTS uses the SVD to jointly transform the weights of different LoRA models into an aligned space, where existing merging methods can be applied. In addition, we introduce a new benchmark that explicitly evaluates whether merged models are general models. Notably, KnOTS consistently improves LoRA merging by up to 4.3% across several vision and language benchmarks, including our new setting. We release our code at: https://github.com/gstoica27/KnOTS.", "authors": ["George Stoica", "Pratik Ramesh", "Boglarka Ecsedi", "Leshem Choshen", "Judy Hoffman"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-25", "url": "https://arxiv.org/abs/2410.19735", "pdf_url": "https://arxiv.org/pdf/2410.19735v1", "arxiv_id": "2410.19735", "doi": "10.48550/arXiv.2410.19735", "citation_count": 93, "influential_citation_count": 24, "has_code": true, "code_url": "https://github.com/gstoica27/KnOTS", "venue": "International Conference on Learning Representations", "quality_score": 0.699}
{"id": "c4c7553446d715a16d03c7292c0aefd08e17152af636d7003788964bfeb24076", "sources": ["arxiv", "semantic_scholar"], "title": "Language Models are Symbolic Learners in Arithmetic", "abstract": "The prevailing question in LM performing arithmetic is whether these models learn to truly compute or if they simply master superficial pattern matching. In this paper, we argues for the latter, presenting evidence that LMs act as greedy symbolic learners, prioritizing the simplest possible shortcuts to fit the stats of dataset to solve arithmetic tasks. To investigate this, we introduce subgroup induction, a practical framework adapted from Solomonoff Induction (SI), one of the most powerful universal predictors. Our framework analyzes arithmetic problems by breaking them down into subgroups-minimal mappings between a few input digits and a single output digit. Our primary metric, subgroup quality, measures the viability of these shortcuts. Experiments reveal a distinct U-shaped accuracy pattern in multi-digit multiplication: LMs quickly master the first and last output digits while struggling with those in the middle. We demonstrate this U-shape is not coincidental; it perfectly mirrors the quality of the simplest possible subgroups, those requiring the fewest input tokens. This alignment suggests a core learning mechanism: LMs first learn easy, low-token shortcuts and only incorporate more complex, multi-token patterns as training progresses. They do not learn the algorithm of multiplication but rather a hierarchy of increasingly complex symbol-to-symbol mappings. Ultimately, our findings suggest that the path to arithmetic mastery for LMs is not paved with algorithms, but with a cascade of simple, hierarchically-learned symbolic shortcuts.", "authors": ["Chunyuan Deng", "Zhiqi Li", "Roy Xie", "Ruidi Chang", "Hanjie Chen"], "categories": ["cs.LG", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-21", "url": "https://arxiv.org/abs/2410.15580", "pdf_url": "https://arxiv.org/pdf/2410.15580v2", "arxiv_id": "2410.15580", "doi": "10.48550/arXiv.2410.15580", "citation_count": 9, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/chili-lab/Symbolic-Arithmetic", "venue": null, "quality_score": 0.25}
{"id": "239e46784f5fe03662669501b8e7b379bd63ebe42470eb0d843e525692743765", "sources": ["arxiv", "semantic_scholar"], "title": "Analysing the Residual Stream of Language Models Under Knowledge Conflicts", "abstract": "Large language models (LLMs) can store a significant amount of factual knowledge in their parameters. However, their parametric knowledge may conflict with the information provided in the context. Such conflicts can lead to undesirable model behaviour, such as reliance on outdated or incorrect information. In this work, we investigate whether LLMs can identify knowledge conflicts and whether it is possible to know which source of knowledge the model will rely on by analysing the residual stream of the LLM. Through probing tasks, we find that LLMs can internally register the signal of knowledge conflict in the residual stream, which can be accurately detected by probing the intermediate model activations. This allows us to detect conflicts within the residual stream before generating the answers without modifying the input or model parameters. Moreover, we find that the residual stream shows significantly different patterns when the model relies on contextual knowledge versus parametric knowledge to resolve conflicts. This pattern can be employed to estimate the behaviour of LLMs when conflict happens and prevent unexpected answers before producing the answers. Our analysis offers insights into how LLMs internally manage knowledge conflicts and provides a foundation for developing methods to control the knowledge selection processes.", "authors": ["Yu Zhao", "Xiaotang Du", "Giwon Hong", "Aryo Pradipta Gema", "Alessio Devoto", "Hongru Wang", "Xuanli He", "Kam-Fai Wong", "Pasquale Minervini"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-21", "url": "https://arxiv.org/abs/2410.16090", "pdf_url": "https://arxiv.org/pdf/2410.16090v2", "arxiv_id": "2410.16090", "doi": "10.48550/arXiv.2410.16090", "citation_count": 13, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2865}
{"id": "dd07633c10a669580a8c78aecdbd956a219c7b151b963f13ced2e1a2fcce2856", "sources": ["arxiv", "semantic_scholar"], "title": "Improving General Text Embedding Model: Tackling Task Conflict and Data Imbalance through Model Merging", "abstract": "Text embeddings are vital for tasks such as text retrieval and semantic textual similarity (STS). Recently, the advent of pretrained language models, along with unified benchmarks like the Massive Text Embedding Benchmark (MTEB), has facilitated the development of versatile general-purpose text embedding models. Advanced embedding models are typically developed using large-scale multi-task data and joint training across multiple tasks. However, our experimental analysis reveals two significant drawbacks of joint training: 1) Task Conflict: Gradients from different tasks interfere with each other, leading to negative transfer. 2) Data Imbalance: Disproportionate data distribution introduces biases that negatively impact performance across tasks. To overcome these challenges, we explore model merging-a technique that combines independently trained models to mitigate gradient conflicts and balance data distribution. We introduce a novel method, Self Positioning, which efficiently searches for optimal model combinations within the interpolation space of task vectors using stochastic gradient descent. Our experiments demonstrate that Self Positioning significantly enhances multi-task performance on the MTEB dataset, achieving an absolute improvement of 0.7 points. It outperforms traditional resampling methods while reducing computational costs. This work offers a robust approach to building generalized text embedding models with superior performance across diverse embedding-related tasks.", "authors": ["Mingxin Li", "Zhijie Nie", "Yanzhao Zhang", "Dingkun Long", "Richong Zhang", "Pengjun Xie"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-19", "url": "https://arxiv.org/abs/2410.15035", "pdf_url": "https://arxiv.org/pdf/2410.15035v1", "arxiv_id": "2410.15035", "doi": "10.48550/arXiv.2410.15035", "citation_count": 12, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2785}
{"id": "9f0c5682ecf27f98bccb9409796de1f3f2e054bba666057a9ac5bfc69d0e2e9d", "sources": ["arxiv", "semantic_scholar"], "title": "SurgeryV2: Bridging the Gap Between Model Merging and Multi-Task Learning with Deep Representation Surgery", "abstract": "Model merging-based multitask learning (MTL) offers a promising approach for performing MTL by merging multiple expert models without requiring access to raw training data. However, in this paper, we examine the merged model's representation distribution and uncover a critical issue of \"representation bias\". This bias arises from a significant distribution gap between the representations of the merged and expert models, leading to the suboptimal performance of the merged MTL model. To address this challenge, we first propose a representation surgery solution called Surgery. Surgery is a lightweight, task-specific module that aligns the final layer representations of the merged model with those of the expert models, effectively alleviating bias and improving the merged model's performance. Despite these improvements, a performance gap remains compared to the traditional MTL method. Further analysis reveals that representation bias phenomena exist at each layer of the merged model, and aligning representations only in the last layer is insufficient for fully reducing systemic bias because biases introduced at each layer can accumulate and interact in complex ways. To tackle this, we then propose a more comprehensive solution, deep representation surgery (also called SurgeryV2), which mitigates representation bias across all layers, and thus bridges the performance gap between model merging-based MTL and traditional MTL. Finally, we design an unsupervised optimization objective to optimize both the Surgery and SurgeryV2 modules. Our experimental results show that incorporating these modules into state-of-the-art (SOTA) model merging schemes leads to significant performance gains. Notably, our SurgeryV2 scheme reaches almost the same level as individual expert models or the traditional MTL model. The code is available at \\url{https://github.com/EnnengYang/SurgeryV2}.", "authors": ["Enneng Yang", "Li Shen", "Zhenyi Wang", "Guibing Guo", "Xingwei Wang", "Xiaocun Cao", "Jie Zhang", "Dacheng Tao"], "categories": ["cs.LG", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-18", "url": "https://arxiv.org/abs/2410.14389", "pdf_url": "https://arxiv.org/pdf/2410.14389v1", "arxiv_id": "2410.14389", "doi": "10.48550/arXiv.2410.14389", "citation_count": 12, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/EnnengYang/SurgeryV2}", "venue": "arXiv.org", "quality_score": 0.2785}
{"id": "32e79e110d4fb43a23535cafabb73ed1495d7cb7b567e8ab8fe682776ec2119b", "sources": ["arxiv", "semantic_scholar"], "title": "Mitigating the Backdoor Effect for Multi-Task Model Merging via Safety-Aware Subspace", "abstract": "Model merging has gained significant attention as a cost-effective approach to integrate multiple single-task fine-tuned models into a unified one that can perform well on multiple tasks. However, existing model merging techniques primarily focus on resolving conflicts between task-specific models, they often overlook potential security threats, particularly the risk of backdoor attacks in the open-source model ecosystem. In this paper, we first investigate the vulnerabilities of existing model merging methods to backdoor attacks, identifying two critical challenges: backdoor succession and backdoor transfer. To address these issues, we propose a novel Defense-Aware Merging (DAM) approach that simultaneously mitigates task interference and backdoor vulnerabilities. Specifically, DAM employs a meta-learning-based optimization method with dual masks to identify a shared and safety-aware subspace for model merging. These masks are alternately optimized: the Task-Shared mask identifies common beneficial parameters across tasks, aiming to preserve task-specific knowledge while reducing interference, while the Backdoor-Detection mask isolates potentially harmful parameters to neutralize security threats. This dual-mask design allows us to carefully balance the preservation of useful knowledge and the removal of potential vulnerabilities. Compared to existing merging methods, DAM achieves a more favorable balance between performance and security, reducing the attack success rate by 2-10 percentage points while sacrificing only about 1% in accuracy. Furthermore, DAM exhibits robust performance and broad applicability across various types of backdoor attacks and the number of compromised models involved in the merging process. Our codes and models are available at https://github.com/Yangjinluan/DAM.", "authors": ["Jinluan Yang", "Anke Tang", "Didi Zhu", "Zhengyu Chen", "Li Shen", "Fei Wu"], "categories": ["cs.CR", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-17", "url": "https://arxiv.org/abs/2410.13910", "pdf_url": "https://arxiv.org/pdf/2410.13910v2", "arxiv_id": "2410.13910", "doi": "10.48550/arXiv.2410.13910", "citation_count": 14, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/Yangjinluan/DAM", "venue": "International Conference on Learning Representations", "quality_score": 0.294}
{"id": "8ceffa6e45efb3ad640f580ba684cbf571fc59b71555f83f42ccef2dd9b0569f", "sources": ["arxiv", "semantic_scholar"], "title": "Unconstrained Model Merging for Enhanced LLM Reasoning", "abstract": "Recent advancements in building domain-specific large language models (LLMs) have shown remarkable success, especially in tasks requiring reasoning abilities like logical inference over complex relationships and multi-step problem solving. However, creating a powerful all-in-one LLM remains challenging due to the need for proprietary data and vast computational resources. As a resource-friendly alternative, we explore the potential of merging multiple expert models into a single LLM. Existing studies on model merging mainly focus on generalist LLMs instead of domain experts, or the LLMs under the same architecture and size. In this work, we propose an unconstrained model merging framework that accommodates both homogeneous and heterogeneous model architectures with a focus on reasoning tasks. A fine-grained layer-wise weight merging strategy is designed for homogeneous models merging, while heterogeneous model merging is built upon the probabilistic distribution knowledge derived from instruction-response fine-tuning data. Across 7 benchmarks and 9 reasoning-optimized LLMs, we reveal key findings that combinatorial reasoning emerges from merging which surpasses simple additive effects. We propose that unconstrained model merging could serve as a foundation for decentralized LLMs, marking a notable progression from the existing centralized LLM framework. This evolution could enhance wider participation and stimulate additional advancement in the field of artificial intelligence, effectively addressing the constraints posed by centralized models.", "authors": ["Yiming Zhang", "Baoyi He", "Shengyu Zhang", "Yuhao Fu", "Qi Zhou", "Zhijie Sang", "Zijin Hong", "Kejing Yang", "Wenjun Wang", "Jianbo Yuan", "Guanghan Ning", "Linyi Li", "Chunlin Ji", "Fei Wu", "Hongxia Yang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-17", "url": "https://arxiv.org/abs/2410.13699", "pdf_url": "https://arxiv.org/pdf/2410.13699v2", "arxiv_id": "2410.13699", "doi": "10.48550/arXiv.2410.13699", "citation_count": 6, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2113}
{"id": "af55f7b6f06f0b8bee8bcc743e5d561f4648cf03803271e8d2a9e10b4308ed47", "sources": ["arxiv", "semantic_scholar"], "title": "LoRA Soups: Merging LoRAs for Practical Skill Composition Tasks", "abstract": "Low-Rank Adaptation (LoRA) is a popular technique for parameter-efficient fine-tuning of Large Language Models (LLMs). We study how different LoRA modules can be merged to achieve skill composition -- testing the performance of the merged model on a target task that involves combining multiple skills, each skill coming from a single LoRA. This setup is favorable when it is difficult to obtain training data for the target task and when it can be decomposed into multiple skills. First, we identify practically occurring use-cases that can be studied under the realm of skill composition, e.g. solving hard math-word problems with code, creating a bot to answer questions on proprietary manuals or about domain-specialized corpora. Our main contribution is to show that concatenation of LoRAs (CAT), which optimally weights LoRAs that were individually trained on different skills, outperforms existing model- and data- merging techniques; for instance on math-word problems, CAT beats these methods by an average of 43% and 12% respectively. Thus, this paper advocates model merging as an efficient way to solve compositional tasks and underscores CAT as a simple, compute-friendly and effective procedure. To our knowledge, this is the first work demonstrating the superiority of model merging over data mixing for binary skill composition tasks. Code and data are available at https://github.com/aksh555/LoRA-Soups", "authors": ["Akshara Prabhakar", "Yuanzhi Li", "Karthik Narasimhan", "Sham Kakade", "Eran Malach", "Samy Jelassi"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-16", "url": "https://arxiv.org/abs/2410.13025", "pdf_url": "https://arxiv.org/pdf/2410.13025v2", "arxiv_id": "2410.13025", "doi": "10.48550/arXiv.2410.13025", "citation_count": 45, "influential_citation_count": 3, "has_code": true, "code_url": "https://github.com/aksh555/LoRA-Soups", "venue": "International Conference on Computational Linguistics", "quality_score": 0.4157}
{"id": "93e529dfbbf3e89c0aa50609e0b37d7cd09fe4f39062e84daf7e4da46ee33dba", "sources": ["arxiv", "semantic_scholar"], "title": "The Non-Local Model Merging Problem: Permutation Symmetries and Variance Collapse", "abstract": "Model merging aims to efficiently combine the weights of multiple expert models, each trained on a specific task, into a single multi-task model, with strong performance across all tasks. When applied to all but the last layer of weights, existing methods -- such as Task Arithmetic, TIES-merging, and TALL mask merging -- work well to combine expert models obtained by fine-tuning a common foundation model, operating within a \"local\" neighborhood of the foundation model. This work explores the more challenging scenario of \"non-local\" merging, which we find arises when an expert model changes significantly during pretraining or where the expert models do not even share a common foundation model. We observe that standard merging techniques often fail to generalize effectively in this non-local setting, even when accounting for permutation symmetries using standard techniques. We identify that this failure is, in part, due to \"variance collapse\", a phenomenon identified also in the setting of linear mode connectivity by Jordan et al. (2023). To address this, we propose a multi-task technique to re-scale and shift the output activations of the merged model for each task, aligning its output statistics with those of the corresponding task-specific expert models. Our experiments demonstrate that this correction significantly improves the performance of various model merging approaches in non-local settings, providing a strong baseline for future research on this problem.", "authors": ["Ekansh Sharma", "Daniel M. Roy", "Gintare Karolina Dziugaite"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-16", "url": "https://arxiv.org/abs/2410.12766", "pdf_url": "https://arxiv.org/pdf/2410.12766v1", "arxiv_id": "2410.12766", "doi": "10.48550/arXiv.2410.12766", "citation_count": 6, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2113}
{"id": "6ecba8ab5deb228fdf962be902666a195f08d7d3d34346cd6991dc0ccb0ab377", "sources": ["arxiv", "semantic_scholar"], "title": "Exploring Model Kinship for Merging Large Language Models", "abstract": "Model merging has emerged as a key technique for enhancing the capabilities and efficiency of Large Language Models (LLMs). The open-source community has driven model evolution by iteratively merging existing models, yet a principled understanding of the gains and underlying factors in model merging remains limited. In this work, we study model evolution through iterative merging, drawing an analogy to biological evolution, and introduce the concept of model kinship, the degree of similarity or relatedness between LLMs. Through comprehensive empirical analysis, we show that model kinship is closely linked to the performance improvements achieved by merging, providing a useful criterion for selecting candidate models. Building on this insight, we propose a new model merging strategy: Top-k Greedy Merging with Model Kinship, which can improve benchmark performance. Specifically, we discover that incorporating model kinship as a guiding criterion enables continuous merging while mitigating performance degradation caused by local optima, thereby facilitating more effective model evolution. Code is available at https://github.com/zjunlp/ModelKinship.", "authors": ["Yedi Hu", "Yunzhi Yao", "Ningyu Zhang", "Huajun Chen", "Shumin Deng"], "categories": ["cs.CL", "cs.AI", "cs.CV", "cs.LG", "cs.MA"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-16", "url": "https://arxiv.org/abs/2410.12613", "pdf_url": "https://arxiv.org/pdf/2410.12613v3", "arxiv_id": "2410.12613", "doi": "10.48550/arXiv.2410.12613", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/zjunlp/ModelKinship", "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.0753}
{"id": "224422e8a91446585bec79c6635bc1dd1daa821c65c329cf93a7d84ec46450aa", "sources": ["arxiv", "semantic_scholar"], "title": "Merge to Learn: Efficiently Adding Skills to Language Models with Model Merging", "abstract": "Adapting general-purpose language models to new skills is currently an expensive process that must be repeated as new instruction datasets targeting new skills are created, or can cause the models to forget older skills. In this work, we investigate the effectiveness of adding new skills to preexisting models by training on the new skills in isolation and later merging with the general model (e.g. using task vectors). In experiments focusing on scientific literature understanding, safety, and coding, we find that the parallel-train-then-merge procedure, which is significantly cheaper than retraining the models on updated data mixtures, is often comparably effective. Our experiments also show that parallel training is especially well-suited for enabling safety features in LMs relative to continued finetuning and retraining, as it dramatically improves model compliance with safe prompts while preserving its ability to refuse dangerous or harmful prompts.", "authors": ["Jacob Morrison", "Noah A. Smith", "Hannaneh Hajishirzi", "Pang Wei Koh", "Jesse Dodge", "Pradeep Dasigi"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-16", "url": "https://arxiv.org/abs/2410.12937", "pdf_url": "https://arxiv.org/pdf/2410.12937v1", "arxiv_id": "2410.12937", "doi": "10.48550/arXiv.2410.12937", "citation_count": 7, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.2258}
{"id": "6469384f47c5df07324224ba6feabff2808b6754d685094ef6864b551b700f97", "sources": ["arxiv", "semantic_scholar"], "title": "Efficiera Residual Networks: Hardware-Friendly Fully Binary Weight with 2-bit Activation Model Achieves Practical ImageNet Accuracy", "abstract": "The edge-device environment imposes severe resource limitations, encompassing computation costs, hardware resource usage, and energy consumption for deploying deep neural network models. Ultra-low-bit quantization and hardware accelerators have been explored as promising approaches to address these challenges. Ultra-low-bit quantization significantly reduces the model size and the computational cost. Despite progress so far, many competitive ultra-low-bit models still partially rely on float or non-ultra-low-bit quantized computation such as the input and output layer. We introduce Efficiera Residual Networks (ERNs), a model optimized for low-resource edge devices. ERNs achieve full ultra-low-bit quantization, with all weights, including the initial and output layers, being binary, and activations set at 2 bits. We introduce the shared constant scaling factor technique to enable integer-valued computation in residual connections, allowing our model to operate without float values until the final convolution layer. Demonstrating competitiveness, ERNs achieve an ImageNet top-1 accuracy of 72.5pt with a ResNet50-compatible architecture and 63.6pt with a model size less than 1MB. Moreover, ERNs exhibit impressive inference times, reaching 300FPS with the smallest model and 60FPS with the largest model on a cost-efficient FPGA device.", "authors": ["Shuntaro Takahashi", "Takuya Wakisaka", "Hiroyuki Tokunaga"], "categories": ["cs.CV", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-15", "url": "https://arxiv.org/abs/2410.11553", "pdf_url": "https://arxiv.org/pdf/2410.11553v1", "arxiv_id": "2410.11553", "doi": "10.48550/arXiv.2410.11553", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/LeapMind/ERN", "venue": "arXiv.org", "quality_score": 0.0}
{"id": "1ef0ccd1d7a8048101b774c2c4929e35ffba37935de4d0a4d1b513762bb13853", "sources": ["arxiv", "semantic_scholar"], "title": "Meta-DT: Offline Meta-RL as Conditional Sequence Modeling with World Model Disentanglement", "abstract": "A longstanding goal of artificial general intelligence is highly capable generalists that can learn from diverse experiences and generalize to unseen tasks. The language and vision communities have seen remarkable progress toward this trend by scaling up transformer-based models trained on massive datasets, while reinforcement learning (RL) agents still suffer from poor generalization capacity under such paradigms. To tackle this challenge, we propose Meta Decision Transformer (Meta-DT), which leverages the sequential modeling ability of the transformer architecture and robust task representation learning via world model disentanglement to achieve efficient generalization in offline meta-RL. We pretrain a context-aware world model to learn a compact task representation, and inject it as a contextual condition to the causal transformer to guide task-oriented sequence generation. Then, we subtly utilize history trajectories generated by the meta-policy as a self-guided prompt to exploit the architectural inductive bias. We select the trajectory segment that yields the largest prediction error on the pretrained world model to construct the prompt, aiming to encode task-specific information complementary to the world model maximally. Notably, the proposed framework eliminates the requirement of any expert demonstration or domain knowledge at test time. Experimental results on MuJoCo and Meta-World benchmarks across various dataset types show that Meta-DT exhibits superior few and zero-shot generalization capacity compared to strong baselines while being more practical with fewer prerequisites. Our code is available at https://github.com/NJU-RL/Meta-DT.", "authors": ["Zhi Wang", "Li Zhang", "Wenhao Wu", "Yuanheng Zhu", "Dongbin Zhao", "Chunlin Chen"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-15", "url": "https://arxiv.org/abs/2410.11448", "pdf_url": "https://arxiv.org/pdf/2410.11448v2", "arxiv_id": "2410.11448", "doi": "10.48550/arXiv.2410.11448", "citation_count": 23, "influential_citation_count": 3, "has_code": true, "code_url": "https://github.com/NJU-RL/Meta-DT", "venue": "Neural Information Processing Systems", "quality_score": 0.3451}
{"id": "ef3e9fe995c5a672c54c672717f21a3da99f8862544b1a5b4446065bebab1e69", "sources": ["arxiv", "semantic_scholar"], "title": "Cusp types of arithmetic hyperbolic manifolds", "abstract": "We establish necessary and sufficient conditions for determining when a flat manifold can occur as a cusp cross-section within a given commensurability class of cusped arithmetic hyperbolic manifolds. This reduces the problem of identifying which commensurability classes of arithmetic hyperbolic manifolds can contain a specific flat manifold as a cusp cross-section to a question involving rational representations of the flat manifold's holonomy group. More generally we show that the holonomy representation provides an obstruction on the quasi-arithmetic manifolds containing a given flat manifold as a cusp cross-section. As applications, we prove that a flat manifold $M$ with a holonomy group of odd order appears as a cusp cross-section in every commensurability class of arithmetic hyperbolic manifolds if and only if $b_1(M)\\geq 3$. We also provide examples of flat manifolds that arise as cusp cross-sections in a unique commensurability class of arithmetic hyperbolic manifolds and exhibit examples of pairs of flat manifolds that can never appear as cusp cross-sections in the same quasi-arithmetic hyperbolic manifold.", "authors": ["Duncan McCoy", "Connor Sell"], "categories": ["math.GT"], "fields_of_study": ["Mathematics"], "published_date": "2024-10-14", "url": "https://arxiv.org/abs/2410.10707", "pdf_url": "https://arxiv.org/pdf/2410.10707v3", "arxiv_id": "2410.10707", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.0}
{"id": "41a332f08630dfa6be0929d95b10697485530ff0464c623d7b4e3ac065c9eeee", "sources": ["arxiv", "semantic_scholar"], "title": "Mix Data or Merge Models? Optimizing for Diverse Multi-Task Learning", "abstract": "Large Language Models (LLMs) have been adopted and deployed worldwide for a broad variety of applications. However, ensuring their safe use remains a significant challenge. Preference training and safety measures often overfit to harms prevalent in Western-centric datasets, and safety protocols frequently fail to extend to multilingual settings. In this work, we explore model merging in a diverse multi-task setting, combining safety and general-purpose tasks within a multilingual context. Each language introduces unique and varied learning challenges across tasks. We find that objective-based merging is more effective than mixing data, with improvements of up to 8% and 10% in general performance and safety respectively. We also find that language-based merging is highly effective -- by merging monolingually fine-tuned models, we achieve a 4% increase in general performance and 7% reduction in harm across all languages on top of the data mixtures method using the same available data. Overall, our comprehensive study of merging approaches provides a useful framework for building strong and safe multilingual models.", "authors": [" Aakanksha", "Arash Ahmadian", "Seraphina Goldfarb-Tarrant", "Beyza Ermis", "Marzieh Fadaee", "Sara Hooker"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-14", "url": "https://arxiv.org/abs/2410.10801", "pdf_url": "https://arxiv.org/pdf/2410.10801v1", "arxiv_id": "2410.10801", "doi": "10.48550/arXiv.2410.10801", "citation_count": 17, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3138}
{"id": "f9e987418a2bb42083c29a6d83726f9cb46a1519245dcf40d1a1abb63caa5fcb", "sources": ["arxiv", "semantic_scholar"], "title": "NegMerge: Sign-Consensual Weight Merging for Machine Unlearning", "abstract": "Machine unlearning aims to selectively remove specific knowledge from a trained model. Existing approaches, such as Task Arithmetic, fine-tune the model on the forget set to create a task vector (i.e., a direction in weight space) for subtraction from the original model's weight. However, their effectiveness is highly sensitive to hyperparameter selection, requiring extensive validation to identify the optimal vector from many fine-tuned candidates. In this paper, we propose a novel method that utilizes all fine-tuned models trained with varying hyperparameters instead of a single selection. Specifically, we aggregate the computed task vectors by retaining only the elements with consistent shared signs. The merged task vector is then negated to induce unlearning on the original model. Evaluations on zero-shot and standard image recognition tasks across twelve datasets and four backbone architectures show that our approach outperforms state-of-the-art methods while requiring similar or fewer computational resources. Code is available at https://github.com/naver-ai/negmerge.", "authors": ["Hyo Seo Kim", "Dongyoon Han", "Junsuk Choe"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-08", "url": "https://arxiv.org/abs/2410.05583", "pdf_url": "https://arxiv.org/pdf/2410.05583v2", "arxiv_id": "2410.05583", "doi": null, "citation_count": 5, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/naver-ai/negmerge", "venue": "International Conference on Machine Learning", "quality_score": 0.1945}
{"id": "737348cf6407440c77c9a62139d8eb93524a9f05108131e80095c8bbcae098df", "sources": ["arxiv", "semantic_scholar"], "title": "What Matters for Model Merging at Scale?", "abstract": "Model merging aims to combine multiple expert models into a more capable single model, offering benefits such as reduced storage and serving costs, improved generalization, and support for decentralized model development. Despite its promise, previous studies have primarily focused on merging a few small models. This leaves many unanswered questions about the effect of scaling model size and how it interplays with other key factors -- like the base model quality and number of expert models -- , to affect the merged model's performance. This work systematically evaluates the utility of model merging at scale, examining the impact of these different factors. We experiment with merging fully fine-tuned models using 4 popular merging methods -- Averaging, Task~Arithmetic, Dare, and TIES -- across model sizes ranging from 1B-64B parameters and merging up to 8 different expert models. We evaluate the merged models on both held-in tasks, i.e., the expert's training tasks, and zero-shot generalization to unseen held-out tasks. Our experiments provide several new insights about model merging at scale and the interplay between different factors. First, we find that merging is more effective when experts are created from strong base models, i.e., models with good zero-shot performance. Second, larger models facilitate easier merging. Third merging consistently improves generalization capabilities. Notably, when merging 8 large expert models, the merged models often generalize better compared to the multitask trained models. Fourth, we can better merge more expert models when working with larger models. Fifth, different merging methods behave very similarly at larger scales. Overall, our findings shed light on some interesting properties of model merging while also highlighting some limitations. We hope that this study will serve as a reference point on large-scale merging for upcoming research.", "authors": ["Prateek Yadav", "Tu Vu", "Jonathan Lai", "Alexandra Chronopoulou", "Manaal Faruqui", "Mohit Bansal", "Tsendsuren Munkhdalai"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-04", "url": "https://arxiv.org/abs/2410.03617", "pdf_url": "https://arxiv.org/pdf/2410.03617v1", "arxiv_id": "2410.03617", "doi": "10.48550/arXiv.2410.03617", "citation_count": 56, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.439}
{"id": "72187bb35bf5fa214ac2ea462dc316e50e4370775132907ef09062e0d3f077b9", "sources": ["arxiv", "semantic_scholar"], "title": "A Comprehensive Review of Propagation Models in Complex Networks: From Deterministic to Deep Learning Approaches", "abstract": "Understanding propagation mechanisms in complex networks is essential for fields like epidemiology and multi-robot networks. This paper reviews various propagation models, from traditional deterministic frameworks to advanced data-driven and deep learning approaches. We differentiate between static and dynamic networks, noting that static models provide foundational insights, while dynamic models capture real-world temporal changes. Deterministic models like the SIR framework offer clear mathematical insights but often lack adaptability to randomness, whereas stochastic models enhance realism at the cost of interpretability. Behavior-based models focus on individual decision-making, demanding more computational resources. Data-driven approaches improve accuracy in nonlinear scenarios by adapting to evolving networks, using either traditional models or model-free machine learning techniques. We explore supervised and unsupervised learning methods, as well as reinforcement learning, which operates without predefined datasets. The application of graph neural networks (GNNs) is also discussed, highlighting their effectiveness in modeling propagation in complex networks. The paper underscores key applications and challenges associated with each model type, emphasizing the increasing importance of hybrid and machine learning-based solutions in contemporary network propagation issues.", "authors": ["Bin Wu", "Sifu Luo", "C. Steve Suh"], "categories": ["cs.SI"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-03", "url": "https://arxiv.org/abs/2410.02118", "pdf_url": "https://arxiv.org/pdf/2410.02118v1", "arxiv_id": "2410.02118", "doi": "10.48550/arXiv.2410.02118", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Journal of Vibration Testing and System Dynamics", "quality_score": 0.0753}
{"id": "cc756c9e77294adec127b6e231ff4ace18dc8dfc599d51fbdca658581d14be10", "sources": ["arxiv", "semantic_scholar"], "title": "DaWin: Training-free Dynamic Weight Interpolation for Robust Adaptation", "abstract": "Adapting a pre-trained foundation model on downstream tasks should ensure robustness against distribution shifts without the need to retrain the whole model. Although existing weight interpolation methods are simple yet effective, we argue that their static nature limits downstream performance while achieving efficiency. In this work, we propose DaWin, a training-free dynamic weight interpolation method that leverages the entropy of individual models over each unlabeled test sample to assess model expertise, and compute per-sample interpolation coefficients dynamically. Unlike previous works that typically rely on additional training to learn such coefficients, our approach requires no training. Then, we propose a mixture modeling approach that greatly reduces inference overhead raised by dynamic interpolation. We validate DaWin on the large-scale visual recognition benchmarks, spanning 14 tasks across robust fine-tuning -- ImageNet and derived five distribution shift benchmarks -- and multi-task learning with eight classification tasks. Results demonstrate that DaWin achieves significant performance gain in considered settings, with minimal computational overhead. We further discuss DaWin's analytic behavior to explain its empirical success.", "authors": ["Changdae Oh", "Yixuan Li", "Kyungwoo Song", "Sangdoo Yun", "Dongyoon Han"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-03", "url": "https://arxiv.org/abs/2410.03782", "pdf_url": "https://arxiv.org/pdf/2410.03782v4", "arxiv_id": "2410.03782", "doi": "10.48550/arXiv.2410.03782", "citation_count": 18, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Conference on Learning Representations", "quality_score": 0.3197}
{"id": "f8c71e01b3cf8c6dbbb4eb4ffca50a0fc1319523009d795f05bf72314cdcf0fc", "sources": ["arxiv", "semantic_scholar"], "title": "Parameter Competition Balancing for Model Merging", "abstract": "While fine-tuning pretrained models has become common practice, these models often underperform outside their specific domains. Recently developed model merging techniques enable the direct integration of multiple models, each fine-tuned for distinct tasks, into a single model. This strategy promotes multitasking capabilities without requiring retraining on the original datasets. However, existing methods fall short in addressing potential conflicts and complex correlations between tasks, especially in parameter-level adjustments, posing a challenge in effectively balancing parameter competition across various tasks. This paper introduces an innovative technique named PCB-Merging (Parameter Competition Balancing), a lightweight and training-free technique that adjusts the coefficients of each parameter for effective model merging. PCB-Merging employs intra-balancing to gauge parameter significance within individual tasks and inter-balancing to assess parameter similarities across different tasks. Parameters with low importance scores are dropped, and the remaining ones are rescaled to form the final merged model. We assessed our approach in diverse merging scenarios, including cross-task, cross-domain, and cross-training configurations, as well as out-of-domain generalization. The experimental results reveal that our approach achieves substantial performance enhancements across multiple modalities, domains, model sizes, number of tasks, fine-tuning forms, and large language models, outperforming existing model merging methods. The code is publicly available at: \\url{https://github.com/duguodong7/pcb-merging}.", "authors": ["Guodong Du", "Junlin Lee", "Jing Li", "Runhua Jiang", "Yifei Guo", "Shuyang Yu", "Hanting Liu", "Sim Kuan Goh", "Ho-Kin Tang", "Daojing He", "Min Zhang"], "categories": ["cs.CV", "cs.AI", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-03", "url": "https://arxiv.org/abs/2410.02396", "pdf_url": "https://arxiv.org/pdf/2410.02396v1", "arxiv_id": "2410.02396", "doi": "10.48550/arXiv.2410.02396", "citation_count": 73, "influential_citation_count": 15, "has_code": true, "code_url": "https://github.com/duguodong7/pcb-merging}", "venue": "Neural Information Processing Systems", "quality_score": 0.6021}
{"id": "739a857f49fc5518dcfe8ac9e6602a0810cbb8495487d5dea9581bccd27aeafb", "sources": ["arxiv", "semantic_scholar"], "title": "Foldable SuperNets: Scalable Merging of Transformers with Different Initializations and Tasks", "abstract": "Recent methods aim to merge neural networks (NNs) with identical architectures trained on different tasks into a single multi-task model. While most works focus on the simpler setup of merging NNs initialized from a common pre-trained network, we target the harder problem of merging large transformers trained on different tasks from distinct initializations. We show that traditional merging methods fail catastrophically in this setup, while Knowledge Distillation (KD) achieves much better results, though at a higher cost. However, KD is data-inefficient, as it does not exploit the original models' weights. To solve this, we introduce \"Foldable SuperNet Merge\" (FS-Merge), which trains a SuperNet containing the original models (with frozen weights) using a feature reconstruction objective. After training, the SuperNet is folded back to the size of a single original model. FS-Merge is simple, data-efficient, has a computational cost comparable to KD, and is proven to have superior expressiveness compared to traditional merging methods on MLP models. It achieves SOTA results when tested on MLPs and transformers across various sizes, tasks, modalities, and distribution shifts, especially in low-data scenarios.", "authors": ["Edan Kinderman", "Itay Hubara", "Haggai Maron", "Daniel Soudry"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-02", "url": "https://arxiv.org/abs/2410.01483", "pdf_url": "https://arxiv.org/pdf/2410.01483v2", "arxiv_id": "2410.01483", "doi": "10.48550/arXiv.2410.01483", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Transactions on Machine Learning Research, 2025", "quality_score": 0.1747}
{"id": "a7985cf63e30c857ed53c7f37a53cd98460c20bcd14ceb8d672980ae623899a1", "sources": ["arxiv", "semantic_scholar"], "title": "AVID: Adapting Video Diffusion Models to World Models", "abstract": "Large-scale generative models have achieved remarkable success in a number of domains. However, for sequential decision-making problems, such as robotics, action-labelled data is often scarce and therefore scaling-up foundation models for decision-making remains a challenge. A potential solution lies in leveraging widely-available unlabelled videos to train world models that simulate the consequences of actions. If the world model is accurate, it can be used to optimize decision-making in downstream tasks. Image-to-video diffusion models are already capable of generating highly realistic synthetic videos. However, these models are not action-conditioned, and the most powerful models are closed-source which means they cannot be finetuned. In this work, we propose to adapt pretrained video diffusion models to action-conditioned world models, without access to the parameters of the pretrained model. Our approach, AVID, trains an adapter on a small domain-specific dataset of action-labelled videos. AVID uses a learned mask to modify the intermediate outputs of the pretrained model and generate accurate action-conditioned videos. We evaluate AVID on video game and real-world robotics data, and show that it outperforms existing baselines for diffusion model adaptation.1 Our results demonstrate that if utilized correctly, pretrained video models have the potential to be powerful tools for embodied AI.", "authors": ["Marc Rigter", "Tarun Gupta", "Agrin Hilmkil", "Chao Ma"], "categories": ["cs.CV", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-01", "url": "https://arxiv.org/abs/2410.12822", "pdf_url": "https://arxiv.org/pdf/2410.12822v2", "arxiv_id": "2410.12822", "doi": "10.48550/arXiv.2410.12822", "citation_count": 29, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3693}
{"id": "2db5735bb4fbbc718c24fd84f6e0ffcad8daa35b7d9e8744ce430de6ccefb608", "sources": ["arxiv", "semantic_scholar"], "title": "HM3: Hierarchical Multi-Objective Model Merging for Pretrained Models", "abstract": "Model merging is a technique that combines multiple large pretrained models into a single model with enhanced performance and broader task adaptability. It has gained popularity in large pretrained model development due to its ability to bypass the need for original training data and further training processes. However, most existing model merging approaches focus solely on exploring the parameter space, merging models with identical architectures. Merging within the architecture space, despite its potential, remains in its early stages due to the vast search space and the challenges of layer compatibility. This paper marks a significant advance toward more flexible and comprehensive model merging techniques by modeling the architecture-space merging process as a reinforcement learning task. We train policy and value networks using offline sampling of weight vectors, which are then employed for the online optimization of merging strategies. Moreover, a multi-objective optimization paradigm is introduced to accommodate users' diverse task preferences, learning the Pareto front of optimal models to offer customized merging suggestions. Experimental results across multiple tasks, including text translation, mathematical reasoning, and code generation, validate the effectiveness and superiority of the proposed framework in model merging. The code will be made publicly available after the review process.", "authors": ["Yu Zhou", "Xingyu Wu", "Jibin Wu", "Liang Feng", "Kay Chen Tan"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-09-27", "url": "https://arxiv.org/abs/2409.18893", "pdf_url": "https://arxiv.org/pdf/2409.18893v1", "arxiv_id": "2409.18893", "doi": "10.48550/arXiv.2409.18893", "citation_count": 10, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2603}
{"id": "d5bbce50a32cab7530559e587a4e9573e079aa85c78a243fa2899bff360d50bf", "sources": ["arxiv", "semantic_scholar"], "title": "HM3: Heterogeneous Multi-Class Model Merging", "abstract": "Foundation language model deployments often include auxiliary guard-rail models to filter or classify text, detecting jailbreak attempts, biased or toxic output, or ensuring topic adherence. These additional models increase the complexity and cost of model inference, especially since many are also large language models. To address this issue, we explore training-free model merging techniques to consolidate these models into a single, multi-functional model. We propose Heterogeneous Multi-Class Model Merging (HM3) as a simple technique for merging multi-class classifiers with heterogeneous label spaces. Unlike parameter-efficient fine-tuning techniques like LoRA, which require extensive training and add complexity during inference, recent advancements allow models to be merged in a training-free manner. We report promising results for merging BERT-based guard models, some of which attain an average F1-score higher than the source models while reducing the inference time by up to 44%. We introduce self-merging to assess the impact of reduced task-vector density, finding that the more poorly performing hate speech classifier benefits from self-merging while higher-performing classifiers do not, which raises questions about using task vector reduction for model tuning.", "authors": ["Stefan Hackmann"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-09-27", "url": "https://arxiv.org/abs/2409.19173", "pdf_url": "https://arxiv.org/pdf/2409.19173v1", "arxiv_id": "2409.19173", "doi": "10.48550/arXiv.2409.19173", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0753}
{"id": "8b75300f0d87d0142ade57e7cd85b0354165287de8204392bc087f412d3c793c", "sources": ["arxiv", "semantic_scholar"], "title": "Layer-wise Model Merging for Unsupervised Domain Adaptation in Segmentation Tasks", "abstract": "Merging parameters of multiple models has resurfaced as an effective strategy to enhance task performance and robustness, but prior work is limited by the high costs of ensemble creation and inference. In this paper, we leverage the abundance of freely accessible trained models to introduce a cost-free approach to model merging. It focuses on a layer-wise integration of merged models, aiming to maintain the distinctiveness of the task-specific final layers while unifying the initial layers, which are primarily associated with feature extraction. This approach ensures parameter consistency across all layers, essential for boosting performance. Moreover, it facilitates seamless integration of knowledge, enabling effective merging of models from different datasets and tasks. Specifically, we investigate its applicability in Unsupervised Domain Adaptation (UDA), an unexplored area for model merging, for Semantic and Panoptic Segmentation. Experimental results demonstrate substantial UDA improvements without additional costs for merging same-architecture models from distinct datasets ($\\uparrow 2.6\\%$ mIoU) and different-architecture models with a shared backbone ($\\uparrow 6.8\\%$ mIoU). Furthermore, merging Semantic and Panoptic Segmentation models increases mPQ by $\\uparrow 7\\%$. These findings are validated across a wide variety of UDA strategies, architectures, and datasets.", "authors": ["Roberto Alcover-Couso", "Juan C. SanMiguel", "Marcos Escudero-Viñolo", "Jose M Martínez"], "categories": ["cs.CV", "cs.AI", "cs.MM"], "fields_of_study": ["Computer Science"], "published_date": "2024-09-24", "url": "https://arxiv.org/abs/2409.15813", "pdf_url": "https://arxiv.org/pdf/2409.15813v1", "arxiv_id": "2409.15813", "doi": "10.1007/s00371-025-03843-7", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "The Visual Computer", "quality_score": 0.1747}
{"id": "ea8e8f994b2e46291cbecf9d61a055f7a6d8fac99c7fcfd1fec6f890b3f69f1a", "sources": ["arxiv", "semantic_scholar"], "title": "Interpreting Arithmetic Mechanism in Large Language Models through Comparative Neuron Analysis", "abstract": "We find arithmetic ability resides within a limited number of attention heads, with each head specializing in distinct operations. To delve into the reason, we introduce the Comparative Neuron Analysis (CNA) method, which identifies an internal logic chain consisting of four distinct stages from input to prediction: feature enhancing with shallow FFN neurons, feature transferring by shallow attention layers, feature predicting by arithmetic heads, and prediction enhancing among deep FFN neurons. Moreover, we identify the human-interpretable FFN neurons within both feature-enhancing and feature-predicting stages. These findings lead us to investigate the mechanism of LoRA, revealing that it enhances prediction probabilities by amplifying the coefficient scores of FFN neurons related to predictions. Finally, we apply our method in model pruning for arithmetic tasks and model editing for reducing gender bias. Code is on https://github.com/zepingyu0512/arithmetic-mechanism.", "authors": ["Zeping Yu", "Sophia Ananiadou"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-09-21", "url": "https://arxiv.org/abs/2409.14144", "pdf_url": "https://arxiv.org/pdf/2409.14144v1", "arxiv_id": "2409.14144", "doi": "10.48550/arXiv.2409.14144", "citation_count": 38, "influential_citation_count": 4, "has_code": true, "code_url": "https://github.com/zepingyu0512/arithmetic-mechanism", "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.3978}
{"id": "e6dd178671b723a542a0cb442407becd16f055802d2a39cf103be13d7214dbd7", "sources": ["arxiv", "semantic_scholar"], "title": "Task Arithmetic for Language Expansion in Speech Translation", "abstract": "Recent progress in large language models (LLMs) has gained interest in speech-text multimodal foundation models, achieving strong performance on instruction-tuned speech translation (ST). However, expanding language pairs is costly due to re-training on combined new and previous datasets. To address this, we aim to build a one-to-many ST system from existing one-to-one ST systems using task arithmetic without re-training. Direct application of task arithmetic in ST leads to language confusion; therefore, we introduce an augmented task arithmetic method incorporating a language control model to ensure correct target language generation. Our experiments on MuST-C and CoVoST-2 show BLEU score improvements of up to 4.66 and 4.92, with COMET gains of 8.87 and 11.83. In addition, we demonstrate our framework can extend to language pairs lacking paired ST training data or pre-trained ST models by synthesizing ST models based on existing machine translation (MT) and ST models via task analogies.", "authors": ["Yao-Fei Cheng", "Hayato Futami", "Yosuke Kashiwagi", "Emiru Tsunoo", "Wen Shen Teo", "Siddhant Arora", "Shinji Watanabe"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-09-17", "url": "https://arxiv.org/abs/2409.11274", "pdf_url": "https://arxiv.org/pdf/2409.11274v3", "arxiv_id": "2409.11274", "doi": "10.48550/arXiv.2409.11274", "citation_count": 6, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2113}
{"id": "44e4eeca8fa36eae46cbeacb03907ff424a6ee3a8279ee55cbc094b452337921", "sources": ["arxiv", "semantic_scholar"], "title": "Existence of an extremal function of Sobolev critical embedding with an $α$-homogeneous weight", "abstract": "In our previous publication [{\\em Calc. Var. Partial Differential Equations}, 60(1):Paper No. 16, 27, 2021], we delved into examining a critical Sobolev-type embedding of a Sobolev weighted space into an exponential weighted Orlicz space. We specifically determined the optimal Moser-type constant for this embedding, utilizing the monomial weight introduced by Cabré and Ros-Oton [{\\em J. Differential Equations}, 255(11):4312--4336, 2013]. Towards the conclusion of that paper, we pledged to explore the existence of an extremal function within this framework. In this current work, we not only provide a positive affirmation to this inquiry but extend it to a broader range of weights known as \\emph{$α$-homogeneous weights}.", "authors": ["Petr Gurka", "Daniel Hauer"], "categories": ["math.AP"], "fields_of_study": ["Mathematics"], "published_date": "2024-09-17", "url": "https://arxiv.org/abs/2409.11193", "pdf_url": "https://arxiv.org/pdf/2409.11193v1", "arxiv_id": "2409.11193", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.0}
{"id": "c5aee12b6cc15de8a9228b164b5a3458dc338bc7d36af93e64e009128764b0d9", "sources": ["arxiv", "semantic_scholar"], "title": "xLAM: A Family of Large Action Models to Empower AI Agent Systems", "abstract": "Autonomous agents powered by large language models (LLMs) have attracted significant research interest. However, the open-source community faces many challenges in developing specialized models for agent tasks, driven by the scarcity of high-quality agent datasets and the absence of standard protocols in this area. We introduce and publicly release xLAM, a series of large action models designed for AI agent tasks. The xLAM series includes five models with both dense and mixture-of-expert architectures, ranging from 1B to 8x22B parameters, trained using a scalable, flexible pipeline that unifies, augments, and synthesizes diverse datasets to enhance AI agents' generalizability and performance across varied environments. Our experimental results demonstrate that xLAM consistently delivers exceptional performance across multiple agent ability benchmarks, notably securing the 1st position on the Berkeley Function-Calling Leaderboard, outperforming GPT-4, Claude-3, and many other models in terms of tool use. By releasing the xLAM series, we aim to advance the performance of open-source LLMs for autonomous AI agents, potentially accelerating progress and democratizing access to high-performance models for agent tasks. Models are available at https://huggingface.co/collections/Salesforce/xlam-models-65f00e2a0a63bbcd1c2dade4", "authors": ["Jianguo Zhang", "Tian Lan", "Ming Zhu", "Zuxin Liu", "Thai Hoang", "Shirley Kokane", "Weiran Yao", "Juntao Tan", "Akshara Prabhakar", "Haolin Chen", "Zhiwei Liu", "Yihao Feng", "Tulika Awalgaonkar", "Rithesh Murthy", "Eric Hu", "Zeyuan Chen", "Ran Xu", "Juan Carlos Niebles", "Shelby Heinecke", "Huan Wang", "Silvio Savarese", "Caiming Xiong"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-09-05", "url": "https://arxiv.org/abs/2409.03215", "pdf_url": "https://arxiv.org/pdf/2409.03215v1", "arxiv_id": "2409.03215", "doi": "10.48550/arXiv.2409.03215", "citation_count": 114, "influential_citation_count": 7, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5152}
{"id": "8c317bd5c6a12b7873ae61d59fb633aa01cc9d6997a7c9714c0d61deea4e7ed2", "sources": ["arxiv", "semantic_scholar"], "title": "Localize-and-Stitch: Efficient Model Merging via Sparse Task Arithmetic", "abstract": "Model merging offers an effective strategy to combine the strengths of multiple finetuned models into a unified model that preserves the specialized capabilities of each. Existing methods merge models in a global manner, performing arithmetic operations across all model parameters. However, such global merging often leads to task interference, degrading the performance of the merged model. In this work, we introduce Localize-and-Stitch, a novel approach that merges models in a localized way. Our algorithm works in two steps: i) Localization: identify tiny ($1\\%$ of the total parameters) localized regions in the finetuned models containing essential skills for the downstream tasks, and ii) Stitching: reintegrate only these essential regions back into the pretrained model for task synergy. We demonstrate that our approach effectively locates sparse regions responsible for finetuned performance, and the localized regions could be treated as compact and interpretable representations of the finetuned models (tasks). Empirically, we evaluate our method on various vision and language benchmarks, showing that it outperforms existing model merging methods under different data availability scenarios. Beyond strong empirical performance, our algorithm also facilitates model compression and preserves pretrained knowledge, enabling flexible and continual skill composition from multiple finetuned models with minimal storage and computational overhead. Our code is available at https://github.com/uiuctml/Localize-and-Stitch.", "authors": ["Yifei He", "Yuzheng Hu", "Yong Lin", "Tong Zhang", "Han Zhao"], "categories": ["cs.LG", "cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-08-24", "url": "https://arxiv.org/abs/2408.13656", "pdf_url": "https://arxiv.org/pdf/2408.13656v2", "arxiv_id": "2408.13656", "doi": "10.48550/arXiv.2408.13656", "citation_count": 43, "influential_citation_count": 3, "has_code": true, "code_url": "https://github.com/uiuctml/Localize-and-Stitch", "venue": null, "quality_score": 0.4109}
{"id": "0cdb05397f92a30c0627ef4d1eab33c90330936fdda7446c5e78d54e8c8050cf", "sources": ["arxiv", "semantic_scholar"], "title": "Weight Scope Alignment: A Frustratingly Easy Method for Model Merging", "abstract": "Merging models becomes a fundamental procedure in some applications that consider model efficiency and robustness. The training randomness or Non-I.I.D. data poses a huge challenge for averaging-based model fusion. Previous research efforts focus on element-wise regularization or neural permutations to enhance model averaging while overlooking weight scope variations among models, which can significantly affect merging effectiveness. In this paper, we reveal variations in weight scope under different training conditions, shedding light on its influence on model merging. Fortunately, the parameters in each layer basically follow the Gaussian distribution, which inspires a novel and simple regularization approach named Weight Scope Alignment (WSA). It contains two key components: 1) leveraging a target weight scope to guide the model training process for ensuring weight scope matching in the subsequent model merging. 2) fusing the weight scope of two or more models into a unified one for multi-stage model fusion. We extend the WSA regularization to two different scenarios, including Mode Connectivity and Federated Learning. Abundant experimental studies validate the effectiveness of our approach.", "authors": ["Yichu Xu", "Xin-Chun Li", "Le Gan", "De-Chuan Zhan"], "categories": ["cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-08-22", "url": "https://arxiv.org/abs/2408.12237", "pdf_url": "https://arxiv.org/pdf/2408.12237v1", "arxiv_id": "2408.12237", "doi": "10.48550/arXiv.2408.12237", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "European Conference on Artificial Intelligence", "quality_score": 0.1505}
{"id": "53b2af75ca435a096749098e555dadd8db6fe0b1fd18f525fc4cb06829895df2", "sources": ["arxiv", "semantic_scholar"], "title": "Pareto Merging: Multi-Objective Optimization for Preference-Aware Model Merging", "abstract": "Model merging, which combines multiple models into a single model, has gained popularity in recent years. By efficiently integrating the capabilities of various models, this significantly reduces the parameter count and memory usage. However, current methods can only produce one single merged model. This necessitates a performance trade-off due to conflicts among the various models, and the resultant one-size-fits-all model may not align with the preferences of different users who may prioritize certain models over others. To address this issue, we propose preference-aware model merging, and formulate this as a multi-objective optimization problem in which the performance of the merged model on each base model's task is treated as an objective. In a single merging process, the proposed parameter-efficient structure generates a Pareto set of merged models, with each representing a Pareto-optimal solution for a preference. Users can then select merged models tailored to their preferences from this learned Pareto set. Experimental results demonstrate that the proposed Pareto Merging produces diverse trade-off models and achieves higher test accuracy compared to state-of-the-art merging baselines.", "authors": ["Weiyu Chen", "James Kwok"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-08-22", "url": "https://arxiv.org/abs/2408.12105", "pdf_url": "https://arxiv.org/pdf/2408.12105v2", "arxiv_id": "2408.12105", "doi": null, "citation_count": 10, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.2603}
{"id": "b6b9621be3e07583fe4e1f30a08a562d43ca89542904b2125a73a0b35b912af2", "sources": ["arxiv", "semantic_scholar"], "title": "MergeRepair: An Exploratory Study on Merging Task-Specific Adapters in Code LLMs for Automated Program Repair", "abstract": "Large Language Models (LLMs) have shown high capabilities in several software development-related tasks such as program repair, documentation, code refactoring, debugging, and testing. However, training these models requires massive amount of data and significant computational resources. Adapters are specialized, small modules designed for parameter efficient fine-tuning of LLMs for specific tasks, domains, or applications without requiring extensive retraining of the entire model. These adapters offer a more efficient way to customize LLMs for particular needs, leveraging the pre-existing capabilities of the large model. Model (and adapter) merging have emerged as a technique to develop one model capable of multiple tasks, with minimal or no training required. Although model and adapter merging has shown promising performance in domains such as natural language processing and computer vision, its applicability to software engineering tasks remains underexplored. In this paper, we investigate the effectiveness of merged adapters within the context of software engineering, with a particular focus on the Automated Program Repair (APR) task, through our approach, MergeRepair. In particular, we merge multiple task-specific adapters using three different merging methods, including weight-averaging, ties, and dare-ties, and evaluate the performance of the merged adapter on the APR task. We introduce a continual merging approach, a novel method in which we sequentially merge the task-specific adapters where the order and weight of the merged adapters play a significant role. We further compare the performance of our approach with a baseline method consisting of equal-weight merging applied on parameters of different adapters, where all adapters are of equal importance.", "authors": ["Meghdad Dehghan", "Jie JW Wu", "Fatemeh H. Fard", "Ali Ouni"], "categories": ["cs.SE", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-08-18", "url": "https://arxiv.org/abs/2408.09568", "pdf_url": "https://arxiv.org/pdf/2408.09568v3", "arxiv_id": "2408.09568", "doi": "10.48550/arXiv.2408.09568", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "0f00e17886fd523c38fc8f0cd029bb52476a749738558d71a094e8deb7fc99b2", "sources": ["arxiv", "semantic_scholar"], "title": "BadMerging: Backdoor Attacks Against Model Merging", "abstract": "Fine-tuning pre-trained models for downstream tasks has led to a proliferation of open-sourced task-specific models. Recently, Model Merging (MM) has emerged as an effective approach to facilitate knowledge transfer among these independently fine-tuned models. MM directly combines multiple fine-tuned task-specific models into a merged model without additional training, and the resulting model shows enhanced capabilities in multiple tasks. Although MM provides great utility, it may come with security risks because an adversary can exploit MM to affect multiple downstream tasks. However, the security risks of MM have barely been studied. In this paper, we first find that MM, as a new learning paradigm, introduces unique challenges for existing backdoor attacks due to the merging process. To address these challenges, we introduce BadMerging, the first backdoor attack specifically designed for MM. Notably, BadMerging allows an adversary to compromise the entire merged model by contributing as few as one backdoored task-specific model. BadMerging comprises a two-stage attack mechanism and a novel feature-interpolation-based loss to enhance the robustness of embedded backdoors against the changes of different merging parameters. Considering that a merged model may incorporate tasks from different domains, BadMerging can jointly compromise the tasks provided by the adversary (on-task attack) and other contributors (off-task attack) and solve the corresponding unique challenges with novel attack designs. Extensive experiments show that BadMerging achieves remarkable attacks against various MM algorithms. Our ablation study demonstrates that the proposed attack designs can progressively contribute to the attack performance. Finally, we show that prior defense mechanisms fail to defend against our attacks, highlighting the need for more advanced defense.", "authors": ["Jinghuai Zhang", "Jianfeng Chi", "Zheng Li", "Kunlin Cai", "Yang Zhang", "Yuan Tian"], "categories": ["cs.CR", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-08-14", "url": "https://arxiv.org/abs/2408.07362", "pdf_url": "https://arxiv.org/pdf/2408.07362v2", "arxiv_id": "2408.07362", "doi": "10.1145/3658644.3690284", "citation_count": 42, "influential_citation_count": 7, "has_code": true, "code_url": null, "venue": "Conference on Computer and Communications Security", "quality_score": 0.4515}
{"id": "b1af9f8ce35a17b670d9034b26c3db939551abd6528349d801076dab7a0e5138", "sources": ["arxiv", "semantic_scholar"], "title": "Model Merging in LLMs, MLLMs, and Beyond: Methods, Theories, Applications and Opportunities", "abstract": "Model merging is an efficient empowerment technique in the machine learning community that does not require the collection of raw training data and does not require expensive computation. As model merging becomes increasingly prevalent across various fields, it is crucial to understand the available model merging techniques comprehensively. However, there is a significant gap in the literature regarding a systematic and thorough review of these techniques. This survey provides a comprehensive overview of model merging methods and theories, their applications in various domains and settings, and future research directions. Specifically, we first propose a new taxonomic approach that exhaustively discusses existing model merging methods. Secondly, we discuss the application of model merging techniques in large language models, multimodal large language models, and more than ten machine learning subfields, including continual learning, multi-task learning, few-shot learning, etc. Finally, we highlight the remaining challenges of model merging and discuss future research directions. A comprehensive list of papers about model merging is available at https://github.com/EnnengYang/Awesome-Model-Merging-Methods-Theories-Applications.", "authors": ["Enneng Yang", "Li Shen", "Guibing Guo", "Xingwei Wang", "Xiaochun Cao", "Jie Zhang", "Dacheng Tao"], "categories": ["cs.LG", "cs.AI", "cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-08-14", "url": "https://arxiv.org/abs/2408.07666", "pdf_url": "https://arxiv.org/pdf/2408.07666v5", "arxiv_id": "2408.07666", "doi": "10.1145/3787849", "citation_count": 253, "influential_citation_count": 8, "has_code": true, "code_url": "https://github.com/EnnengYang/Awesome-Model-Merging-Methods-Theories-Applications", "venue": "ACM Computing Surveys", "quality_score": 0.6012}
{"id": "64738f8a9d00506d17c9dc29de1f558899650290f7572fd9ad93aa21b62e6be0", "sources": ["arxiv", "semantic_scholar"], "title": "KIF: Knowledge Identification and Fusion for Language Model Continual Learning", "abstract": "Language model continual learning (CL) has recently attracted significant interest for its ability to adapt large language models (LLMs) to dynamic real-world scenarios without retraining. A major challenge in this domain is catastrophic forgetting, where models lose previously acquired knowledge upon learning new tasks. Existing approaches commonly utilize multiple parameter-efficient fine-tuning (PEFT) blocks to acquire task-specific knowledge, yet these methods are inefficient and fail to leverage potential knowledge transfer across tasks. In this paper, we introduce a novel CL framework for language models, named Knowledge Identification and Fusion (KIF), which boosts knowledge transfer without depending on memory replay. KIF initially segregates the model into 'skill units' based on parameter dependencies, allowing for more precise control. Subsequently, it employs a novel group-wise knowledge identification technique to ascertain the importance distribution of skill units for a new task. By comparing this importance distribution with those from previous tasks, we implement a fine-grained knowledge fusion strategy that retains task-specific knowledge, thereby preventing forgetting, and updates task-shared knowledge, which facilitates bi-directional knowledge transfer. As a result, KIF achieves an optimal balance between retaining prior knowledge and excelling in new tasks. KIF also demonstrates strong generalizability, making it suitable for various base models and adaptable to PEFT methods like LoRA. Furthermore, it offers notable extensibility, supporting enhancements through integration with memory replay techniques. Comprehensive experiments conducted on two CL benchmarks, involving models ranging from 220M to 7B parameters, affirm the effectiveness of KIF and its variants across different settings.", "authors": ["Yujie Feng", "Xu Chu", "Yongxin Xu", "Zexin Lu", "Bo Liu", "Philip S. Yu", "Xiao-Ming Wu"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-08-09", "url": "https://arxiv.org/abs/2408.05200", "pdf_url": "https://arxiv.org/pdf/2408.05200v4", "arxiv_id": "2408.05200", "doi": null, "citation_count": 9, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.25}
{"id": "ab3a85808cda0fdb98a9eab8a92c3ebd4755cc7244a27ecb14ac5f2f25a27212", "sources": ["arxiv", "semantic_scholar"], "title": "Extend Model Merging from Fine-Tuned to Pre-Trained Large Language Models via Weight Disentanglement", "abstract": "Merging Large Language Models (LLMs) aims to amalgamate multiple homologous LLMs into one with all the capabilities. Ideally, any LLMs sharing the same backbone should be mergeable, irrespective of whether they are Fine-Tuned (FT) with minor parameter changes or Pre-Trained (PT) with substantial parameter shifts. However, existing methods often manually assign the model importance, rendering them feasible only for LLMs with similar parameter alterations, such as multiple FT LLMs. The diverse parameter changed ranges between FT and PT LLMs pose challenges for current solutions in empirically determining the optimal combination. In this paper, we make a pioneering effort to broaden the applicability of merging techniques from FT to PT LLMs. We initially examine the efficacy of current methods in merging FT and PT LLMs, discovering that they struggle to deal with PT LLMs. Subsequently, we introduce an approach based on WeIght DisENtanglement (WIDEN) to effectively extend the merging scope, which first disentangles model weights into magnitude and direction components, and then performs adaptive fusion by considering their respective contributions. In the experiments, we merge Qwen1.5-Chat (an FT LLM with instruction-following skills) with Sailor (a PT LLM with multilingual abilities) across 7B and 14B model scales. Results reveal that: (1) existing solutions usually fail when merging Sailor, either losing both abilities or only retaining instruction-following skills; (2) WIDEN successfully injects the multilingual abilities of Sailor into Qwen1.5-Chat and make it proficient in Southeast Asian languages, achieving enhancements in the fundamental capabilities. In light of previous research, we also merge multiple 13B FT LLMs and observe that WIDEN achieves a balanced amalgamation of instruction following, mathematical reasoning, and code generation skills.", "authors": ["Le Yu", "Bowen Yu", "Haiyang Yu", "Fei Huang", "Yongbin Li"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-08-06", "url": "https://arxiv.org/abs/2408.03092", "pdf_url": "https://arxiv.org/pdf/2408.03092v1", "arxiv_id": "2408.03092", "doi": "10.48550/arXiv.2408.03092", "citation_count": 12, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2785}
{"id": "dd06caabbf5230acfec64015b4935066fbee578acc008781b6c01fef46779c49", "sources": ["arxiv", "semantic_scholar"], "title": "Dynamical Degrees, Arithmetic Degrees, and Canonical Heights: History, Conjectures, and Future Directions", "abstract": "In this note we give an overview of various quantities that are used to measure the complexity of an algebraic dynamical system f:X-->X, including the dynamical degree d(f), which gives a coarse measure of the geometric complexity of the iterates of f, the arithmetic degree a(f,P), which gives a coarse measure of the arithmetic complexity of the orbit of a an algebraic point P in X, and various versions of the canonical height h_f(P) that provide more refined measures of arithmetic complexity. Emphasis is placed on open problems and directions for further exploration.", "authors": ["Joseph H. Silverman"], "categories": ["math.NT", "math.DS"], "fields_of_study": ["Mathematics"], "published_date": "2024-08-02", "url": "https://arxiv.org/abs/2408.01559", "pdf_url": "https://arxiv.org/pdf/2408.01559v1", "arxiv_id": "2408.01559", "doi": null, "citation_count": 2, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1505}
{"id": "91540ede9380eeb4ed9cf59bc03226fb8bb9095e537e10d8056bfeb22c1ef523", "sources": ["arxiv", "semantic_scholar"], "title": "Principled Understanding of Generalization for Generative Transformer Models in Arithmetic Reasoning Tasks", "abstract": "Transformer-based models excel in various tasks but their generalization capabilities, especially in arithmetic reasoning, remain incompletely understood. Arithmetic tasks provide a controlled framework to explore these capabilities, yet performance anomalies persist, such as inconsistent effectiveness in multiplication and erratic generalization in modular addition (e.g., modulo 100 vs. 101). This paper develops a unified theoretical framework for understanding the generalization behaviors of transformers in arithmetic tasks, focusing on length generalization. Through detailed analysis of addition, multiplication, and modular operations, we reveal that translation invariance in addition aligns with relative positional encoding for robust generalization, while base mismatch in modular operations disrupts this alignment. Experiments across GPT-family models validate our framework, confirming its ability to predict generalization behaviors. Our work highlights the importance of task structure and training data distribution for achieving data-efficient and structure-aware training, providing a systematic approach to understanding of length generalization in transformers.", "authors": ["Xingcheng Xu", "Zibo Zhao", "Haipeng Zhang", "Yanqing Yang"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-07-25", "url": "https://arxiv.org/abs/2407.17963", "pdf_url": "https://arxiv.org/pdf/2407.17963v2", "arxiv_id": "2407.17963", "doi": "10.18653/v1/2025.acl-long.235", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.1193}
{"id": "e1b38e09517478168ecf7716ece031705bd25d633c35f5a7a2dccea19ac25dc4", "sources": ["arxiv", "semantic_scholar"], "title": "Training Foundation Models as Data Compression: On Information, Model Weights and Copyright Law", "abstract": "The training process of foundation models as for other classes of deep learning systems is based on minimizing the reconstruction error over a training set. For this reason, they are susceptible to the memorization and subsequent reproduction of training samples. In this paper, we introduce a training-as-compressing perspective, wherein the model's weights embody a compressed representation of the training data. From a copyright standpoint, this point of view implies that the weights can be considered a reproduction or, more likely, a derivative work of a potentially protected set of works. We investigate the technical and legal challenges that emerge from this framing of the copyright of outputs generated by foundation models, including their implications for practitioners and researchers. We demonstrate that adopting an information-centric approach to the problem presents a promising pathway for tackling these emerging complex legal issues.", "authors": ["Giorgio Franceschelli", "Claudia Cevenini", "Mirco Musolesi"], "categories": ["cs.CY", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-07-18", "url": "https://arxiv.org/abs/2407.13493", "pdf_url": "https://arxiv.org/pdf/2407.13493v4", "arxiv_id": "2407.13493", "doi": "10.48550/arXiv.2407.13493", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1193}
{"id": "27ff2a8e575fe0cdd4db3cc9aa2634b64ac7d2a91a1a9e536938fea6037b6ecd", "sources": ["arxiv", "semantic_scholar"], "title": "Model Tells You Where to Merge: Adaptive KV Cache Merging for LLMs on Long-Context Tasks", "abstract": "How to efficiently serve Large Language Models (LLMs) has become a pressing issue because of their huge computational cost in their autoregressive generation process. To mitigate computational costs, LLMs often employ the KV Cache technique to improve the generation speed. While improving the computational efficiency, the storage requirements of the KV cache are substantial, particularly in long-context scenarios, leading to significant memory consumption. Existing KV cache eviction methods often degrade the performance of LLMs in long-context scenarios due to the information loss introduced by eviction. In this paper, we propose a novel KV cache merging approach, called KVMerger, to achieve adaptive KV cache compression for long-context tasks without significant performance degradation under constrained memory budgets. Our approach is inspired by the intriguing observation that key states exhibit high similarity at the token level within a single sequence. To facilitate merging, we develop an effective yet straightforward merging set identification algorithm to identify suitable KV states for merging. Our merging set identification algorithm stimulates the second observation that KV cache sparsity, from similarity perspective, is independent of the dataset and remains persistent at the model level. Subsequently, we propose a Gaussian kernel weighted merging algorithm to selectively merge all states within each merging set. We conduct extensive experiments to demonstrate the effectiveness of KVMerger for long-context tasks under constrained memory budgets, applying it to models including Llama2-7B-chat and Llama2-13B-chat. Using the LongBench and ZeroScroll benchmarks, we compare our method with other KV cache compression techniques, including H2O and CaM, showing that our method achieves superior performance across tasks with both 50% and 35% KV cache budgets.", "authors": ["Zheng Wang", "Boxiao Jin", "Zhongzhi Yu", "Minjia Zhang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-07-11", "url": "https://arxiv.org/abs/2407.08454", "pdf_url": "https://arxiv.org/pdf/2407.08454v2", "arxiv_id": "2407.08454", "doi": "10.48550/arXiv.2407.08454", "citation_count": 79, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4758}
{"id": "1072cc5aa57e1d308edc4f735a559b20ac60152dc27ce743a7f2152941526c1e", "sources": ["arxiv", "semantic_scholar"], "title": "Fine-Tuning Attention Modules Only: Enhancing Weight Disentanglement in Task Arithmetic", "abstract": "In recent years, task arithmetic has garnered increasing attention. This approach edits pre-trained models directly in weight space by combining the fine-tuned weights of various tasks into a unified model. Its efficiency and cost-effectiveness stem from its training-free combination, contrasting with traditional methods that require model training on large datasets for multiple tasks. However, applying such a unified model to individual tasks can lead to interference from other tasks (lack of weight disentanglement). To address this issue, Neural Tangent Kernel (NTK) linearization has been employed to leverage a \"kernel behavior\", facilitating weight disentanglement and mitigating adverse effects from unrelated tasks. Despite its benefits, NTK linearization presents drawbacks, including doubled training costs, as well as reduced performance of individual models. To tackle this problem, we propose a simple yet effective and efficient method that is to finetune the attention modules only in the Transformer. Our study reveals that the attention modules exhibit kernel behavior, and fine-tuning the attention modules only significantly improves weight disentanglement. To further understand how our method improves the weight disentanglement of task arithmetic, we present a comprehensive study of task arithmetic by differentiating the role of the representation module and task-specific module. In particular, we find that the representation module plays an important role in improving weight disentanglement whereas the task-specific modules such as the classification heads can degenerate the weight disentanglement performance. (The code is available at https://github.com/kyrie-23/task_arithmetic_tangent)", "authors": ["Ruochen Jin", "Bojian Hou", "Jiancong Xiao", "Weijie Su", "Li Shen"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-07-09", "url": "https://arxiv.org/abs/2407.07089", "pdf_url": "https://arxiv.org/pdf/2407.07089v2", "arxiv_id": "2407.07089", "doi": null, "citation_count": 14, "influential_citation_count": 4, "has_code": true, "code_url": "https://github.com/kyrie-23/task_arithmetic_tangent", "venue": "International Conference on Learning Representations", "quality_score": 0.3495}
{"id": "da16c6515decbd6022f39e31f2f69e21432c3f6c3c2346b0178f53c67e482268", "sources": ["arxiv", "semantic_scholar"], "title": "MagMax: Leveraging Model Merging for Seamless Continual Learning", "abstract": "This paper introduces a continual learning approach named MagMax, which utilizes model merging to enable large pre-trained models to continuously learn from new data without forgetting previously acquired knowledge. Distinct from traditional continual learning methods that aim to reduce forgetting during task training, MagMax combines sequential fine-tuning with a maximum magnitude weight selection for effective knowledge integration across tasks. Our initial contribution is an extensive examination of model merging techniques, revealing that simple approaches like weight averaging and random weight selection surprisingly hold up well in various continual learning contexts. More importantly, we present MagMax, a novel model-merging strategy that enables continual learning of large pre-trained models for successive tasks. Our thorough evaluation demonstrates the superiority of MagMax in various scenarios, including class- and domain-incremental learning settings. The code is available at this URL: https://github.com/danielm1405/magmax.", "authors": ["Daniel Marczak", "Bartłomiej Twardowski", "Tomasz Trzciński", "Sebastian Cygert"], "categories": ["cs.LG", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-07-08", "url": "https://arxiv.org/abs/2407.06322", "pdf_url": "https://arxiv.org/pdf/2407.06322v2", "arxiv_id": "2407.06322", "doi": "10.48550/arXiv.2407.06322", "citation_count": 67, "influential_citation_count": 9, "has_code": true, "code_url": "https://github.com/danielm1405/magmax", "venue": "European Conference on Computer Vision", "quality_score": 0.5}
{"id": "c788f5b51542f4d199eed0cf94afbbc7a382f2916c704b6d0b76f91376a1194e", "sources": ["arxiv", "semantic_scholar"], "title": "Learning Scalable Model Soup on a Single GPU: An Efficient Subspace Training Strategy", "abstract": "Pre-training followed by fine-tuning is widely adopted among practitioners. The performance can be improved by \"model soups\"~\\cite{wortsman2022model} via exploring various hyperparameter configurations.The Learned-Soup, a variant of model soups, significantly improves the performance but suffers from substantial memory and time costs due to the requirements of (i) having to load all fine-tuned models simultaneously, and (ii) a large computational graph encompassing all fine-tuned models. In this paper, we propose Memory Efficient Hyperplane Learned Soup (MEHL-Soup) to tackle this issue by formulating the learned soup as a hyperplane optimization problem and introducing block coordinate gradient descent to learn the mixing coefficients. At each iteration, MEHL-Soup only needs to load a few fine-tuned models and build a computational graph with one combined model. We further extend MEHL-Soup to MEHL-Soup+ in a layer-wise manner. Experimental results on various ViT models and data sets show that MEHL-Soup(+) outperforms Learned-Soup(+) in terms of test accuracy, and also reduces memory usage by more than $13\\times$. Moreover, MEHL-Soup(+) can be run on a single GPU and achieves $9\\times$ speed up in soup construction compared with the Learned-Soup. The code is released at https://github.com/nblt/MEHL-Soup.", "authors": ["Tao Li", "Weisen Jiang", "Fanghui Liu", "Xiaolin Huang", "James T. Kwok"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-07-04", "url": "https://arxiv.org/abs/2407.03641", "pdf_url": "https://arxiv.org/pdf/2407.03641v2", "arxiv_id": "2407.03641", "doi": "10.48550/arXiv.2407.03641", "citation_count": 4, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/nblt/MEHL-Soup", "venue": "European Conference on Computer Vision", "quality_score": 0.1747}
{"id": "8b34bb7bb9e6ed48c5e4c6ba71d9712bb66b8b3d81b4eee6d31b2244e0489d4f", "sources": ["arxiv", "semantic_scholar"], "title": "Unlocking the Potential of Model Merging for Low-Resource Languages", "abstract": "Adapting large language models (LLMs) to new languages typically involves continual pre-training (CT) followed by supervised fine-tuning (SFT). However, this CT-then-SFT approach struggles with limited data in the context of low-resource languages, failing to balance language modeling and task-solving capabilities. We thus propose model merging as an alternative for low-resource languages, combining models with distinct capabilities into a single model without additional training. We use model merging to develop task-solving LLMs for low-resource languages without SFT data in the target languages. Our experiments based on Llama-2-7B demonstrate that model merging effectively endows LLMs for low-resource languages with task-solving abilities, outperforming CT-then-SFT in scenarios with extremely scarce data. Observing performance saturation in model merging with more training tokens, we further analyze the merging process and introduce a slack variable to the model merging algorithm to mitigate the loss of important parameters, thereby enhancing performance. We hope that model merging can benefit more human languages suffering from data scarcity with its higher data efficiency.", "authors": ["Mingxu Tao", "Chen Zhang", "Quzhe Huang", "Tianyao Ma", "Songfang Huang", "Dongyan Zhao", "Yansong Feng"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-07-04", "url": "https://arxiv.org/abs/2407.03994", "pdf_url": "https://arxiv.org/pdf/2407.03994v3", "arxiv_id": "2407.03994", "doi": "10.18653/v1/2024.findings-emnlp.508", "citation_count": 19, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.3253}
{"id": "c1375debacf9afcf81ab2ee8519f1ab8a9d5d53b02ff4a30c31164c4c9f5b7e6", "sources": ["arxiv", "semantic_scholar"], "title": "On the Workflows and Smells of Leaderboard Operations (LBOps): An Exploratory Study of Foundation Model Leaderboards", "abstract": "Foundation models (FM), such as large language models (LLMs), which are large-scale machine learning (ML) models, have demonstrated remarkable adaptability in various downstream software engineering (SE) tasks, such as code completion, code understanding, and software development. As a result, FM leaderboards have become essential tools for SE teams to compare and select the best third-party FMs for their specific products and purposes. However, the lack of standardized guidelines for FM evaluation and comparison threatens the transparency of FM leaderboards and limits stakeholders' ability to perform effective FM selection. As a first step towards addressing this challenge, our research focuses on understanding how these FM leaderboards operate in real-world scenarios (\"leaderboard operations\") and identifying potential pitfalls and areas for improvement (\"leaderboard smells\"). In this regard, we collect up to 1,045 FM leaderboards from five different sources: GitHub, Hugging Face Spaces, Papers With Code, spreadsheet and independent platform, to examine their documentation and engage in direct communication with leaderboard operators to understand their workflows. Through card sorting and negotiated agreement, we identify five distinct workflow patterns and develop a domain model that captures the key components and their interactions within these workflows. We then identify eight unique types of leaderboard smells in LBOps. By mitigating these smells, SE teams can improve transparency, accountability, and collaboration in current LBOps practices, fostering a more robust and responsible ecosystem for FM comparison and selection.", "authors": ["Zhimin Zhao", "Abdul Ali Bangash", "Filipe Roseiro Côgo", "Bram Adams", "Ahmed E. Hassan"], "categories": ["cs.SE", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-07-04", "url": "https://arxiv.org/abs/2407.04065", "pdf_url": "https://arxiv.org/pdf/2407.04065v4", "arxiv_id": "2407.04065", "doi": "10.1109/TSE.2025.3533972", "citation_count": 5, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/SAILResearch/awesome-foundation-model-leaderboards;", "venue": "IEEE Transactions on Software Engineering", "quality_score": 0.1945}
{"id": "587882956ba87c3577e1dc6258013f52fa5659c1482767a30f80c9649cd4f79f", "sources": ["arxiv", "semantic_scholar"], "title": "PLeaS -- Merging Models with Permutations and Least Squares", "abstract": "The democratization of machine learning systems has made the process of fine-tuning accessible to practitioners, leading to a wide range of open-source models fine-tuned on specialized tasks and datasets. Recent work has proposed to merge such models to combine their functionalities. However, prior approaches are usually restricted to models that are fine-tuned from the same base model. Furthermore, the final merged model is typically required to be of the same size as the original models. In this work, we propose a new two-step algorithm to merge models -- termed PLeaS -- which relaxes these constraints. First, leveraging the Permutation symmetries inherent in the two models, PLeaS partially matches nodes in each layer by maximizing alignment. Next, PLeaS computes the weights of the merged model as a layer-wise Least Squares solution to minimize the approximation error between the features of the merged model and the permuted features of the original models. PLeaS allows a practitioner to merge two models sharing the same architecture into a single performant model of a desired size, even when the two original models are fine-tuned from different base models. We also demonstrate how our method can be extended to address a challenging scenario where no data is available from the fine-tuning domains. We demonstrate our method to merge ResNet and ViT models trained with shared and different label spaces, and show improvement over the state-of-the-art merging methods of up to 15 percentage points for the same target compute while merging models trained on DomainNet and fine-grained classification tasks. Our code is open-sourced at https://github.com/SewoongLab/PLeaS-Merging .", "authors": ["Anshul Nasery", "Jonathan Hayase", "Pang Wei Koh", "Sewoong Oh"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-07-02", "url": "https://arxiv.org/abs/2407.02447", "pdf_url": "https://arxiv.org/pdf/2407.02447v2", "arxiv_id": "2407.02447", "doi": "10.1109/CVPR52734.2025.02839", "citation_count": 18, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/SewoongLab/PLeaS-Merging", "venue": "Computer Vision and Pattern Recognition", "quality_score": 0.3197}
{"id": "a51cf2a5e180223bc809e902c949084e72a02828357f3f3956ef7e8d8b6fb74d", "sources": ["arxiv", "semantic_scholar"], "title": "An Attribute Interpolation Method in Speech Synthesis by Model Merging", "abstract": "With the development of speech synthesis, recent research has focused on challenging tasks, such as speaker generation and emotion intensity control. Attribute interpolation is a common approach to these tasks. However, most previous methods for attribute interpolation require specific modules or training methods. We propose an attribute interpolation method in speech synthesis by model merging. Model merging is a method that creates new parameters by only averaging the parameters of base models. The merged model can generate an output with an intermediate feature of the base models. This method is easily applicable without specific modules or training methods, as it uses only existing trained base models. We merged two text-to-speech models to achieve attribute interpolation and evaluated its performance on speaker generation and emotion intensity control tasks. As a result, our proposed method achieved smooth attribute interpolation while keeping the linguistic content in both tasks.", "authors": ["Masato Murata", "Koichi Miyazaki", "Tomoki Koriyama"], "categories": ["cs.SD", "eess.AS"], "fields_of_study": ["Computer Science", "Engineering"], "published_date": "2024-06-30", "url": "https://arxiv.org/abs/2407.00766", "pdf_url": "https://arxiv.org/pdf/2407.00766v1", "arxiv_id": "2407.00766", "doi": "10.48550/arXiv.2407.00766", "citation_count": 9, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Interspeech", "quality_score": 0.25}
{"id": "5aeb5ba4b46d602849facefacefc9fec794eac7c96b493bff7468d6826e0761c", "sources": ["arxiv", "semantic_scholar"], "title": "Model Merging and Safety Alignment: One Bad Model Spoils the Bunch", "abstract": "Merging Large Language Models (LLMs) is a cost-effective technique for combining multiple expert LLMs into a single versatile model, retaining the expertise of the original ones. However, current approaches often overlook the importance of safety alignment during merging, leading to highly misaligned models. This work investigates the effects of model merging on alignment. We evaluate several popular model merging techniques, demonstrating that existing methods do not only transfer domain expertise but also propagate misalignment. We propose a simple two-step approach to address this problem: (i) generating synthetic safety and domain-specific data, and (ii) incorporating these generated data into the optimization process of existing data-aware model merging techniques. This allows us to treat alignment as a skill that can be maximized in the resulting merged LLM. Our experiments illustrate the effectiveness of integrating alignment-related data during merging, resulting in models that excel in both domain expertise and alignment.", "authors": ["Hasan Abed Al Kader Hammoud", "Umberto Michieli", "Fabio Pizzati", "Philip Torr", "Adel Bibi", "Bernard Ghanem", "Mete Ozay"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-06-20", "url": "https://arxiv.org/abs/2406.14563", "pdf_url": "https://arxiv.org/pdf/2406.14563v1", "arxiv_id": "2406.14563", "doi": "10.48550/arXiv.2406.14563", "citation_count": 41, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.4058}
{"id": "b0ff744e2b561fa8a13f7998793624ef93725138e13c7873f8e4aed52cc95990", "sources": ["arxiv", "semantic_scholar"], "title": "MetaGPT: Merging Large Language Models Using Model Exclusive Task Arithmetic", "abstract": "The advent of large language models (LLMs) like GPT-4 has catalyzed the exploration of multi-task learning (MTL), in which a single model demonstrates proficiency across diverse tasks. Task arithmetic has emerged as a cost-effective approach for MTL. It enables performance enhancement across multiple tasks by adding their corresponding task vectors to a pre-trained model. However, the current lack of a method that can simultaneously achieve optimal performance, computational efficiency, and data privacy limits their application to LLMs. In this paper, we propose \\textbf{M}odel \\textbf{E}xclusive \\textbf{T}ask \\textbf{A}rithmetic for merging \\textbf{GPT}-scale models, which formalizes the objective of model merging into a multi-task learning framework, aiming to minimize the average loss difference between the merged model and each individual task model. Since data privacy limits the use of multi-task training data, we leverage LLMs' local linearity and task vectors' orthogonality to separate the data term and scaling coefficients term and derive a model-exclusive task arithmetic method. Our proposed MetaGPT is data-agnostic and bypasses the heavy search process, making it cost-effective and easy to implement for LLMs.Extensive experiments demonstrate that MetaGPT leads to improvements in task arithmetic and achieves state-of-the-art performance on multiple tasks.", "authors": ["Yuyan Zhou", "Liang Song", "Bingning Wang", "Weipeng Chen"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-06-17", "url": "https://arxiv.org/abs/2406.11385", "pdf_url": "https://arxiv.org/pdf/2406.11385v2", "arxiv_id": "2406.11385", "doi": "10.48550/arXiv.2406.11385", "citation_count": 55, "influential_citation_count": 4, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.437}
{"id": "663b3a3bdf2c9af092b864e5e454c7ef63c305c5904f6d31d94f7e7da286ea92", "sources": ["arxiv", "semantic_scholar"], "title": "Safety Arithmetic: A Framework for Test-time Safety Alignment of Language Models by Steering Parameters and Activations", "abstract": "Ensuring the safe alignment of large language models (LLMs) with human values is critical as they become integral to applications like translation and question answering. Current alignment methods struggle with dynamic user intentions and complex objectives, making models vulnerable to generating harmful content. We propose Safety Arithmetic, a training-free framework enhancing LLM safety across different scenarios: Base models, Supervised fine-tuned models (SFT), and Edited models. Safety Arithmetic involves Harm Direction Removal to avoid harmful content and Safety Alignment to promote safe responses. Additionally, we present NoIntentEdit, a dataset highlighting edit instances that could compromise model safety if used unintentionally. Our experiments show that Safety Arithmetic significantly improves safety measures, reduces over-safety, and maintains model utility, outperforming existing methods in ensuring safe content generation.", "authors": ["Rima Hazra", "Sayan Layek", "Somnath Banerjee", "Soujanya Poria"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-06-17", "url": "https://arxiv.org/abs/2406.11801", "pdf_url": "https://arxiv.org/pdf/2406.11801v2", "arxiv_id": "2406.11801", "doi": "10.48550/arXiv.2406.11801", "citation_count": 29, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/declare-lab/safety-arithmetic", "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.3693}
{"id": "ca04cc046fc3016e24da77bc876466175b0ea20fba6a921b34818fe496209135", "sources": ["arxiv", "semantic_scholar"], "title": "Twin-Merging: Dynamic Integration of Modular Expertise in Model Merging", "abstract": "In the era of large language models, model merging is a promising way to combine multiple task-specific models into a single multitask model without extra training. However, two challenges remain: (a) interference between different models and (b) heterogeneous data during testing. Traditional model merging methods often show significant performance gaps compared to fine-tuned models due to these issues. Additionally, a one-size-fits-all model lacks flexibility for diverse test data, leading to performance degradation. We show that both shared and exclusive task-specific knowledge are crucial for merging performance, but directly merging exclusive knowledge hinders overall performance. In view of this, we propose Twin-Merging, a method that encompasses two principal stages: (1) modularizing knowledge into shared and exclusive components, with compression to reduce redundancy and enhance efficiency; (2) dynamically merging shared and task-specific knowledge based on the input. This approach narrows the performance gap between merged and fine-tuned models and improves adaptability to heterogeneous data. Extensive experiments on $20$ datasets for both language and vision tasks demonstrate the effectiveness of our method, showing an average improvement of $28.34\\%$ in absolute normalized score for discriminative tasks and even surpassing the fine-tuned upper bound on the generative tasks. Our implementation is available in \\url{https://github.com/LZY-the-boys/Twin-Merging}", "authors": ["Zhenyi Lu", "Chenghao Fan", "Wei Wei", "Xiaoye Qu", "Dangyang Chen", "Yu Cheng"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-06-17", "url": "https://arxiv.org/abs/2406.15479", "pdf_url": "https://arxiv.org/pdf/2406.15479v2", "arxiv_id": "2406.15479", "doi": "10.48550/arXiv.2406.15479", "citation_count": 121, "influential_citation_count": 23, "has_code": true, "code_url": "https://github.com/LZY-the-boys/Twin-Merging}", "venue": "Neural Information Processing Systems", "quality_score": 0.6901}
{"id": "803bd660373d2ef31e056b94ea76004bddf11fbf67e4ba2ebf0dbef515b8e317", "sources": ["arxiv", "semantic_scholar"], "title": "Performance Improvement of Language-Queried Audio Source Separation Based on Caption Augmentation From Large Language Models for DCASE Challenge 2024 Task 9", "abstract": "We present a prompt-engineering-based text-augmentation approach applied to a language-queried audio source separation (LASS) task. To enhance the performance of LASS, the proposed approach utilizes large language models (LLMs) to generate multiple captions corresponding to each sentence of the training dataset. To this end, we first perform experiments to identify the most effective prompts for caption augmentation with a smaller number of captions. A LASS model trained with these augmented captions demonstrates improved performance on the DCASE 2024 Task 9 validation set compared to that trained without augmentation. This study highlights the effectiveness of LLM-based caption augmentation in advancing language-queried audio source separation.", "authors": ["Do Hyun Lee", "Yoonah Song", "Hong Kook Kim"], "categories": ["eess.AS", "cs.AI", "cs.SD"], "fields_of_study": ["Engineering", "Computer Science"], "published_date": "2024-06-17", "url": "https://arxiv.org/abs/2406.11248", "pdf_url": "https://arxiv.org/pdf/2406.11248v2", "arxiv_id": "2406.11248", "doi": "10.48550/arXiv.2406.11248", "citation_count": 5, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.301}
{"id": "a0b889915e59aa267fe45468e946ca116b286d9eacad0de08906231fdcb12d78", "sources": ["arxiv", "semantic_scholar"], "title": "Diffusion Soup: Model Merging for Text-to-Image Diffusion Models", "abstract": "We present Diffusion Soup, a compartmentalization method for Text-to-Image Generation that averages the weights of diffusion models trained on sharded data. By construction, our approach enables training-free continual learning and unlearning with no additional memory or inference costs, since models corresponding to data shards can be added or removed by re-averaging. We show that Diffusion Soup samples from a point in weight space that approximates the geometric mean of the distributions of constituent datasets, which offers anti-memorization guarantees and enables zero-shot style mixing. Empirically, Diffusion Soup outperforms a paragon model trained on the union of all data shards and achieves a 30% improvement in Image Reward (.34 $\\to$ .44) on domain sharded data, and a 59% improvement in IR (.37 $\\to$ .59) on aesthetic data. In both cases, souping also prevails in TIFA score (respectively, 85.5 $\\to$ 86.5 and 85.6 $\\to$ 86.8). We demonstrate robust unlearning -- removing any individual domain shard only lowers performance by 1% in IR (.45 $\\to$ .44) -- and validate our theoretical insights on anti-memorization using real data. Finally, we showcase Diffusion Soup's ability to blend the distinct styles of models finetuned on different shards, resulting in the zero-shot generation of hybrid styles.", "authors": ["Benjamin Biggs", "Arjun Seshadri", "Yang Zou", "Achin Jain", "Aditya Golatkar", "Yusheng Xie", "Alessandro Achille", "Ashwin Swaminathan", "Stefano Soatto"], "categories": ["cs.CV", "cs.AI", "cs.CR", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-06-12", "url": "https://arxiv.org/abs/2406.08431", "pdf_url": "https://arxiv.org/pdf/2406.08431v1", "arxiv_id": "2406.08431", "doi": "10.48550/arXiv.2406.08431", "citation_count": 27, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "European Conference on Computer Vision", "quality_score": 0.3618}
{"id": "388692881245549a0319c1050d7404d3807ccb4ad4a854c623e99ee382cee642", "sources": ["arxiv", "semantic_scholar"], "title": "Language Models Do Hard Arithmetic Tasks Easily and Hardly Do Easy Arithmetic Tasks", "abstract": "The ability (and inability) of large language models (LLMs) to perform arithmetic tasks has been the subject of much theoretical and practical debate. We show that LLMs are frequently able to correctly and confidently predict the first digit of n-digit by m-digit multiplication tasks without using chain of thought reasoning, despite these tasks require compounding operations to solve. Simultaneously, LLMs in practice often fail to correctly or confidently predict the last digit of an n-digit by m-digit multiplication, a task equivalent to 1-digit by 1-digit multiplication which can be easily learned or memorized. We show that the latter task can be solved more robustly when the LLM is conditioned on all of the correct higher-order digits, which on average increases the confidence of the correct last digit on 5-digit by 5-digit multiplication tasks using Llama 2-13B by over 230% (0.13 to 0.43) and Mistral-7B by 150% (0.22 to 0.55).", "authors": ["Andrew Gambardella", "Yusuke Iwasawa", "Yutaka Matsuo"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-06-04", "url": "https://arxiv.org/abs/2406.02356", "pdf_url": "https://arxiv.org/pdf/2406.02356v1", "arxiv_id": "2406.02356", "doi": "10.48550/arXiv.2406.02356", "citation_count": 24, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.3495}
{"id": "787161f5f32bc6ba5a64b73a21464b7a9fd4887e9b69ddfc8f9a3dcdff30ce62", "sources": ["arxiv", "semantic_scholar"], "title": "EMR-Merging: Tuning-Free High-Performance Model Merging", "abstract": "The success of pretrain-finetune paradigm brings about the release of numerous model weights. In this case, merging models finetuned on different tasks to enable a single model with multi-task capabilities is gaining increasing attention for its practicability. Existing model merging methods usually suffer from (1) significant performance degradation or (2) requiring tuning by additional data or training. In this paper, we rethink and analyze the existing model merging paradigm. We discover that using a single model's weights can hardly simulate all the models' performance. To tackle this issue, we propose Elect, Mask & Rescale-Merging (EMR-Merging). We first (a) elect a unified model from all the model weights and then (b) generate extremely lightweight task-specific modulators, including masks and rescalers, to align the direction and magnitude between the unified model and each specific model, respectively. EMR-Merging is tuning-free, thus requiring no data availability or any additional training while showing impressive performance. We find that EMR-Merging shows outstanding performance compared to existing merging methods under different classical and newly-established settings, including merging different numbers of vision models (up to 30), NLP models, PEFT models, and multi-modal models.", "authors": ["Chenyu Huang", "Peng Ye", "Tao Chen", "Tong He", "Xiangyu Yue", "Wanli Ouyang"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-05-23", "url": "https://arxiv.org/abs/2405.17461", "pdf_url": "https://arxiv.org/pdf/2405.17461v2", "arxiv_id": "2405.17461", "doi": "10.48550/arXiv.2405.17461", "citation_count": 112, "influential_citation_count": 21, "has_code": false, "code_url": null, "venue": "Neural Information Processing Systems", "quality_score": 0.6712}
{"id": "0be567ad738c8e26c2f9b02f7ac838f8c10312fb3b7fe28bbf47899e4eaf6476", "sources": ["arxiv", "semantic_scholar"], "title": "Octo: An Open-Source Generalist Robot Policy", "abstract": "Large policies pretrained on diverse robot datasets have the potential to transform robotic learning: instead of training new policies from scratch, such generalist robot policies may be finetuned with only a little in-domain data, yet generalize broadly. However, to be widely applicable across a range of robotic learning scenarios, environments, and tasks, such policies need to handle diverse sensors and action spaces, accommodate a variety of commonly used robotic platforms, and finetune readily and efficiently to new domains. In this work, we aim to lay the groundwork for developing open-source, widely applicable, generalist policies for robotic manipulation. As a first step, we introduce Octo, a large transformer-based policy trained on 800k trajectories from the Open X-Embodiment dataset, the largest robot manipulation dataset to date. It can be instructed via language commands or goal images and can be effectively finetuned to robot setups with new sensory inputs and action spaces within a few hours on standard consumer GPUs. In experiments across 9 robotic platforms, we demonstrate that Octo serves as a versatile policy initialization that can be effectively finetuned to new observation and action spaces. We also perform detailed ablations of design decisions for the Octo model, from architecture to training data, to guide future research on building generalist robot models.", "authors": [" Octo Model Team", "Dibya Ghosh", "Homer Walke", "Karl Pertsch", "Kevin Black", "Oier Mees", "Sudeep Dasari", "Joey Hejna", "Tobias Kreiman", "Charles Xu", "Jianlan Luo", "You Liang Tan", "Lawrence Yunliang Chen", "Pannag Sanketi", "Quan Vuong", "Ted Xiao", "Dorsa Sadigh", "Chelsea Finn", "Sergey Levine"], "categories": ["cs.RO", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-05-20", "url": "https://arxiv.org/abs/2405.12213", "pdf_url": "https://arxiv.org/pdf/2405.12213v2", "arxiv_id": "2405.12213", "doi": "10.48550/arXiv.2405.12213", "citation_count": 1386, "influential_citation_count": 112, "has_code": true, "code_url": null, "venue": null, "quality_score": 1.0}
{"id": "636285c8a1bbc80db69644b73c0a345703c3e1bea06177ba15a2cfcb0d457987", "sources": ["arxiv", "semantic_scholar"], "title": "Localizing Task Information for Improved Model Merging and Compression", "abstract": "Model merging and task arithmetic have emerged as promising scalable approaches to merge multiple single-task checkpoints to one multi-task model, but their applicability is reduced by significant performance loss. Previous works have linked these drops to interference in the weight space and erasure of important task-specific features. Instead, in this work we show that the information required to solve each task is still preserved after merging as different tasks mostly use non-overlapping sets of weights. We propose TALL-masks, a method to identify these task supports given a collection of task vectors and show that one can retrieve >99% of the single task accuracy by applying our masks to the multi-task vector, effectively compressing the individual checkpoints. We study the statistics of intersections among constructed masks and reveal the existence of selfish and catastrophic weights, i.e., parameters that are important exclusively to one task and irrelevant to all tasks but detrimental to multi-task fusion. For this reason, we propose Consensus Merging, an algorithm that eliminates such weights and improves the general performance of existing model merging approaches. Our experiments in vision and NLP benchmarks with up to 20 tasks, show that Consensus Merging consistently improves existing approaches. Furthermore, our proposed compression scheme reduces storage from 57Gb to 8.2Gb while retaining 99.7% of original performance.", "authors": ["Ke Wang", "Nikolaos Dimitriadis", "Guillermo Ortiz-Jimenez", "François Fleuret", "Pascal Frossard"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-05-13", "url": "https://arxiv.org/abs/2405.07813", "pdf_url": "https://arxiv.org/pdf/2405.07813v1", "arxiv_id": "2405.07813", "doi": "10.48550/arXiv.2405.07813", "citation_count": 115, "influential_citation_count": 21, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.6712}
{"id": "bfd623b425c00dbee51e6e151042612cc13a44beafdc202984bd5e268770c649", "sources": ["arxiv", "semantic_scholar"], "title": "D-NLP at SemEval-2024 Task 2: Evaluating Clinical Inference Capabilities of Large Language Models", "abstract": "Large language models (LLMs) have garnered significant attention and widespread usage due to their impressive performance in various tasks. However, they are not without their own set of challenges, including issues such as hallucinations, factual inconsistencies, and limitations in numerical-quantitative reasoning. Evaluating LLMs in miscellaneous reasoning tasks remains an active area of research. Prior to the breakthrough of LLMs, Transformers had already proven successful in the medical domain, effectively employed for various natural language understanding (NLU) tasks. Following this trend, LLMs have also been trained and utilized in the medical domain, raising concerns regarding factual accuracy, adherence to safety protocols, and inherent limitations. In this paper, we focus on evaluating the natural language inference capabilities of popular open-source and closed-source LLMs using clinical trial reports as the dataset. We present the performance results of each LLM and further analyze their performance on a development set, particularly focusing on challenging instances that involve medical abbreviations and require numerical-quantitative reasoning. Gemini, our leading LLM, achieved a test set F1-score of 0.748, securing the ninth position on the task scoreboard. Our work is the first of its kind, offering a thorough examination of the inference capabilities of LLMs within the medical domain.", "authors": ["Duygu Altinok"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-05-07", "url": "https://arxiv.org/abs/2405.04170", "pdf_url": "https://arxiv.org/pdf/2405.04170v1", "arxiv_id": "2405.04170", "doi": "10.48550/arXiv.2405.04170", "citation_count": 4, "influential_citation_count": 1, "has_code": true, "code_url": null, "venue": "International Workshop on Semantic Evaluation", "quality_score": 0.1747}
{"id": "f52e3ce9787f706639c634424e7a3255a6701978a78d05afd218b8f960a6902b", "sources": ["arxiv", "semantic_scholar"], "title": "Beating Posits at Their Own Game: Takum Arithmetic", "abstract": "Recent evaluations have highlighted the tapered posit number format as a promising alternative to the uniform precision IEEE 754 floating-point numbers, which suffer from various deficiencies. Although the posit encoding scheme offers superior coding efficiency at values close to unity, its efficiency markedly diminishes with deviation from unity. This reduction in efficiency leads to suboptimal encodings and a consequent diminution in dynamic range, thereby rendering posits suboptimal for general-purpose computer arithmetic. This paper introduces and formally proves 'takum' as a novel general-purpose logarithmic tapered-precision number format, synthesising the advantages of posits in low-bit applications with high encoding efficiency for numbers distant from unity. Takums exhibit an asymptotically constant dynamic range in terms of bit string length, which is delineated in the paper to be suitable for a general-purpose number format. It is demonstrated that takums either match or surpass existing alternatives. Moreover, takums address several issues previously identified in posits while unveiling novel and beneficial arithmetic properties.", "authors": ["Laslo Hunhold"], "categories": ["math.NA", "cs.DS"], "fields_of_study": ["Mathematics", "Computer Science"], "published_date": "2024-04-29", "url": "https://arxiv.org/abs/2404.18603", "pdf_url": "https://arxiv.org/pdf/2404.18603v2", "arxiv_id": "2404.18603", "doi": "10.1007/978-3-031-72709-2_1", "citation_count": 11, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference for Next Generation Arithmetic", "quality_score": 0.2698}
{"id": "0b1b5f1fdb9edbd2c43e700698b1cb25791270793d3215ffdaeb07f58574f7e6", "sources": ["arxiv", "semantic_scholar"], "title": "MetaCheckGPT -- A Multi-task Hallucination Detector Using LLM Uncertainty and Meta-models", "abstract": "Hallucinations in large language models (LLMs) have recently become a significant problem. A recent effort in this direction is a shared task at Semeval 2024 Task 6, SHROOM, a Shared-task on Hallucinations and Related Observable Overgeneration Mistakes. This paper describes our winning solution ranked 1st and 2nd in the 2 sub-tasks of model agnostic and model aware tracks respectively. We propose a meta-regressor framework of LLMs for model evaluation and integration that achieves the highest scores on the leaderboard. We also experiment with various transformer-based models and black box methods like ChatGPT, Vectara, and others. In addition, we perform an error analysis comparing GPT4 against our best model which shows the limitations of the former.", "authors": ["Rahul Mehta", "Andrew Hoblitzell", "Jack O'Keefe", "Hyeju Jang", "Vasudeva Varma"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-04-10", "url": "https://arxiv.org/abs/2404.06948", "pdf_url": "https://arxiv.org/pdf/2404.06948v2", "arxiv_id": "2404.06948", "doi": "10.48550/arXiv.2404.06948", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "9a91cb93fa6c2647f13d4defbe371745635a57d8008c9891e46640e1de775cb7", "sources": ["arxiv", "semantic_scholar"], "title": "Have You Merged My Model? On The Robustness of Large Language Model IP Protection Methods Against Model Merging", "abstract": "Model merging is a promising lightweight model empowerment technique that does not rely on expensive computing devices (e.g., GPUs) or require the collection of specific training data. Instead, it involves editing different upstream model parameters to absorb their downstream task capabilities. However, uncertified model merging can infringe upon the Intellectual Property (IP) rights of the original upstream models. In this paper, we conduct the first study on the robustness of IP protection methods under model merging scenarios. Specifically, we investigate two state-of-the-art IP protection techniques: Quantization Watermarking and Instructional Fingerprint, along with various advanced model merging technologies, such as Task Arithmetic, TIES-MERGING, and so on. Experimental results indicate that current Large Language Model (LLM) watermarking techniques cannot survive in the merged models, whereas model fingerprinting techniques can. Our research aims to highlight that model merging should be an indispensable consideration in the robustness assessment of model IP protection techniques, thereby promoting the healthy development of the open-source LLM community. Our code is available at https://github.com/ThuCCSLab/MergeGuard.", "authors": ["Tianshuo Cong", "Delong Ran", "Zesen Liu", "Xinlei He", "Jinyuan Liu", "Yichen Gong", "Qi Li", "Anyu Wang", "Xiaoyun Wang"], "categories": ["cs.CR", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-04-08", "url": "https://arxiv.org/abs/2404.05188", "pdf_url": "https://arxiv.org/pdf/2404.05188v2", "arxiv_id": "2404.05188", "doi": "10.1145/3689217.3690614", "citation_count": 26, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/ThuCCSLab/MergeGuard", "venue": null, "quality_score": 0.3578}
{"id": "1469cdc9c9095736f4356cb25e3de34f993edbbc2cfcf243ac0ec6d0cfd61824", "sources": ["arxiv", "semantic_scholar"], "title": "Continual Learning with Weight Interpolation", "abstract": "Continual learning poses a fundamental challenge for modern machine learning systems, requiring models to adapt to new tasks while retaining knowledge from previous ones. Addressing this challenge necessitates the development of efficient algorithms capable of learning from data streams and accumulating knowledge over time. This paper proposes a novel approach to continual learning utilizing the weight consolidation method. Our method, a simple yet powerful technique, enhances robustness against catastrophic forgetting by interpolating between old and new model weights after each novel task, effectively merging two models to facilitate exploration of local minima emerging after arrival of new concepts. Moreover, we demonstrate that our approach can complement existing rehearsal-based replay approaches, improving their accuracy and further mitigating the forgetting phenomenon. Additionally, our method provides an intuitive mechanism for controlling the stability-plasticity trade-off. Experimental results showcase the significant performance enhancement to state-of-the-art experience replay algorithms the proposed weight consolidation approach offers. Our algorithm can be downloaded from https://github.com/jedrzejkozal/weight-interpolation-cl.", "authors": ["Jędrzej Kozal", "Jan Wasilewski", "Bartosz Krawczyk", "Michał Woźniak"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-04-05", "url": "https://arxiv.org/abs/2404.04002", "pdf_url": "https://arxiv.org/pdf/2404.04002v2", "arxiv_id": "2404.04002", "doi": "10.1109/CVPRW63382.2024.00422", "citation_count": 15, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/jedrzejkozal/weight-interpolation-cl", "venue": null, "quality_score": 0.301}
{"id": "4857998265fa2b359e0905a90ce234ebee48c7f220f8a631e732aa89096a18e8", "sources": ["arxiv", "semantic_scholar"], "title": "Classification properties for some ternary structures", "abstract": "We provide a model-theoretic classification of the countable homogeneous $\\mathbf{H}_4$-free 3-hypertournament studied by Cherlin, Hubička, Konečný, and Nešetřil. Our main result is that the theory of this structure is $\\mathrm{SOP}_3$, $\\mathrm{TP}_2$, and $\\mathrm{NSOP}_4$. We offer two proofs of this fact: one is a direct proof, and the other employs part of the abstract machinery recently developed by Mutchnik.", "authors": ["Alberto Miguel-Gómez"], "categories": ["math.LO", "math.CO"], "fields_of_study": ["Mathematics"], "published_date": "2024-04-05", "url": "https://arxiv.org/abs/2404.04381", "pdf_url": "https://arxiv.org/pdf/2404.04381v2", "arxiv_id": "2404.04381", "doi": "10.2140/mt.2025.4.203", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Model Th. 4 (2025) 203-230", "quality_score": 0.1193}
{"id": "ac5faef1e81fb6cdd8a38d57fd1377aed25aefb0fdd6de4d2f17374c48676fe3", "sources": ["arxiv", "semantic_scholar"], "title": "Model Stock: All we need is just a few fine-tuned models", "abstract": "This paper introduces an efficient fine-tuning method for large pre-trained models, offering strong in-distribution (ID) and out-of-distribution (OOD) performance. Breaking away from traditional practices that need a multitude of fine-tuned models for averaging, our approach employs significantly fewer models to achieve final weights yet yield superior accuracy. Drawing from key insights in the weight space of fine-tuned weights, we uncover a strong link between the performance and proximity to the center of weight space. Based on this, we introduce a method that approximates a center-close weight using only two fine-tuned models, applicable during or after training. Our innovative layer-wise weight averaging technique surpasses state-of-the-art model methods such as Model Soup, utilizing only two fine-tuned models. This strategy can be aptly coined Model Stock, highlighting its reliance on selecting a minimal number of models to draw a more optimized-averaged model. We demonstrate the efficacy of Model Stock with fine-tuned models based upon pre-trained CLIP architectures, achieving remarkable performance on both ID and OOD tasks on the standard benchmarks, all while barely bringing extra computational demands. Our code and pre-trained models are available at https://github.com/naver-ai/model-stock.", "authors": ["Dong-Hwan Jang", "Sangdoo Yun", "Dongyoon Han"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-03-28", "url": "https://arxiv.org/abs/2403.19522", "pdf_url": "https://arxiv.org/pdf/2403.19522v2", "arxiv_id": "2403.19522", "doi": "10.48550/arXiv.2403.19522", "citation_count": 91, "influential_citation_count": 15, "has_code": true, "code_url": "https://github.com/naver-ai/model-stock", "venue": "European Conference on Computer Vision", "quality_score": 0.6021}
{"id": "441520eb5673f885666fe05b1dbb703a5b9bb0b60817a445d9914c634465f209", "sources": ["arxiv", "semantic_scholar"], "title": "Vi-Mistral-X: Building a Vietnamese Language Model with Advanced Continual Pre-training", "abstract": "The advancement of Large Language Models (LLMs) has significantly transformed the field of natural language processing, although the focus on English-centric models has created a noticeable research gap for specific languages, including Vietnamese. To address this issue, this paper presents vi-mistral-x, an innovative Large Language Model designed expressly for the Vietnamese language. It utilizes a unique method of continual pre-training, based on the Mistral architecture, which incorporates grouped-query attention and sliding window attention techniques. This model, vi-Mistral-X, marks a significant step forward in improving the understanding and generation of the Vietnamese language. It introduces an additional phase of continual pre-training, specifically adapted for Vietnamese, enhancing the model's capability in understanding complex language nuances and generating accurate, context-aware Vietnamese text. Through comprehensive testing on various benchmarks, vi-mistral-x has shown to outperform existing Vietnamese LLMs in several key areas, including text classification, question answering, and text generation. Particularly, in the Vietnamese Multitask Language Understanding (VMLU) benchmark, vi-mistral-x sets a new standard, outperforming other available models significantly. This paper highlights the critical role of continual pre-training in advancing language-specific LLMs and opens new avenues for the development of multilingual models. We aim for vi-mistral-x to not just be an important asset for processing the Vietnamese language but also to encourage more advancements in creating large language models for languages that are less represented.", "authors": ["James Vo"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-03-20", "url": "https://arxiv.org/abs/2403.15470", "pdf_url": "https://arxiv.org/pdf/2403.15470v1", "arxiv_id": "2403.15470", "doi": "10.48550/arXiv.2403.15470", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "fcdd2b3c5b1a5f70733ba404bbc4e018d67ea83819e1928b34937c142d32f565", "sources": ["arxiv", "semantic_scholar"], "title": "Arcee's MergeKit: A Toolkit for Merging Large Language Models", "abstract": "The rapid expansion of the open-source language model landscape presents an opportunity to merge the competencies of these model checkpoints by combining their parameters. Advances in transfer learning, the process of fine-tuning pretrained models for specific tasks, has resulted in the development of vast amounts of task-specific models, typically specialized in individual tasks and unable to utilize each other's strengths. Model merging facilitates the creation of multitask models without the need for additional training, offering a promising avenue for enhancing model performance and versatility. By preserving the intrinsic capabilities of the original models, model merging addresses complex challenges in AI - including the difficulties of catastrophic forgetting and multitask learning. To support this expanding area of research, we introduce MergeKit, a comprehensive, open-source library designed to facilitate the application of model merging strategies. MergeKit offers an extensible framework to efficiently merge models on any hardware, providing utility to researchers and practitioners. To date, thousands of models have been merged by the open-source community, leading to the creation of some of the worlds most powerful open-source model checkpoints, as assessed by the Open LLM Leaderboard. The library is accessible at https://github.com/arcee-ai/MergeKit.", "authors": ["Charles Goddard", "Shamane Siriwardhana", "Malikeh Ehghaghi", "Luke Meyers", "Vlad Karpukhin", "Brian Benedict", "Mark McQuade", "Jacob Solawetz"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-03-20", "url": "https://arxiv.org/abs/2403.13257", "pdf_url": "https://arxiv.org/pdf/2403.13257v3", "arxiv_id": "2403.13257", "doi": "10.48550/arXiv.2403.13257", "citation_count": 217, "influential_citation_count": 22, "has_code": true, "code_url": "https://github.com/arcee-ai/MergeKit", "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.6809}
{"id": "b36056ccece100cdbfa0322bb5799f56248bfa4dba061412b9dfabd4a287a485", "sources": ["arxiv", "semantic_scholar"], "title": "MedMerge: Merging Models for Effective Transfer Learning to Medical Imaging Tasks", "abstract": "Transfer learning has become a powerful tool to initialize deep learning models to achieve faster convergence and higher performance. This is especially useful in the medical imaging analysis domain, where data scarcity limits possible performance gains for deep learning models. Some advancements have been made in boosting the transfer learning performance gain by merging models starting from the same initialization. However, in the medical imaging analysis domain, there is an opportunity to merge models starting from different initializations, thus combining the features learned from different tasks. In this work, we propose MedMerge, a method whereby the weights of different models can be merged, and their features can be effectively utilized to boost performance on a new task. With MedMerge, we learn kernel-level weights that can later be used to merge the models into a single model, even when starting from different initializations. Testing on various medical imaging analysis tasks, we show that our merged model can achieve significant performance gains, with up to 7% improvement on the F1 score. The code implementation of this work is available at github.com/BioMedIA-MBZUAI/MedMerge.", "authors": ["Ibrahim Almakky", "Santosh Sanjeev", "Anees Ur Rehman Hashmi", "Mohammad Areeb Qazi", "Hu Wang", "Mohammad Yaqub"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-03-18", "url": "https://arxiv.org/abs/2403.11646", "pdf_url": "https://arxiv.org/pdf/2403.11646v2", "arxiv_id": "2403.11646", "doi": "10.48550/arXiv.2403.11646", "citation_count": 8, "influential_citation_count": 1, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2386}
{"id": "d64c937bf0c435ce13ae3d20f9f8bfe8db0a251fd0f127df005991e7828d36bf", "sources": ["arxiv", "semantic_scholar"], "title": "Thermal Earth Model for the Conterminous United States Using an Interpolative Physics-Informed Graph Neural Network (InterPIGNN)", "abstract": "This study presents a data-driven spatial interpolation algorithm based on physics-informed graph neural networks used to develop national temperature-at-depth maps for the conterminous United States. The model was trained to approximately satisfy the three-dimensional heat conduction law by simultaneously predicting subsurface temperature, surface heat flow, and rock thermal conductivity. In addition to bottomhole temperature measurements, we incorporated other physical quantities as model inputs, such as depth, geographic coordinates, elevation, sediment thickness, magnetic anomaly, gravity anomaly, gamma-ray flux of radioactive elements, seismicity, and electric conductivity. We constructed surface heat flow, and temperature and thermal conductivity predictions for depths of 0-7 km at an interval of 1 km with spatial resolution of 18 km$^2$ per grid cell. Our model showed superior temperature, surface heat flow and thermal conductivity mean absolute errors of 4.8° C, 5.817 mW/m$^2$ and 0.022 W/(C-m)$, respectively. The predictions were visualized in two-dimensional spatial maps across the modeled depths. This thorough modeling of the Earth's thermal processes is crucial to understanding subsurface phenomena and exploiting natural underground resources.", "authors": ["Mohammad J. Aljubran", "Roland N. Horne"], "categories": ["physics.geo-ph", "cs.LG"], "fields_of_study": ["Physics", "Computer Science"], "published_date": "2024-03-15", "url": "https://arxiv.org/abs/2403.09961", "pdf_url": "https://arxiv.org/pdf/2403.09961v1", "arxiv_id": "2403.09961", "doi": "10.1186/s40517-024-00304-7", "citation_count": 21, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "Geothermal Energy", "quality_score": 0.3356}
{"id": "17788e19a5a78b04f17432163734845be109d6162adfdf0b579fcd4d610c1a95", "sources": ["arxiv", "semantic_scholar"], "title": "Fisher Mask Nodes for Language Model Merging", "abstract": "Fine-tuning pre-trained models provides significant advantages in downstream performance. The ubiquitous nature of pre-trained models such as BERT and its derivatives in natural language processing has also led to a proliferation of task-specific fine-tuned models. As these models typically only perform one task well, additional training or ensembling is required in multi-task scenarios. The growing field of model merging provides a solution, dealing with the challenge of combining multiple task-specific models into a single multi-task model. In this study, we introduce a novel model merging method for Transformers, combining insights from previous work in Fisher-weighted averaging and the use of Fisher information in model pruning. Utilizing the Fisher information of mask nodes within the Transformer architecture, we devise a computationally efficient weighted-averaging scheme. Our method exhibits a regular and significant performance increase across various models in the BERT family, outperforming full-scale Fisher-weighted averaging in a fraction of the computational cost, with baseline performance improvements of up to +6.5 and a speedup between 57.4x and 321.7x across models. Our results prove the potential of our method in current multi-task learning environments and suggest its scalability and adaptability to new model architectures and learning scenarios.", "authors": ["Thennal D K", "Ganesh Nathan", "Suchithra M S"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-03-14", "url": "https://arxiv.org/abs/2403.09891", "pdf_url": "https://arxiv.org/pdf/2403.09891v3", "arxiv_id": "2403.09891", "doi": "10.48550/arXiv.2403.09891", "citation_count": 9, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Conference on Language Resources and Evaluation", "quality_score": 0.25}
{"id": "e12395dc31065654c16f4d000b45fe690fcf31589b8facdc5e5df13f6b6ce5ea", "sources": ["arxiv", "semantic_scholar"], "title": "SemEval-2024 Shared Task 6: SHROOM, a Shared-task on Hallucinations and Related Observable Overgeneration Mistakes", "abstract": "This paper presents the results of the SHROOM, a shared task focused on detecting hallucinations: outputs from natural language generation (NLG) systems that are fluent, yet inaccurate. Such cases of overgeneration put in jeopardy many NLG applications, where correctness is often mission-critical. The shared task was conducted with a newly constructed dataset of 4000 model outputs labeled by 5 annotators each, spanning 3 NLP tasks: machine translation, paraphrase generation and definition modeling. The shared task was tackled by a total of 58 different users grouped in 42 teams, out of which 27 elected to write a system description paper; collectively, they submitted over 300 prediction sets on both tracks of the shared task. We observe a number of key trends in how this approach was tackled -- many participants rely on a handful of model, and often rely either on synthetic data for fine-tuning or zero-shot prompting strategies. While a majority of the teams did outperform our proposed baseline system, the performances of top-scoring systems are still consistent with a random handling of the more challenging items.", "authors": ["Timothee Mickus", "Elaine Zosa", "Raúl Vázquez", "Teemu Vahtola", "Jörg Tiedemann", "Vincent Segonne", "Alessandro Raganato", "Marianna Apidianaki"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-03-12", "url": "https://arxiv.org/abs/2403.07726", "pdf_url": "https://arxiv.org/pdf/2403.07726v3", "arxiv_id": "2403.07726", "doi": "10.48550/arXiv.2403.07726", "citation_count": 42, "influential_citation_count": 6, "has_code": false, "code_url": null, "venue": "International Workshop on Semantic Evaluation", "quality_score": 0.4225}
{"id": "5716178a6ce3a64b49b758de7ebe44d8aa43a51bebbb3ab5aee6710e73c37e47", "sources": ["arxiv", "semantic_scholar"], "title": "A Segmentation Foundation Model for Diverse-type Tumors", "abstract": "Large pre-trained models with their numerous model parameters and extensive training datasets have shown excellent performance in various tasks. Many publicly available medical image datasets do not have a sufficient amount of data so there are few large-scale models in medical imaging. We propose a large-scale Tumor Segmentation Foundation Model (TSFM) with 1.6 billion parameters using Resblock-backbone and Transformer-bottleneck,which has good transfer ability for downstream tasks. To make TSFM exhibit good performance in tumor segmentation, we make full use of the strong spatial correlation between tumors and organs in the medical image, innovatively fuse 7 tumor datasets and 3 multi-organ datasets to build a 3D medical dataset pool, including 2779 cases with totally 300k medical images, whose size currently exceeds many other single publicly available datasets. TSFM is the pre-trained model for medical image segmentation, which also can be transferred to multiple downstream tasks for fine-tuning learning. The average performance of our pre-trained model is 2% higher than that of nnU-Net across various tumor types. In the transfer learning task, TSFM only needs 5% training epochs of nnU-Net to achieve similar performance and can surpass nnU-Net by 2% on average with 10% training epoch. Pre-trained TSFM and its code will be released soon.", "authors": ["Jianhao Xie", "Ziang Zhang", "Guibo Luo", "Yuesheng Zhu"], "categories": ["eess.IV", "cs.CV"], "fields_of_study": ["Engineering", "Computer Science"], "published_date": "2024-03-11", "url": "https://arxiv.org/abs/2403.06396", "pdf_url": "https://arxiv.org/pdf/2403.06396v1", "arxiv_id": "2403.06396", "doi": "10.48550/arXiv.2403.06396", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0}
{"id": "3c504b3ad9a012ce3011989387018499c5d0ab82bb18c6ab6d7cfcf3567bd7bd", "sources": ["arxiv", "semantic_scholar"], "title": "Training-Free Pretrained Model Merging", "abstract": "Recently, model merging techniques have surfaced as a solution to combine multiple single-talent models into a single multi-talent model. However, previous endeavors in this field have either necessitated additional training or fine-tuning processes, or require that the models possess the same pre-trained initialization. In this work, we identify a common drawback in prior works w.r.t. the inconsistency of unit similarity in the weight space and the activation space. To address this inconsistency, we propose an innovative model merging framework, coined as merging under dual-space constraints (MuDSC). Specifically, instead of solely maximizing the objective of a single space, we advocate for the exploration of permutation matrices situated in a region with a unified high similarity in the dual space, achieved through the linear combination of activation and weight similarity matrices. In order to enhance usability, we have also incorporated adaptations for group structure, including Multi-Head Attention and Group Normalization. Comprehensive experimental comparisons demonstrate that MuDSC can significantly boost the performance of merged models with various task combinations and architectures. Furthermore, the visualization of the merged model within the multi-task loss landscape reveals that MuDSC enables the merged model to reside in the overlapping segment, featuring a unified lower loss for each task. Our code is publicly available at https://github.com/zju-vipa/training_free_model_merging.", "authors": ["Zhengqi Xu", "Ke Yuan", "Huiqiong Wang", "Yong Wang", "Mingli Song", "Jie Song"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-03-04", "url": "https://arxiv.org/abs/2403.01753", "pdf_url": "https://arxiv.org/pdf/2403.01753v3", "arxiv_id": "2403.01753", "doi": "10.1109/CVPR52733.2024.00565", "citation_count": 36, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/zju-vipa/training_free_model_merging", "venue": "Computer Vision and Pattern Recognition", "quality_score": 0.3921}
{"id": "97bbc26f25c0cfecc3ce52bebc8455cd8694b30312be2db9f7969c9522ce0271", "sources": ["arxiv", "semantic_scholar"], "title": "Merging Text Transformer Models from Different Initializations", "abstract": "Recent work on permutation-based model merging has shown impressive low- or zero-barrier mode connectivity between models from completely different initializations. However, this line of work has not yet extended to the Transformer architecture, despite its dominant popularity in the language domain. Therefore, in this work, we investigate the extent to which separate Transformer minima learn similar features, and propose a model merging technique to investigate the relationship between these minima in the loss landscape. The specifics of the architecture, like its residual connections, multi-headed attention, and discrete, sequential input, require specific interventions in order to compute model permutations that remain within the same functional equivalence class. In merging these models with our method, we consistently find lower loss barriers between minima compared to model averaging, across models trained on a masked-language modeling task or fine-tuned on a language understanding benchmark. Our results show that the minima of these models are less sharp and isolated than previously understood, and provide a basis for future work on merging separately trained Transformer models.", "authors": ["Neha Verma", "Maha Elbayad"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-03-01", "url": "https://arxiv.org/abs/2403.00986", "pdf_url": "https://arxiv.org/pdf/2403.00986v3", "arxiv_id": "2403.00986", "doi": "10.48550/arXiv.2403.00986", "citation_count": 15, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.301}
{"id": "72f48530cbcd9beeb969f2ab0c42661455ff0d298a65e8591be8d7aa8def508e", "sources": ["arxiv", "semantic_scholar"], "title": "Language Models are Homer Simpson! Safety Re-Alignment of Fine-tuned Language Models through Task Arithmetic", "abstract": "Aligned language models face a significant limitation as their fine-tuning often results in compromised safety. To tackle this, we propose a simple method RESTA that performs LLM safety realignment. RESTA stands for REstoring Safety through Task Arithmetic. At its core, it involves a simple arithmetic addition of a safety vector to the weights of the compromised model. We demonstrate the effectiveness of RESTA in both parameter-efficient and full fine-tuning, covering a wide range of downstream tasks, including instruction following in Chinese, English, and Hindi, as well as problem-solving capabilities in Code and Math. We also showcase the generalizability of RESTA on three existing safety evaluation benchmarks and a multilingual benchmark dataset proposed as a part of this work, consisting of 550 harmful questions covering 11 categories, each with 5 sub-categories of harm. Overall, RESTA decreases the harmfulness of the compromised model from 18.6% to 5.1% and from 9.2% to 1.5% in parameter-efficient and full fine-tuning, respectively, while maintaining most of the model's performance on the task. We release the source codes at: https://github.com/declare-lab/resta.", "authors": ["Rishabh Bhardwaj", "Do Duc Anh", "Soujanya Poria"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-02-19", "url": "https://arxiv.org/abs/2402.11746", "pdf_url": "https://arxiv.org/pdf/2402.11746v1", "arxiv_id": "2402.11746", "doi": "10.48550/arXiv.2402.11746", "citation_count": 106, "influential_citation_count": 14, "has_code": true, "code_url": "https://github.com/declare-lab/resta", "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.588}
{"id": "a1e3c7ec9b0520560ab9bd60709056519850bb92b6d4a63f65984a68f51b4726", "sources": ["arxiv", "semantic_scholar"], "title": "WKVQuant: Quantizing Weight and Key/Value Cache for Large Language Models Gains More", "abstract": "Large Language Models (LLMs) face significant deployment challenges due to their substantial memory requirements and the computational demands of auto-regressive text generation process. This paper addresses these challenges by focusing on the quantization of LLMs, a technique that reduces memory consumption by converting model parameters and activations into low-bit integers. We critically analyze the existing quantization approaches, identifying their limitations in balancing the accuracy and efficiency of the quantized LLMs. To advance beyond these limitations, we propose WKVQuant, a PTQ framework especially designed for quantizing weights and the key/value (KV) cache of LLMs. Specifically, we incorporates past-only quantization to improve the computation of attention. Additionally, we introduce two-dimensional quantization strategy to handle the distribution of KV cache, along with a cross-block reconstruction regularization for parameter optimization. Experiments show that WKVQuant achieves almost comparable memory savings to weight-activation quantization, while also approaching the performance of weight-only quantization.", "authors": ["Yuxuan Yue", "Zhihang Yuan", "Haojie Duanmu", "Sifan Zhou", "Jianlong Wu", "Liqiang Nie"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-02-19", "url": "https://arxiv.org/abs/2402.12065", "pdf_url": "https://arxiv.org/pdf/2402.12065v2", "arxiv_id": "2402.12065", "doi": "10.48550/arXiv.2402.12065", "citation_count": 85, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4836}
{"id": "c2e884c616d99c0b45de7f42f49d7c491805537ba6d607e924412c2438ec1a06", "sources": ["arxiv", "semantic_scholar"], "title": "Learning From Failure: Integrating Negative Examples when Fine-tuning Large Language Models as Agents", "abstract": "Large language models (LLMs) have achieved success in acting as agents, which interact with environments through tools such as search engines. However, LLMs are optimized for language generation instead of tool use during training or alignment, limiting their effectiveness as agents. To resolve this problem, previous work has first collected interaction trajectories between LLMs and environments, using only trajectories that successfully finished the task to fine-tune smaller models, making fine-tuning data scarce and acquiring it both difficult and costly. Discarding failed trajectories also leads to significant wastage of data and resources and limits the possible optimization paths during fine-tuning. In this paper, we argue that unsuccessful trajectories offer valuable insights, and LLMs can learn from these trajectories through appropriate quality control and fine-tuning strategies. By simply adding a prefix or suffix that tells the model whether to generate a successful trajectory during training, we improve model performance by a large margin on mathematical reasoning, multi-hop question answering, and strategic question answering tasks. We further analyze the inference results and find that our method provides a better trade-off between valuable information and errors in unsuccessful trajectories. To our knowledge, we are the first to demonstrate the value of negative trajectories and their application in agent-tunning scenarios. Our findings offer guidance for developing better agent-tuning methods and low-resource data usage techniques.", "authors": ["Renxi Wang", "Haonan Li", "Xudong Han", "Yixuan Zhang", "Timothy Baldwin"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-02-18", "url": "https://arxiv.org/abs/2402.11651", "pdf_url": "https://arxiv.org/pdf/2402.11651v2", "arxiv_id": "2402.11651", "doi": "10.48550/arXiv.2402.11651", "citation_count": 46, "influential_citation_count": 4, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.418}
{"id": "6dfa928b137534c5e038f8544820da357cf14b81f471f03c71ceeed777ab191f", "sources": ["arxiv", "semantic_scholar"], "title": "Pelican Soup Framework: A Theoretical Framework for Language Model Capabilities", "abstract": "In this work, we propose a simple theoretical framework, Pelican Soup, aiming to better understand how pretraining allows LLMs to (1) generalize to unseen instructions and (2) perform in-context learning, even when the verbalizers are irrelevant to the task. To this end, in our framework, we introduce the notion of \"knowledge base\" and \"reference-sense association\" and a simple formalism for natural language processing tasks. Our framework demonstrates how linguistic, psychology, and philosophy studies can inform our understanding of the language model and is connected to several other existing theoretical results. As an illustration of the usage of our framework, we derive a bound on in-context learning loss with our framework. Finally, we support our framework with empirical experiments and provide possible future research directions.", "authors": ["Ting-Rui Chiang", "Dani Yogatama"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-02-16", "url": "https://arxiv.org/abs/2402.10424", "pdf_url": "https://arxiv.org/pdf/2402.10424v2", "arxiv_id": "2402.10424", "doi": "10.18653/v1/2026.findings-eacl.23", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference of the European Chapter of the Association for Computational Linguistics", "quality_score": 0.1505}
{"id": "7c68eda12ca260f0e2fe8896f5fb5567c8b2f57a8ec118ca9a8d5a19f564b530", "sources": ["arxiv", "semantic_scholar"], "title": "Representation Surgery for Multi-Task Model Merging", "abstract": "Multi-task learning (MTL) compresses the information from multiple tasks into a unified backbone to improve computational efficiency and generalization. Recent work directly merges multiple independently trained models to perform MTL instead of collecting their raw data for joint training, greatly expanding the application scenarios of MTL. However, by visualizing the representation distribution of existing model merging schemes, we find that the merged model often suffers from the dilemma of representation bias. That is, there is a significant discrepancy in the representation distribution between the merged and individual models, resulting in poor performance of merged MTL. In this paper, we propose a representation surgery solution called \"Surgery\" to reduce representation bias in the merged model. Specifically, Surgery is a lightweight task-specific module that takes the representation of the merged model as input and attempts to output the biases contained in the representation from the merged model. We then designed an unsupervised optimization objective that updates the Surgery module by minimizing the distance between the merged model's representation and the individual model's representation. Extensive experiments demonstrate significant MTL performance improvements when our Surgery module is applied to state-of-the-art (SOTA) model merging schemes.", "authors": ["Enneng Yang", "Li Shen", "Zhenyi Wang", "Guibing Guo", "Xiaojun Chen", "Xingwei Wang", "Dacheng Tao"], "categories": ["cs.LG", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-02-05", "url": "https://arxiv.org/abs/2402.02705", "pdf_url": "https://arxiv.org/pdf/2402.02705v2", "arxiv_id": "2402.02705", "doi": "10.48550/arXiv.2402.02705", "citation_count": 104, "influential_citation_count": 11, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.5396}
{"id": "7638c58ec905a473a4ea4112c7d4ef10a0947a2817b4a515efc3b12aadefa56d", "sources": ["arxiv", "semantic_scholar"], "title": "Merging Multi-Task Models via Weight-Ensembling Mixture of Experts", "abstract": "Merging various task-specific Transformer-based models trained on different tasks into a single unified model can execute all the tasks concurrently. Previous methods, exemplified by task arithmetic, have been proven to be both effective and scalable. Existing methods have primarily focused on seeking a static optimal solution within the original model parameter space. A notable challenge is mitigating the interference between parameters of different models, which can substantially deteriorate performance. In this paper, we propose to merge most of the parameters while upscaling the MLP of the Transformer layers to a weight-ensembling mixture of experts (MoE) module, which can dynamically integrate shared and task-specific knowledge based on the input, thereby providing a more flexible solution that can adapt to the specific needs of each instance. Our key insight is that by identifying and separating shared knowledge and task-specific knowledge, and then dynamically integrating them, we can mitigate the parameter interference problem to a great extent. We conduct the conventional multi-task model merging experiments and evaluate the generalization and robustness of our method. The results demonstrate the effectiveness of our method and provide a comprehensive understanding of our method. The code is available at https://github.com/tanganke/weight-ensembling_MoE", "authors": ["Anke Tang", "Li Shen", "Yong Luo", "Nan Yin", "Lefei Zhang", "Dacheng Tao"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-02-01", "url": "https://arxiv.org/abs/2402.00433", "pdf_url": "https://arxiv.org/pdf/2402.00433v2", "arxiv_id": "2402.00433", "doi": "10.48550/arXiv.2402.00433", "citation_count": 99, "influential_citation_count": 13, "has_code": true, "code_url": "https://github.com/tanganke/weight-ensembling_MoE", "venue": "International Conference on Machine Learning", "quality_score": 0.5731}
{"id": "e8bd04765f835107f832036521f2a71da6fdc098b00987eb1f729ad8b42039a1", "sources": ["arxiv", "semantic_scholar"], "title": "RADIN: Souping on a Budget", "abstract": "Model Soups, extending Stochastic Weights Averaging (SWA), combine models fine-tuned with different hyperparameters. Yet, their adoption is hindered by computational challenges due to subset selection issues. In this paper, we propose to speed up model soups by approximating soups performance using averaged ensemble logits performances. Theoretical insights validate the congruence between ensemble logits and weight averaging soups across any mixing ratios. Our Resource ADjusted soups craftINg (RADIN) procedure stands out by allowing flexible evaluation budgets, enabling users to adjust his budget of exploration adapted to his resources while increasing performance at lower budget compared to previous greedy approach (up to 4% on ImageNet).", "authors": ["Thibaut Menes", "Olivier Risser-Maroix"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-01-31", "url": "https://arxiv.org/abs/2401.17790", "pdf_url": "https://arxiv.org/pdf/2401.17790v1", "arxiv_id": "2401.17790", "doi": "10.48550/arXiv.2401.17790", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0753}
{"id": "9cbc2ae410bc3e48828c92d7ca8ef5165650e95657c3000c5c4bcd4ffdc7ca50", "sources": ["arxiv", "semantic_scholar"], "title": "Active Inference as a Model of Agency", "abstract": "Is there a canonical way to think of agency beyond reward maximisation? In this paper, we show that any type of behaviour complying with physically sound assumptions about how macroscopic biological agents interact with the world canonically integrates exploration and exploitation in the sense of minimising risk and ambiguity about states of the world. This description, known as active inference, refines the free energy principle, a popular descriptive framework for action and perception originating in neuroscience. Active inference provides a normative Bayesian framework to simulate and model agency that is widely used in behavioural neuroscience, reinforcement learning (RL) and robotics. The usefulness of active inference for RL is three-fold. \\emph{a}) Active inference provides a principled solution to the exploration-exploitation dilemma that usefully simulates biological agency. \\emph{b}) It provides an explainable recipe to simulate behaviour, whence behaviour follows as an explainable mixture of exploration and exploitation under a generative world model, and all differences in behaviour are explicit in differences in world model. \\emph{c}) This framework is universal in the sense that it is theoretically possible to rewrite any RL algorithm conforming to the descriptive assumptions of active inference as an active inference algorithm. Thus, active inference can be used as a tool to uncover and compare the commitments and assumptions of more specific models of agency.", "authors": ["Lancelot Da Costa", "Samuel Tenka", "Dominic Zhao", "Noor Sajid"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-01-23", "url": "https://arxiv.org/abs/2401.12917", "pdf_url": "https://arxiv.org/pdf/2401.12917v1", "arxiv_id": "2401.12917", "doi": "10.48550/arXiv.2401.12917", "citation_count": 16, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3076}
{"id": "12843b8c954813f221897042e5d835db3ef3aa5d4ffd6b6611b5d29eda95802d", "sources": ["arxiv", "semantic_scholar"], "title": "CLIP Model for Images to Textual Prompts Based on Top-k Neighbors", "abstract": "Text-to-image synthesis, a subfield of multimodal generation, has gained significant attention in recent years. We propose a cost-effective approach for image-to-prompt generation that leverages generative models to generate textual prompts without the need for large amounts of annotated data. We divide our method into two stages: online stage and offline stage. We use a combination of the CLIP model and K-nearest neighbors (KNN) algorithm. The proposed system consists of two main parts: an offline task and an online task. Our method owns the highest metric 0.612 among these models, which is 0.013, 0.055, 0.011 higher than Clip, Clip + KNN(top 10) respectively.", "authors": ["Xin Zhang", "Xin Zhang", "YeMing Cai", "Tianzhi Jia"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-01-18", "url": "https://arxiv.org/abs/2401.09763", "pdf_url": "https://arxiv.org/pdf/2401.09763v1", "arxiv_id": "2401.09763", "doi": "10.1109/EIECS59936.2023.10435489", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1193}
{"id": "18fcca95943e51d2702adcd30900841dc0e750f405eeefdf754550b624b5021e", "sources": ["arxiv", "semantic_scholar"], "title": "Erasing Undesirable Influence in Diffusion Models", "abstract": "Diffusion models are highly effective at generating high-quality images but pose risks, such as the unintentional generation of NSFW (not safe for work) content. Although various techniques have been proposed to mitigate unwanted influences in diffusion models while preserving overall performance, achieving a balance between these goals remains challenging. In this work, we introduce EraseDiff, an algorithm designed to preserve the utility of the diffusion model on retained data while removing the unwanted information associated with the data to be forgotten. Our approach formulates this task as a constrained optimization problem using the value function, resulting in a natural first-order algorithm for solving the optimization problem. By altering the generative process to deviate away from the ground-truth denoising trajectory, we update parameters for preservation while controlling constraint reduction to ensure effective erasure, striking an optimal trade-off. Extensive experiments and thorough comparisons with state-of-the-art algorithms demonstrate that EraseDiff effectively preserves the model's utility, efficacy, and efficiency.", "authors": ["Jing Wu", "Trung Le", "Munawar Hayat", "Mehrtash Harandi"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-01-11", "url": "https://arxiv.org/abs/2401.05779", "pdf_url": "https://arxiv.org/pdf/2401.05779v4", "arxiv_id": "2401.05779", "doi": "10.1109/CVPR52734.2025.02632", "citation_count": 41, "influential_citation_count": 6, "has_code": false, "code_url": null, "venue": "Computer Vision and Pattern Recognition", "quality_score": 0.4225}
{"id": "24418c7947f9ccffd4ab0a6c799f77800098fb070f1e012311313c976d9396a4", "sources": ["arxiv", "semantic_scholar"], "title": "Merging Vision Transformers from Different Tasks and Domains", "abstract": "This work targets to merge various Vision Transformers (ViTs) trained on different tasks (i.e., datasets with different object categories) or domains (i.e., datasets with the same categories but different environments) into one unified model, yielding still good performance on each task or domain. Previous model merging works focus on either CNNs or NLP models, leaving the ViTs merging research untouched. To fill this gap, we first explore and find that existing model merging methods cannot well handle the merging of the whole ViT models and still have improvement space. To enable the merging of the whole ViT, we propose a simple-but-effective gating network that can both merge all kinds of layers (e.g., Embedding, Norm, Attention, and MLP) and select the suitable classifier. Specifically, the gating network is trained by unlabeled datasets from all the tasks (domains), and predicts the probability of which task (domain) the input belongs to for merging the models during inference. To further boost the performance of the merged model, especially when the difficulty of merging tasks increases, we design a novel metric of model weight similarity, and utilize it to realize controllable and combined weight merging. Comprehensive experiments on kinds of newly established benchmarks, validate the superiority of the proposed ViT merging framework for different tasks and domains. Our method can even merge beyond 10 ViT models from different vision tasks with a negligible effect on the performance of each task.", "authors": ["Peng Ye", "Chenyu Huang", "Mingzhu Shen", "Tao Chen", "Yongqi Huang", "Yuning Zhang", "Wanli Ouyang"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-12-25", "url": "https://arxiv.org/abs/2312.16240", "pdf_url": "https://arxiv.org/pdf/2312.16240v1", "arxiv_id": "2312.16240", "doi": "10.48550/arXiv.2312.16240", "citation_count": 20, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3306}
{"id": "90bb504f9c9dabcbee76afa6327ad677fa167d68c23e0a099c905bcc21c45b9a", "sources": ["arxiv", "semantic_scholar"], "title": "Model Breadcrumbs: Scaling Multi-Task Model Merging with Sparse Masks", "abstract": "The rapid development of AI systems has been greatly influenced by the emergence of foundation models. A common approach for targeted problems involves fine-tuning these pre-trained foundation models for specific target tasks, resulting in a rapid spread of models fine-tuned across a diverse array of tasks. This work focuses on the problem of merging multiple fine-tunings of the same foundation model derived from a spectrum of auxiliary tasks. We introduce a new simple method, Model Breadcrumbs, which consists of a sparsely defined weight set that guides model adaptation within the weight space of a pre-trained model. These breadcrumbs are constructed by subtracting the weights from a pre-trained model before and after fine-tuning, followed by a sparsification process that eliminates weight outliers and negligible perturbations. Our experiments demonstrate the effectiveness of Model Breadcrumbs to simultaneously improve performance across multiple tasks. This contribution aligns with the evolving paradigm of updatable machine learning, reminiscent of the collaborative principles underlying open-source software development, fostering a community-driven effort to reliably update machine learning models. Our method is shown to be more efficient and unlike previous proposals does not require hyperparameter tuning for each new task added. Through extensive experimentation involving various models, tasks, and modalities we establish that integrating Model Breadcrumbs offers a simple, efficient, and highly effective approach for constructing multi-task models and facilitating updates to foundation models.", "authors": ["MohammadReza Davari", "Eugene Belilovsky"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-12-11", "url": "https://arxiv.org/abs/2312.06795", "pdf_url": "https://arxiv.org/pdf/2312.06795v2", "arxiv_id": "2312.06795", "doi": "10.48550/arXiv.2312.06795", "citation_count": 130, "influential_citation_count": 11, "has_code": true, "code_url": null, "venue": "European Conference on Computer Vision", "quality_score": 0.5396}
{"id": "8fc405edf05abccb4ed4920eba8116a92f73e9bd758c5deb8a34d5ef250687e2", "sources": ["arxiv", "semantic_scholar"], "title": "Merging by Matching Models in Task Parameter Subspaces", "abstract": "Model merging aims to cheaply combine individual task-specific models into a single multitask model. In this work, we view past merging methods as leveraging different notions of a ''task parameter subspace'' in which models are matched before being merged. We connect the task parameter subspace of a given model to its loss landscape and formalize how this approach to model merging can be seen as solving a linear system of equations. While past work has generally been limited to linear systems that have a closed-form solution, we consider using the conjugate gradient method to find a solution. We show that using the conjugate gradient method can outperform closed-form solutions, enables merging via linear systems that are otherwise intractable to solve, and flexibly allows choosing from a wide variety of initializations and estimates for the ''task parameter subspace''. We ultimately demonstrate that our merging framework called ''Matching Models in their Task Parameter Subspace'' (MaTS) achieves state-of-the-art results in multitask and intermediate-task model merging. We release all of the code and checkpoints used in our work at https://github.com/r-three/mats.", "authors": ["Derek Tam", "Mohit Bansal", "Colin Raffel"], "categories": ["cs.LG", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-12-07", "url": "https://arxiv.org/abs/2312.04339", "pdf_url": "https://arxiv.org/pdf/2312.04339v2", "arxiv_id": "2312.04339", "doi": null, "citation_count": 30, "influential_citation_count": 3, "has_code": true, "code_url": "https://github.com/r-three/mats", "venue": null, "quality_score": 0.3728}
{"id": "6ac5bc2b17c92fb1f9cae5098e3b4430b28e582c22fe804209d3ae5815ac5d53", "sources": ["arxiv", "semantic_scholar"], "title": "Advances in the equivariant minimal model program and their applications in complex and arithmetic dynamics", "abstract": "This note reports some advances in the Equivariant Minimal Model Program (EMMP) for non-isomorphic surjective endomorphisms and their applications in complex and arithmetic dynamics.", "authors": ["Sheng Meng", "De-Qi Zhang"], "categories": ["math.AG", "math.DS", "math.NT"], "fields_of_study": ["Mathematics"], "published_date": "2023-11-27", "url": "https://arxiv.org/abs/2311.16369", "pdf_url": "https://arxiv.org/pdf/2311.16369v1", "arxiv_id": "2311.16369", "doi": "10.1007/978-3-032-04048-0_4", "citation_count": 8, "influential_citation_count": 4, "has_code": false, "code_url": null, "venue": "DeMarco, L., Jonsson, M. (eds) Algebraic, Complex, and Arithmetic Dynamics. Simons Symposia. Springer, Cham. yr 2026, pages 99-123", "quality_score": 0.3495}
{"id": "df1e874c85efbc5727e0229ca9923523e0a29352361e4ce75edabba001f445bd", "sources": ["arxiv", "semantic_scholar"], "title": "Model Theory of Ultrafinitism II: Deconstructing the Term Model (First Draft)", "abstract": "This paper presents a novel possible worlds semantics, designed to elucidate the underpinnings of ultrafinitism. By constructing a careful modification of the well-known Kripke models for inuitionistic logic, we seek to extend our comprehension of the ultra-finite mindset. As it turns out, the passage from standard constructivist mathematics to the ultrafinite is in a sense an operation of deconstruction of familiar mathematical entities, most notably clear when it comes to N.", "authors": ["Mirco A. Mannucci"], "categories": ["math.LO", "cs.LO"], "fields_of_study": ["Mathematics", "Computer Science"], "published_date": "2023-11-26", "url": "https://arxiv.org/abs/2311.17931", "pdf_url": "https://arxiv.org/pdf/2311.17931v1", "arxiv_id": "2311.17931", "doi": "10.48550/arXiv.2311.17931", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0}
{"id": "48dd35cd0752349bfb25bba9eb2cf109a8349769b764f9b7d68b2c611b76b1d7", "sources": ["arxiv", "semantic_scholar"], "title": "Orca 2: Teaching Small Language Models How to Reason", "abstract": "Orca 1 learns from rich signals, such as explanation traces, allowing it to outperform conventional instruction-tuned models on benchmarks like BigBench Hard and AGIEval. In Orca 2, we continue exploring how improved training signals can enhance smaller LMs' reasoning abilities. Research on training small LMs has often relied on imitation learning to replicate the output of more capable models. We contend that excessive emphasis on imitation may restrict the potential of smaller models. We seek to teach small LMs to employ different solution strategies for different tasks, potentially different from the one used by the larger model. For example, while larger models might provide a direct answer to a complex task, smaller models may not have the same capacity. In Orca 2, we teach the model various reasoning techniques (step-by-step, recall then generate, recall-reason-generate, direct answer, etc.). More crucially, we aim to help the model learn to determine the most effective solution strategy for each task. We evaluate Orca 2 using a comprehensive set of 15 diverse benchmarks (corresponding to approximately 100 tasks and over 36,000 unique prompts). Orca 2 significantly surpasses models of similar size and attains performance levels similar or better to those of models 5-10x larger, as assessed on complex tasks that test advanced reasoning abilities in zero-shot settings. make Orca 2 weights publicly available at aka.ms/orca-lm to support research on the development, evaluation, and alignment of smaller LMs", "authors": ["Arindam Mitra", "Luciano Del Corro", "Shweti Mahajan", "Andres Codas", "Clarisse Simoes", "Sahaj Agarwal", "Xuxi Chen", "Anastasia Razdaibiedina", "Erik Jones", "Kriti Aggarwal", "Hamid Palangi", "Guoqing Zheng", "Corby Rosset", "Hamed Khanpour", "Ahmed Awadallah"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-11-18", "url": "https://arxiv.org/abs/2311.11045", "pdf_url": "https://arxiv.org/pdf/2311.11045v2", "arxiv_id": "2311.11045", "doi": "10.48550/arXiv.2311.11045", "citation_count": 203, "influential_citation_count": 17, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.6276}
{"id": "c73feb5738386fecd98077c0ba0bd3eb71297f63cb73d62cc32c5e53ce00b84b", "sources": ["arxiv", "semantic_scholar"], "title": "PPTC Benchmark: Evaluating Large Language Models for PowerPoint Task Completion", "abstract": "Recent evaluations of Large Language Models (LLMs) have centered around testing their zero-shot/few-shot capabilities for basic natural language tasks and their ability to translate instructions into tool APIs. However, the evaluation of LLMs utilizing complex tools to finish multi-turn, multi-modal instructions in a complex multi-modal environment has not been investigated. To address this gap, we introduce the PowerPoint Task Completion (PPTC) benchmark to assess LLMs' ability to create and edit PPT files based on user instructions. It contains 279 multi-turn sessions covering diverse topics and hundreds of instructions involving multi-modal operations. We also propose the PPTX-Match Evaluation System that evaluates if LLMs finish the instruction based on the prediction file rather than the label API sequence, thus it supports various LLM-generated API sequences. We measure 3 closed LLMs and 6 open-source LLMs. The results show that GPT-4 outperforms other LLMs with 75.1\\% accuracy in single-turn dialogue testing but faces challenges in completing entire sessions, achieving just 6\\% session accuracy. We find three main error causes in our benchmark: error accumulation in the multi-turn session, long PPT template processing, and multi-modality perception. These pose great challenges for future LLM and agent systems. We release the data, code, and evaluation system of PPTC at \\url{https://github.com/gydpku/PPTC}.", "authors": ["Yiduo Guo", "Zekai Zhang", "Yaobo Liang", "Dongyan Zhao", "Nan Duan"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-11-03", "url": "https://arxiv.org/abs/2311.01767", "pdf_url": "https://arxiv.org/pdf/2311.01767v2", "arxiv_id": "2311.01767", "doi": "10.48550/arXiv.2311.01767", "citation_count": 30, "influential_citation_count": 4, "has_code": true, "code_url": "https://github.com/gydpku/PPTC}", "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.3728}
{"id": "f40bb48f8ff9401d0a44c442ec42131e4a5ceccda687c8cd30c7e15db8481d37", "sources": ["arxiv", "semantic_scholar"], "title": "Model Merging by Uncertainty-Based Gradient Matching", "abstract": "Models trained on different datasets can be merged by a weighted-averaging of their parameters, but why does it work and when can it fail? Here, we connect the inaccuracy of weighted-averaging to mismatches in the gradients and propose a new uncertainty-based scheme to improve the performance by reducing the mismatch. The connection also reveals implicit assumptions in other schemes such as averaging, task arithmetic, and Fisher-weighted averaging. Our new method gives consistent improvements for large language models and vision transformers, both in terms of performance and robustness to hyperparameters. Code available here.", "authors": ["Nico Daheim", "Thomas Möllenhoff", "Edoardo Maria Ponti", "Iryna Gurevych", "Mohammad Emtiyaz Khan"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-10-19", "url": "https://arxiv.org/abs/2310.12808", "pdf_url": "https://arxiv.org/pdf/2310.12808v2", "arxiv_id": "2310.12808", "doi": "10.48550/arXiv.2310.12808", "citation_count": 90, "influential_citation_count": 7, "has_code": true, "code_url": "https://github.com/UKPLab/iclr2024-model-merging", "venue": "International Conference on Learning Representations", "quality_score": 0.4898}
{"id": "3e2a7ecf50bf40e41986eb2987bc7740bfc76f6ff42dfa4d9975106a6b584a49", "sources": ["arxiv", "semantic_scholar"], "title": "AdaMerging: Adaptive Model Merging for Multi-Task Learning", "abstract": "Multi-task learning (MTL) aims to empower a model to tackle multiple tasks simultaneously. A recent development known as task arithmetic has revealed that several models, each fine-tuned for distinct tasks, can be directly merged into a single model to execute MTL without necessitating a retraining process using the initial training data. Nevertheless, this direct addition of models often leads to a significant deterioration in the overall performance of the merged model. This decline occurs due to potential conflicts and intricate correlations among the multiple tasks. Consequently, the challenge emerges of how to merge pre-trained models more effectively without using their original training data. This paper introduces an innovative technique called Adaptive Model Merging (AdaMerging). This approach aims to autonomously learn the coefficients for model merging, either in a task-wise or layer-wise manner, without relying on the original training data. Specifically, our AdaMerging method operates as an automatic, unsupervised task arithmetic scheme. It leverages entropy minimization on unlabeled test samples from the multi-task setup as a surrogate objective function to iteratively refine the merging coefficients of the multiple models. Our experimental findings across eight tasks demonstrate the efficacy of the AdaMerging scheme we put forth. Compared to the current state-of-the-art task arithmetic merging scheme, AdaMerging showcases a remarkable 11\\% improvement in performance. Notably, AdaMerging also exhibits superior generalization capabilities when applied to unseen downstream tasks. Furthermore, it displays a significantly enhanced robustness to data distribution shifts that may occur during the testing phase.", "authors": ["Enneng Yang", "Zhenyi Wang", "Li Shen", "Shiwei Liu", "Guibing Guo", "Xingwei Wang", "Dacheng Tao"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2023-10-04", "url": "https://arxiv.org/abs/2310.02575", "pdf_url": "https://arxiv.org/pdf/2310.02575v2", "arxiv_id": "2310.02575", "doi": "10.48550/arXiv.2310.02575", "citation_count": 251, "influential_citation_count": 49, "has_code": false, "code_url": null, "venue": "International Conference on Learning Representations", "quality_score": 0.8495}
{"id": "f9f5af4b05e69d84a46f2720a830267b7626e1098965fa1e851a5d3b98fa04d9", "sources": ["arxiv", "semantic_scholar"], "title": "Soft Merging: A Flexible and Robust Soft Model Merging Approach for Enhanced Neural Network Performance", "abstract": "Stochastic Gradient Descent (SGD), a widely used optimization algorithm in deep learning, is often limited to converging to local optima due to the non-convex nature of the problem. Leveraging these local optima to improve model performance remains a challenging task. Given the inherent complexity of neural networks, the simple arithmetic averaging of the obtained local optima models in undesirable results. This paper proposes a {\\em soft merging} method that facilitates rapid merging of multiple models, simplifies the merging of specific parts of neural networks, and enhances robustness against malicious models with extreme values. This is achieved by learning gate parameters through a surrogate of the $l_0$ norm using hard concrete distribution without modifying the model weights of the given local optima models. This merging process not only enhances the model performance by converging to a better local optimum, but also minimizes computational costs, offering an efficient and explicit learning process integrated with stochastic gradient descent. Thorough experiments underscore the effectiveness and superior performance of the merged neural networks.", "authors": ["Hao Chen", "Yusen Wu", "Phuong Nguyen", "Chao Liu", "Yelena Yesha"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-09-21", "url": "https://arxiv.org/abs/2309.12259", "pdf_url": "https://arxiv.org/pdf/2309.12259v1", "arxiv_id": "2309.12259", "doi": "10.48550/arXiv.2309.12259", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0}
{"id": "1228aed58d0055d19b8c959ff70dd01706e62c39d913f12f1e0e3648a1b174a6", "sources": ["arxiv", "semantic_scholar"], "title": "Jais and Jais-chat: Arabic-Centric Foundation and Instruction-Tuned Open Generative Large Language Models", "abstract": "We introduce Jais and Jais-chat, new state-of-the-art Arabic-centric foundation and instruction-tuned open generative large language models (LLMs). The models are based on the GPT-3 decoder-only architecture and are pretrained on a mixture of Arabic and English texts, including source code in various programming languages. With 13 billion parameters, they demonstrate better knowledge and reasoning capabilities in Arabic than any existing open Arabic and multilingual models by a sizable margin, based on extensive evaluation. Moreover, the models are competitive in English compared to English-centric open models of similar size, despite being trained on much less English data. We provide a detailed description of the training, the tuning, the safety alignment, and the evaluation of the models. We release two open versions of the model -- the foundation Jais model, and an instruction-tuned Jais-chat variant -- with the aim of promoting research on Arabic LLMs. Available at https://huggingface.co/inception-mbzuai/jais-13b-chat", "authors": ["Neha Sengupta", "Sunil Kumar Sahu", "Bokang Jia", "Satheesh Katipomu", "Haonan Li", "Fajri Koto", "William Marshall", "Gurpreet Gosal", "Cynthia Liu", "Zhiming Chen", "Osama Mohammed Afzal", "Samta Kamboj", "Onkar Pandit", "Rahul Pal", "Lalit Pradhan", "Zain Muhammad Mujahid", "Massa Baali", "Xudong Han", "Sondos Mahmoud Bsharat", "Alham Fikri Aji", "Zhiqiang Shen", "Zhengzhong Liu", "Natalia Vassilieva", "Joel Hestness", "Andy Hock", "Andrew Feldman", "Jonathan Lee", "Andrew Jackson", "Hector Xuguang Ren", "Preslav Nakov", "Timothy Baldwin", "Eric Xing"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-08-30", "url": "https://arxiv.org/abs/2308.16149", "pdf_url": "https://arxiv.org/pdf/2308.16149v2", "arxiv_id": "2308.16149", "doi": "10.48550/arXiv.2308.16149", "citation_count": 81, "influential_citation_count": 8, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4785}
{"id": "0c4d4c25e58d3df50baf1e9e97d2ef83020f666d639d233208e3cc873bc3875e", "sources": ["arxiv", "semantic_scholar"], "title": "Do the Frankenstein, or how to achieve better out-of-distribution performance with manifold mixing model soup", "abstract": "The standard recipe applied in transfer learning is to finetune a pretrained model on the task-specific dataset with different hyperparameter settings and pick the model with the highest accuracy on the validation dataset. Unfortunately, this leads to models which do not perform well under distribution shifts, e.g. when the model is given graphical sketches of the object as input instead of photos. In order to address this, we propose the manifold mixing model soup, an algorithm which mixes together the latent space manifolds of multiple finetuned models in an optimal way in order to generate a fused model. We show that the fused model gives significantly better out-of-distribution performance (+3.5 % compared to best individual model) when finetuning a CLIP model for image classification. In addition, it provides also better accuracy on the original dataset where the finetuning has been done.", "authors": ["Hannes Fassold"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-08-28", "url": "https://arxiv.org/abs/2309.08610", "pdf_url": "https://arxiv.org/pdf/2309.08610v1", "arxiv_id": "2309.08610", "doi": "10.48550/arXiv.2309.08610", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1193}
{"id": "359a9f43b73be1ae8b68d02d70a11b2c929954c81d4809c08be6a4d172098efe", "sources": ["arxiv", "semantic_scholar"], "title": "Testing different Log Bases For Vector Model Weighting Technique", "abstract": "Information retrieval systems retrieves relevant documents based on a query submitted by the user. The documents are initially indexed and the words in the documents are assigned weights using a weighting technique called TFIDF which is the product of Term Frequency (TF) and Inverse Document Frequency (IDF). TF represents the number of occurrences of a term in a document. IDF measures whether the term is common or rare across all documents. It is computed by dividing the total number of documents in the system by the number of documents containing the term and then computing the logarithm of the quotient. By default, we use base 10 to calculate the logarithm. In this paper, we are going to test this weighting technique by using a range of log bases from 0.1 to 100.0 to calculate the IDF. Testing different log bases for vector model weighting technique is to highlight the importance of understanding the performance of the system at different weighting values. We use the documents of MED, CRAN, NPL, LISA, and CISI test collections that scientists assembled explicitly for experiments in data information retrieval systems.", "authors": ["Kamel Assaf"], "categories": ["cs.IR", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-07-12", "url": "https://arxiv.org/abs/2307.06213", "pdf_url": "https://arxiv.org/pdf/2307.06213v1", "arxiv_id": "2307.06213", "doi": "10.5121/ijnlc.2023.12301", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Journal on Natural Language Computing", "quality_score": 0.0}
{"id": "449368457b7e4ac759e65dcfcfc211dea651c7098e34835201caf6a17dddfc58", "sources": ["arxiv", "semantic_scholar"], "title": "TBGC: Task-level Backbone-Oriented Gradient Clip for Multi-Task Foundation Model Learning", "abstract": "The AllInOne training paradigm squeezes a wide range of tasks into a unified model in a multi-task learning manner. However, optimization in multi-task learning is more challenge than single-task learning, as the gradient norm from different tasks may vary greatly, making the backbone overly biased towards one specific task. To address this issue, we propose the task-level backbone-oriented gradient clip paradigm, compared with the vanilla gradient clip method, it has two points of emphasis:1) gradient clip is performed independently for each task. 2) backbone gradients generated from each task are rescaled to the same norm scale. Based on the experimental results, we argue that the task-level backbone-oriented gradient clip paradigm can relieve the gradient bias problem to some extent. We also propose a novel multi-branch data augmentation strategy where conflict augmentations are placed in different branches. Our approach has been shown to be effective and finally achieve 1st place in the Leaderboard A and 2nd place in the Leaderboard B of the CVPR2023 Foundation Model Challenge. It's worth noting that instead of evaluating all three tasks(detection, segmentation and fine-grained classification) in Leaderboard A, the segmentation task is not evaluated in Leaderboard B, in which our team has a huge advantage.", "authors": ["Zelun Zhang", "Xue Pan"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-07-07", "url": "https://arxiv.org/abs/2307.03465", "pdf_url": "https://arxiv.org/pdf/2307.03465v1", "arxiv_id": "2307.03465", "doi": "10.48550/arXiv.2307.03465", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0}
{"id": "8c1b1df24db248ecde782eddfc6095117f94be642a0d0cf3da7df321d92613d2", "sources": ["arxiv", "semantic_scholar"], "title": "A Critical Look at the Current Usage of Foundation Model for Dense Recognition Task", "abstract": "In recent years large model trained on huge amount of cross-modality data, which is usually be termed as foundation model, achieves conspicuous accomplishment in many fields, such as image recognition and generation. Though achieving great success in their original application case, it is still unclear whether those foundation models can be applied to other different downstream tasks. In this paper, we conduct a short survey on the current methods for discriminative dense recognition tasks, which are built on the pretrained foundation model. And we also provide some preliminary experimental analysis of an existing open-vocabulary segmentation method based on Stable Diffusion, which indicates the current way of deploying diffusion model for segmentation is not optimal. This aims to provide insights for future research on adopting foundation model for downstream task.", "authors": ["Shiqi Yang", "Atsushi Hashimoto", "Yoshitaka Ushiku"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2023-07-06", "url": "https://arxiv.org/abs/2307.02862", "pdf_url": "https://arxiv.org/pdf/2307.02862v2", "arxiv_id": "2307.02862", "doi": "10.48550/arXiv.2307.02862", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0753}
{"id": "5b041de7260623b1492c0988a7d3868b43ecb49875106769cd84c61c6de33980", "sources": ["arxiv", "semantic_scholar"], "title": "Sparse Model Soups: A Recipe for Improved Pruning via Model Averaging", "abstract": "Neural networks can be significantly compressed by pruning, yielding sparse models with reduced storage and computational demands while preserving predictive performance. Model soups (Wortsman et al., 2022) enhance generalization and out-of-distribution (OOD) performance by averaging the parameters of multiple models into a single one, without increasing inference time. However, achieving both sparsity and parameter averaging is challenging as averaging arbitrary sparse models reduces the overall sparsity due to differing sparse connectivities. This work addresses these challenges by demonstrating that exploring a single retraining phase of Iterative Magnitude Pruning (IMP) with varied hyperparameter configurations such as batch ordering or weight decay yields models suitable for averaging, sharing identical sparse connectivity by design. Averaging these models significantly enhances generalization and OOD performance over their individual counterparts. Building on this, we introduce Sparse Model Soups (SMS), a novel method for merging sparse models by initiating each prune-retrain cycle with the averaged model from the previous phase. SMS preserves sparsity, exploits sparse network benefits, is modular and fully parallelizable, and substantially improves IMP's performance. We further demonstrate that SMS can be adapted to enhance state-of-the-art pruning-during-training approaches.", "authors": ["Max Zimmer", "Christoph Spiegel", "Sebastian Pokutta"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-06-29", "url": "https://arxiv.org/abs/2306.16788", "pdf_url": "https://arxiv.org/pdf/2306.16788v3", "arxiv_id": "2306.16788", "doi": "10.48550/arXiv.2306.16788", "citation_count": 22, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "International Conference on Learning Representations", "quality_score": 0.3404}
{"id": "fad81fd31133cd7516f98da301cae63fcab4654ed40279c46d410dbcc32115d1", "sources": ["arxiv", "semantic_scholar"], "title": "Low-Rank Prune-And-Factorize for Language Model Compression", "abstract": "The components underpinning PLMs -- large weight matrices -- were shown to bear considerable redundancy. Matrix factorization, a well-established technique from matrix theory, has been utilized to reduce the number of parameters in PLM. However, it fails to retain satisfactory performance under moderate to high compression rate. In this paper, we identify the \\textit{full-rankness} of fine-tuned PLM as the fundamental bottleneck for the failure of matrix factorization and explore the use of network pruning to extract low-rank sparsity pattern desirable to matrix factorization. We find such low-rank sparsity pattern exclusively exists in models generated by first-order pruning, which motivates us to unite the two approaches and achieve more effective model compression. We further propose two techniques: sparsity-aware SVD and mixed-rank fine-tuning, which improve the initialization and training of the compression procedure, respectively. Experiments on GLUE and question-answering tasks show that the proposed method has superior compression-performance trade-off compared to existing approaches.", "authors": ["Siyu Ren", "Kenny Q. Zhu"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-06-25", "url": "https://arxiv.org/abs/2306.14152", "pdf_url": "https://arxiv.org/pdf/2306.14152v1", "arxiv_id": "2306.14152", "doi": "10.48550/arXiv.2306.14152", "citation_count": 19, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "International Conference on Language Resources and Evaluation", "quality_score": 0.3253}
{"id": "2e4c5da564e29bd0fec7fba2ee0db41b2dafbdb3b61df2f0ff07f0f4688fc024", "sources": ["arxiv", "semantic_scholar"], "title": "Rewarded soups: towards Pareto-optimal alignment by interpolating weights fine-tuned on diverse rewards", "abstract": "Foundation models are first pre-trained on vast unsupervised datasets and then fine-tuned on labeled data. Reinforcement learning, notably from human feedback (RLHF), can further align the network with the intended usage. Yet the imperfections in the proxy reward may hinder the training and lead to suboptimal results; the diversity of objectives in real-world tasks and human opinions exacerbate the issue. This paper proposes embracing the heterogeneity of diverse rewards by following a multi-policy strategy. Rather than focusing on a single a priori reward, we aim for Pareto-optimal generalization across the entire space of preferences. To this end, we propose rewarded soup, first specializing multiple networks independently (one for each proxy reward) and then interpolating their weights linearly. This succeeds empirically because we show that the weights remain linearly connected when fine-tuned on diverse rewards from a shared pre-trained initialization. We demonstrate the effectiveness of our approach for text-to-text (summarization, Q&A, helpful assistant, review), text-image (image captioning, text-to-image generation, visual grounding, VQA), and control (locomotion) tasks. We hope to enhance the alignment of deep models, and how they interact with the world in all its diversity.", "authors": ["Alexandre Ramé", "Guillaume Couairon", "Mustafa Shukor", "Corentin Dancette", "Jean-Baptiste Gaya", "Laure Soulier", "Matthieu Cord"], "categories": ["cs.LG", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2023-06-07", "url": "https://arxiv.org/abs/2306.04488", "pdf_url": "https://arxiv.org/pdf/2306.04488v2", "arxiv_id": "2306.04488", "doi": "10.48550/arXiv.2306.04488", "citation_count": 252, "influential_citation_count": 35, "has_code": false, "code_url": null, "venue": "Neural Information Processing Systems", "quality_score": 0.7782}
{"id": "8d45ab4b36ccb812b56bc774b27276e415acde39ec1875f3e0aa013eda05b006", "sources": ["arxiv", "semantic_scholar"], "title": "TIES-Merging: Resolving Interference When Merging Models", "abstract": "Transfer learning - i.e., further fine-tuning a pre-trained model on a downstream task - can confer significant advantages, including improved downstream performance, faster convergence, and better sample efficiency. These advantages have led to a proliferation of task-specific fine-tuned models, which typically can only perform a single task and do not benefit from one another. Recently, model merging techniques have emerged as a solution to combine multiple task-specific models into a single multitask model without performing additional training. However, existing merging methods often ignore the interference between parameters of different models, resulting in large performance drops when merging multiple models. In this paper, we demonstrate that prior merging techniques inadvertently lose valuable information due to two major sources of interference: (a) interference due to redundant parameter values and (b) disagreement on the sign of a given parameter's values across models. To address this, we propose our method, TRIM, ELECT SIGN & MERGE (TIES-Merging), which introduces three novel steps when merging models: (1) resetting parameters that only changed a small amount during fine-tuning, (2) resolving sign conflicts, and (3) merging only the parameters that are in alignment with the final agreed-upon sign. We find that TIES-Merging outperforms several existing methods in diverse settings covering a range of modalities, domains, number of tasks, model sizes, architectures, and fine-tuning settings. We further analyze the impact of different types of interference on model parameters, and highlight the importance of resolving sign interference. Our code is available at https://github.com/prateeky2806/ties-merging", "authors": ["Prateek Yadav", "Derek Tam", "Leshem Choshen", "Colin Raffel", "Mohit Bansal"], "categories": ["cs.LG", "cs.AI", "cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2023-06-02", "url": "https://arxiv.org/abs/2306.01708", "pdf_url": "https://arxiv.org/pdf/2306.01708v2", "arxiv_id": "2306.01708", "doi": "10.52202/075280-0310", "citation_count": 763, "influential_citation_count": 207, "has_code": true, "code_url": "https://github.com/prateeky2806/ties-merging", "venue": "Neural Information Processing Systems", "quality_score": 1.0}
{"id": "3997689b467b637c16b76e9dd3e22e3bd84348a74367fef05e99a2e5c0e70069", "sources": ["arxiv", "semantic_scholar"], "title": "Latent Diffusion Model Based Foley Sound Generation System For DCASE Challenge 2023 Task 7", "abstract": "Foley sound presents the background sound for multimedia content and the generation of Foley sound involves computationally modelling sound effects with specialized techniques. In this work, we proposed a system for DCASE 2023 challenge task 7: Foley Sound Synthesis. The proposed system is based on AudioLDM, which is a diffusion-based text-to-audio generation model. To alleviate the data-hungry problem, the system first trained with large-scale datasets and then downstreamed into this DCASE task via transfer learning. Through experiments, we found out that the feature extracted by the encoder can significantly affect the performance of the generation model. Hence, we improve the results by leveraging the input label with related text embedding features obtained by a significant language model, i.e., contrastive language-audio pertaining (CLAP). In addition, we utilize a filtering strategy to further refine the output, i.e. by selecting the best results from the candidate clips generated in terms of the similarity score between the sound and target labels. The overall system achieves a Frechet audio distance (FAD) score of 4.765 on average among all seven different classes, substantially outperforming the baseline system which performs a FAD score of 9.7.", "authors": ["Yi Yuan", "Haohe Liu", "Xubo Liu", "Xiyuan Kang", "Mark D. Plumbley", "Wenwu Wang"], "categories": ["cs.SD", "cs.MM", "eess.AS"], "fields_of_study": ["Computer Science", "Engineering"], "published_date": "2023-05-25", "url": "https://arxiv.org/abs/2305.15905", "pdf_url": "https://arxiv.org/pdf/2305.15905v3", "arxiv_id": "2305.15905", "doi": "10.48550/arXiv.2305.15905", "citation_count": 11, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2698}
{"id": "7420aa9ccdb2f04bfd32a53a745bc0cc773b11a92c9e4374e0fd95f3d8f5219e", "sources": ["arxiv", "semantic_scholar"], "title": "Adapting Language Models to Compress Contexts", "abstract": "Transformer-based language models (LMs) are powerful and widely-applicable tools, but their usefulness is constrained by a finite context window and the expensive computational cost of processing long text documents. We propose to adapt pre-trained LMs into AutoCompressors. These language models are capable of compressing long contexts into compact summary vectors, which are then accessible to the model as soft prompts. Summary vectors are trained with an unsupervised objective, whereby long documents are processed in segments, and summary vectors from all previous segments are used in language modeling. We fine-tune OPT and Llama-2 models on sequences of up to 30,720 tokens and show that AutoCompressors can utilize long contexts to improve perplexity. We evaluate AutoCompressors on in-context learning by compressing task demonstrations and find that summary vectors are good substitutes for plain-text demonstrations, increasing accuracy while reducing inference costs. Finally, we explore the benefits of pre-computing summary vectors for large corpora by applying summary vectors to retrievalaugmented language modeling and a passage re-ranking task. Overall, AutoCompressors emerge as a simple and inexpensive solution to extend the context window of LMs while speeding up inference over long contexts.", "authors": ["Alexis Chevalier", "Alexander Wettig", "Anirudh Ajith", "Danqi Chen"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-05-24", "url": "https://arxiv.org/abs/2305.14788", "pdf_url": "https://arxiv.org/pdf/2305.14788v2", "arxiv_id": "2305.14788", "doi": "10.48550/arXiv.2305.14788", "citation_count": 340, "influential_citation_count": 30, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.7457}
{"id": "8955556fa2222a75dbf62343a5a1dee7cfe69be86edbe6ebfc960feb9af8fe95", "sources": ["arxiv", "semantic_scholar"], "title": "Task Arithmetic in the Tangent Space: Improved Editing of Pre-Trained Models", "abstract": "Task arithmetic has recently emerged as a cost-effective and scalable approach to edit pre-trained models directly in weight space: By adding the fine-tuned weights of different tasks, the model's performance can be improved on these tasks, while negating them leads to task forgetting. Yet, our understanding of the effectiveness of task arithmetic and its underlying principles remains limited. We present a comprehensive study of task arithmetic in vision-language models and show that weight disentanglement is the crucial factor that makes it effective. This property arises during pre-training and manifests when distinct directions in weight space govern separate, localized regions in function space associated with the tasks. Notably, we show that fine-tuning models in their tangent space by linearizing them amplifies weight disentanglement. This leads to substantial performance improvements across multiple task arithmetic benchmarks and diverse models. Building on these findings, we provide theoretical and empirical analyses of the neural tangent kernel (NTK) of these models and establish a compelling link between task arithmetic and the spatial localization of the NTK eigenfunctions. Overall, our work uncovers novel insights into the fundamental mechanisms of task arithmetic and offers a more reliable and effective approach to edit pre-trained models through the NTK linearization.", "authors": ["Guillermo Ortiz-Jimenez", "Alessandro Favero", "Pascal Frossard"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2023-05-22", "url": "https://arxiv.org/abs/2305.12827", "pdf_url": "https://arxiv.org/pdf/2305.12827v3", "arxiv_id": "2305.12827", "doi": "10.48550/arXiv.2305.12827", "citation_count": 221, "influential_citation_count": 16, "has_code": false, "code_url": null, "venue": "Neural Information Processing Systems", "quality_score": 0.6152}
{"id": "4dbeedd6e5b4a89c200ca4afad9b337ed039a4ee6d92cc4f67d1850032837765", "sources": ["arxiv", "semantic_scholar"], "title": "MetaMorphosis: Task-oriented Privacy Cognizant Feature Generation for Multi-task Learning", "abstract": "With the growth of computer vision applications, deep learning, and edge computing contribute to ensuring practical collaborative intelligence (CI) by distributing the workload among edge devices and the cloud. However, running separate single-task models on edge devices is inefficient regarding the required computational resource and time. In this context, multi-task learning allows leveraging a single deep learning model for performing multiple tasks, such as semantic segmentation and depth estimation on incoming video frames. This single processing pipeline generates common deep features that are shared among multi-task modules. However, in a collaborative intelligence scenario, generating common deep features has two major issues. First, the deep features may inadvertently contain input information exposed to the downstream modules (violating input privacy). Second, the generated universal features expose a piece of collective information than what is intended for a certain task, in which features for one task can be utilized to perform another task (violating task privacy). This paper proposes a novel deep learning-based privacy-cognizant feature generation process called MetaMorphosis that limits inference capability to specific tasks at hand. To achieve this, we propose a channel squeeze-excitation based feature metamorphosis module, Cross-SEC, to achieve distinct attention of all tasks and a de-correlation loss function with differential-privacy to train a deep learning model that produces distinct privacy-aware features as an output for the respective tasks. With extensive experimentation on four datasets consisting of diverse images related to scene understanding and facial attributes, we show that MetaMorphosis outperforms recent adversarial learning and universal feature generation methods by guaranteeing privacy requirements in an efficient way for image and video analytics.", "authors": ["Md Adnan Arefeen", "Zhouyu Li", "Md Yusuf Sarwar Uddin", "Anupam Das"], "categories": ["cs.CV", "cs.CR", "cs.DC"], "fields_of_study": ["Computer Science"], "published_date": "2023-05-13", "url": "https://arxiv.org/abs/2305.07815", "pdf_url": "https://arxiv.org/pdf/2305.07815v1", "arxiv_id": "2305.07815", "doi": "10.1145/3576842.3582372", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Conference on Internet-of-Things Design and Implementation", "quality_score": 0.0}
{"id": "b2c950f56cf7e45ba34a0e041e7ac60db327e0fabd6cdb051d141c7ed740c04e", "sources": ["arxiv", "semantic_scholar"], "title": "HACK: Learning a Parametric Head and Neck Model for High-fidelity Animation", "abstract": "Significant advancements have been made in developing parametric models for digital humans, with various approaches concentrating on parts such as the human body, hand, or face. Nevertheless, connectors such as the neck have been overlooked in these models, with rich anatomical priors often unutilized. In this paper, we introduce HACK (Head-And-neCK), a novel parametric model for constructing the head and cervical region of digital humans. Our model seeks to disentangle the full spectrum of neck and larynx motions, facial expressions, and appearance variations, providing personalized and anatomically consistent controls, particularly for the neck regions. To build our HACK model, we acquire a comprehensive multi-modal dataset of the head and neck under various facial expressions. We employ a 3D ultrasound imaging scheme to extract the inner biomechanical structures, namely the precise 3D rotation information of the seven vertebrae of the cervical spine. We then adopt a multi-view photometric approach to capture the geometry and physically-based textures of diverse subjects, who exhibit a diverse range of static expressions as well as sequential head-and-neck movements. Using the multi-modal dataset, we train the parametric HACK model by separating the 3D head and neck depiction into various shape, pose, expression, and larynx blendshapes from the neutral expression and the rest skeletal pose. We adopt an anatomically-consistent skeletal design for the cervical region, and the expression is linked to facial action units for artist-friendly controls. HACK addresses the head and neck as a unified entity, offering more accurate and expressive controls, with a new level of realism, particularly for the neck regions. This approach has significant benefits for numerous applications and enables inter-correlation analysis between head and neck for fine-grained motion synthesis and transfer.", "authors": ["Longwen Zhang", "Zijun Zhao", "Xinzhou Cong", "Qixuan Zhang", "Shuqi Gu", "Yuchong Gao", "Rui Zheng", "Wei Yang", "Lan Xu", "Jingyi Yu"], "categories": ["cs.GR"], "fields_of_study": ["Computer Science"], "published_date": "2023-05-08", "url": "https://arxiv.org/abs/2305.04469", "pdf_url": "https://arxiv.org/pdf/2305.04469v1", "arxiv_id": "2305.04469", "doi": "10.1145/3592093", "citation_count": 17, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/ZoneLikeWonderland/HACK-Model", "venue": "ACM Transactions on Graphics", "quality_score": 0.3138}
{"id": "947af38b39c5d4f1064239c6549c7181cc287e1c3a68090e3fd5c01b095e0e60", "sources": ["arxiv", "semantic_scholar"], "title": "ZipIt! Merging Models from Different Tasks without Training", "abstract": "Typical deep visual recognition models are capable of performing the one task they were trained on. In this paper, we tackle the extremely difficult problem of combining distinct models with different initializations, each solving a separate task, into one multi-task model without any additional training. Prior work in model merging permutes one model to the space of the other then averages them together. While this works for models trained on the same task, we find that this fails to account for the differences in models trained on disjoint tasks. Thus, we introduce \"ZipIt!\", a general method for merging two arbitrary models of the same architecture that incorporates two simple strategies. First, in order to account for features that aren't shared between models, we expand the model merging problem to allow for merging features within each model by defining a general \"zip\" operation. Second, we add support for partially zipping the models up until a specified layer, naturally creating a multi-head model. We find that these two changes combined account for 20-60% improvement over prior work, making it more feasible to merge models trained on disjoint tasks without retraining.", "authors": ["George Stoica", "Daniel Bolya", "Jakob Bjorner", "Pratik Ramesh", "Taylor Hearn", "Judy Hoffman"], "categories": ["cs.CV", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-05-04", "url": "https://arxiv.org/abs/2305.03053", "pdf_url": "https://arxiv.org/pdf/2305.03053v3", "arxiv_id": "2305.03053", "doi": "10.48550/arXiv.2305.03053", "citation_count": 195, "influential_citation_count": 19, "has_code": false, "code_url": null, "venue": "International Conference on Learning Representations", "quality_score": 0.6505}
{"id": "088e45e71740d3b664934a2f93796c41ba02689c97245e1f9e0541c2ef39e815", "sources": ["arxiv", "semantic_scholar"], "title": "Predict NAS Multi-Task by Stacking Ensemble Models using GP-NAS", "abstract": "Accurately predicting the performance of architecture with small sample training is an important but not easy task. How to analysis and train dataset to overcome overfitting is the core problem we should deal with. Meanwhile if there is the mult-task problem, we should also think about if we can take advantage of their correlation and estimate as fast as we can. In this track, Super Network builds a search space based on ViT-Base. The search space contain depth, num-heads, mpl-ratio and embed-dim. What we done firstly are pre-processing the data based on our understanding of this problem which can reduce complexity of problem and probability of over fitting. Then we tried different kind of models and different way to combine them. Finally we choose stacking ensemble models using GP-NAS with cross validation. Our stacking model ranked 1st in CVPR 2022 Track 2 Challenge.", "authors": ["Ke Zhang"], "categories": ["cs.LG", "cs.CV", "stat.AP", "stat.CO"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2023-05-02", "url": "https://arxiv.org/abs/2305.01667", "pdf_url": "https://arxiv.org/pdf/2305.01667v1", "arxiv_id": "2305.01667", "doi": "10.48550/arXiv.2305.01667", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0}
{"id": "a0a3232939aed41524a12267dac1b66f5754adc771eff0e4b0a4840c328471d6", "sources": ["arxiv", "semantic_scholar"], "title": "An Empirical Study of Multimodal Model Merging", "abstract": "Model merging (e.g., via interpolation or task arithmetic) fuses multiple models trained on different tasks to generate a multi-task solution. The technique has been proven successful in previous studies, where the models are trained on similar tasks and with the same initialization. In this paper, we expand on this concept to a multimodal setup by merging transformers trained on different modalities. Furthermore, we conduct our study for a novel goal where we can merge vision, language, and cross-modal transformers of a modality-specific architecture to create a parameter-efficient modality-agnostic architecture. Through comprehensive experiments, we systematically investigate the key factors impacting model performance after merging, including initialization, merging mechanisms, and model architectures. We also propose two metrics that assess the distance between weights to be merged and can serve as an indicator of the merging outcomes. Our analysis leads to an effective training recipe for matching the performance of the modality-agnostic baseline (i.e., pre-trained from scratch) via model merging. Our method also outperforms naive merging significantly on various tasks, with improvements of 3% on VQA, 7% on COCO retrieval, 25% on NLVR2, 14% on Flickr30k and 3% on ADE20k. Our code is available at https://github.com/ylsung/vl-merging", "authors": ["Yi-Lin Sung", "Linjie Li", "Kevin Lin", "Zhe Gan", "Mohit Bansal", "Lijuan Wang"], "categories": ["cs.CV", "cs.AI", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-04-28", "url": "https://arxiv.org/abs/2304.14933", "pdf_url": "https://arxiv.org/pdf/2304.14933v2", "arxiv_id": "2304.14933", "doi": "10.48550/arXiv.2304.14933", "citation_count": 57, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/ylsung/vl-merging", "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.4409}
{"id": "a8abd931ab3d779ef8a4fcba309820cfe86a4e1a9412aa7b71918bf7aa1a1156", "sources": ["arxiv", "semantic_scholar"], "title": "HuaTuo: Tuning LLaMA Model with Chinese Medical Knowledge", "abstract": "Large Language Models (LLMs), such as the LLaMA model, have demonstrated their effectiveness in various general-domain natural language processing (NLP) tasks. Nevertheless, LLMs have not yet performed optimally in biomedical domain tasks due to the need for medical expertise in the responses. In response to this challenge, we propose HuaTuo, a LLaMA-based model that has been supervised-fine-tuned with generated QA (Question-Answer) instances. The experimental results demonstrate that HuaTuo generates responses that possess more reliable medical knowledge. Our proposed HuaTuo model is accessible at https://github.com/SCIR-HI/Huatuo-Llama-Med-Chinese.", "authors": ["Haochun Wang", "Chi Liu", "Nuwa Xi", "Zewen Qiang", "Sendong Zhao", "Bing Qin", "Ting Liu"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-04-14", "url": "https://arxiv.org/abs/2304.06975", "pdf_url": "https://arxiv.org/pdf/2304.06975v1", "arxiv_id": "2304.06975", "doi": "10.48550/arXiv.2304.06975", "citation_count": 290, "influential_citation_count": 25, "has_code": true, "code_url": "https://github.com/SCIR-HI/Huatuo-Llama-Med-Chinese", "venue": "arXiv.org", "quality_score": 0.7075}
{"id": "05043c4892f1a9f7493c0a4f01f2d58b449fa9b8727ac4c843f3043dbe31e1ea", "sources": ["arxiv", "semantic_scholar"], "title": "Adjust factor with volatility model using MAXFLAT low-pass filter and construct portfolio in China A share market", "abstract": "In the field of quantitative finance, volatility models, such as ARCH, GARCH, FIGARCH, SV, EWMA, play the key role in risk and portfolio management. Meanwhile, factor investing is more and more famous since mid of 20 century. CAPM, Fama French three factor model, Fama French five-factor model, MSCI Barra factor model are mentioned and developed during this period. In this paper, we will show why we need adjust group of factors by our MAXFLAT low-pass volatility model. All of our experiments are under China's CSI 300 and CSI 500 universe which represent China's large cap stocks and mid-small cap stocks. Our result shows adjust factors by MAXFLAT volatility model have better performance in both large cap and small cap universe than original factors or other risk adjust factors in China A share. Also the portfolio constructed by MAXFLAT risk adjust factors have continuous excess return and lower beta compare with benchmark index.", "authors": ["Ke Zhang"], "categories": ["q-fin.RM", "q-fin.ST"], "fields_of_study": ["Economics"], "published_date": "2023-03-29", "url": "https://arxiv.org/abs/2304.04676", "pdf_url": "https://arxiv.org/pdf/2304.04676v2", "arxiv_id": "2304.04676", "doi": null, "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.0753}
{"id": "36a257c08aa10dfefe4939d4cdd4a01b23eea6ad4581143a61ec24473191c7bd", "sources": ["arxiv", "semantic_scholar"], "title": "Merging Decision Transformers: Weight Averaging for Forming Multi-Task Policies", "abstract": "Recent work has shown the promise of creating generalist, transformer-based, models for language, vision, and sequential decision-making problems. To create such models, we generally require centralized training objectives, data, and compute. It is of interest if we can more flexibly create generalist policies by merging together multiple, task-specific, individually trained policies. In this work, we take a preliminary step in this direction through merging, or averaging, subsets of Decision Transformers in parameter space trained on different MuJoCo locomotion problems, forming multi-task models without centralized training. We also demonstrate the importance of various methodological choices when merging policies, such as utilizing common pre-trained initializations, increasing model capacity, and utilizing Fisher information for weighting parameter importance. In general, we believe research in this direction could help democratize and distribute the process that forms multi-task robotics policies. Our implementation is available at https://github.com/daniellawson9999/merging-decision-transformers.", "authors": ["Daniel Lawson", "Ahmed H. Qureshi"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-03-14", "url": "https://arxiv.org/abs/2303.07551", "pdf_url": "https://arxiv.org/pdf/2303.07551v3", "arxiv_id": "2303.07551", "doi": "10.1109/ICRA57147.2024.10610919", "citation_count": 15, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/daniellawson9999/merging-decision-transformers", "venue": "IEEE International Conference on Robotics and Automation", "quality_score": 0.301}
{"id": "0731e37079725049d04fb30f9c08fbdf15dc29a128da801b1852ecdab8654b8e", "sources": ["arxiv", "semantic_scholar"], "title": "Toward Defining a Domain Complexity Measure Across Domains", "abstract": "Artificial Intelligence (AI) systems planned for deployment in real-world applications frequently are researched and developed in closed simulation environments where all variables are controlled and known to the simulator or labeled benchmark datasets are used. Transition from these simulators, testbeds, and benchmark datasets to more open-world domains poses significant challenges to AI systems, including significant increases in the complexity of the domain and the inclusion of real-world novelties; the open-world environment contains numerous out-of-distribution elements that are not part in the AI systems' training set. Here, we propose a path to a general, domain-independent measure of domain complexity level. We distinguish two aspects of domain complexity: intrinsic and extrinsic. The intrinsic domain complexity is the complexity that exists by itself without any action or interaction from an AI agent performing a task on that domain. This is an agent-independent aspect of the domain complexity. The extrinsic domain complexity is agent- and task-dependent. Intrinsic and extrinsic elements combined capture the overall complexity of the domain. We frame the components that define and impact domain complexity levels in a domain-independent light. Domain-independent measures of complexity could enable quantitative predictions of the difficulty posed to AI systems when transitioning from one testbed or environment to another, when facing out-of-distribution data in open-world tasks, and when navigating the rapidly expanding solution and search spaces encountered in open-world domains.", "authors": ["Katarina Doctor", "Christine Task", "Eric Kildebeck", "Mayank Kejriwal", "Lawrence Holder", "Russell Leong"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-03-07", "url": "https://arxiv.org/abs/2303.04141", "pdf_url": "https://arxiv.org/pdf/2303.04141v1", "arxiv_id": "2303.04141", "doi": "10.48550/arXiv.2303.04141", "citation_count": 10, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2603}
{"id": "7642ac038b3bc81b3b6c1b1ea1d0d0856133425cc041d6bdcd7b62946d2cc230", "sources": ["arxiv", "semantic_scholar"], "title": "Hitachi at SemEval-2023 Task 3: Exploring Cross-lingual Multi-task Strategies for Genre and Framing Detection in Online News", "abstract": "This paper explains the participation of team Hitachi to SemEval-2023 Task 3 \"Detecting the genre, the framing, and the persuasion techniques in online news in a multi-lingual setup.'' Based on the multilingual, multi-task nature of the task and the low-resource setting, we investigated different cross-lingual and multi-task strategies for training the pretrained language models. Through extensive experiments, we found that (a) cross-lingual/multi-task training, and (b) collecting an external balanced dataset, can benefit the genre and framing detection. We constructed ensemble models from the results and achieved the highest macro-averaged F1 scores in Italian and Russian genre categorization subtasks.", "authors": ["Yuta Koreeda", "Ken-ichi Yokote", "Hiroaki Ozaki", "Atsuki Yamaguchi", "Masaya Tsunokake", "Yasuhiro Sogawa"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-03-03", "url": "https://arxiv.org/abs/2303.01794", "pdf_url": "https://arxiv.org/pdf/2303.01794v2", "arxiv_id": "2303.01794", "doi": "10.18653/v1/2023.semeval-1.237", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Workshop on Semantic Evaluation", "quality_score": 0.1505}
{"id": "8726ddc7761ff438489e794d0f1ddfa171f7211f5fc0e1a28712e0ae622a8acf", "sources": ["arxiv", "semantic_scholar"], "title": "Seasoning Model Soups for Robustness to Adversarial and Natural Distribution Shifts", "abstract": "Adversarial training is widely used to make classifiers robust to a specific threat or adversary, such as $\\ell_p$-norm bounded perturbations of a given $p$-norm. However, existing methods for training classifiers robust to multiple threats require knowledge of all attacks during training and remain vulnerable to unseen distribution shifts. In this work, we describe how to obtain adversarially-robust model soups (i.e., linear combinations of parameters) that smoothly trade-off robustness to different $\\ell_p$-norm bounded adversaries. We demonstrate that such soups allow us to control the type and level of robustness, and can achieve robustness to all threats without jointly training on all of them. In some cases, the resulting model soups are more robust to a given $\\ell_p$-norm adversary than the constituent model specialized against that same adversary. Finally, we show that adversarially-robust model soups can be a viable tool to adapt to distribution shifts from a few examples.", "authors": ["Francesco Croce", "Sylvestre-Alvise Rebuffi", "Evan Shelhamer", "Sven Gowal"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2023-02-20", "url": "https://arxiv.org/abs/2302.10164", "pdf_url": "https://arxiv.org/pdf/2302.10164v1", "arxiv_id": "2302.10164", "doi": "10.1109/CVPR52729.2023.01185", "citation_count": 23, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "Computer Vision and Pattern Recognition", "quality_score": 0.3451}
{"id": "60a8eb09320628f095ae2ce7a720d9ef1231ca8665bb3576e769ab6bb1c6261f", "sources": ["arxiv", "semantic_scholar"], "title": "Zilber's notion of logically perfect structure: Universal Covers", "abstract": "We sketch recent interactions between model theory and a roughly 150-year old study of analytic functions involving complex analysis, algebraic topology, and number theory, centered in canonicity of universal covers. Towards this goal we discuss in a systematic and unified way several examples indicating the main ideas of the proofs and the necessary changes in method for different situations: exponential covers, modular and Shimura curves, Shimura and abelian varieties, and coherent families of smooth covers.", "authors": ["John T. Baldwin", "Andrés Villaveces"], "categories": ["math.LO"], "fields_of_study": ["Mathematics"], "published_date": "2023-02-09", "url": "https://arxiv.org/abs/2302.04650", "pdf_url": "https://arxiv.org/pdf/2302.04650v3", "arxiv_id": "2302.04650", "doi": "10.2140/mt.2024.3.647", "citation_count": 7, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Model Th. 3 (2024) 647-683", "quality_score": 0.2258}
{"id": "16776bc8de2ca19bcb0882e78081e34a989d813fc5e8325d2e6ec73b1b976eda", "sources": ["arxiv", "semantic_scholar"], "title": "Backward Compatibility During Data Updates by Weight Interpolation", "abstract": "Backward compatibility of model predictions is a desired property when updating a machine learning driven application. It allows to seamlessly improve the underlying model without introducing regression bugs. In classification tasks these bugs occur in the form of negative flips. This means an instance that was correctly classified by the old model is now classified incorrectly by the updated model. This has direct negative impact on the user experience of such systems e.g. a frequently used voice assistant query is suddenly misclassified. A common reason to update the model is when new training data becomes available and needs to be incorporated. Simply retraining the model with the updated data introduces the unwanted negative flips. We study the problem of regression during data updates and propose Backward Compatible Weight Interpolation (BCWI). This method interpolates between the weights of the old and new model and we show in extensive experiments that it reduces negative flips without sacrificing the improved accuracy of the new model. BCWI is straight forward to implement and does not increase inference cost. We also explore the use of importance weighting during interpolation and averaging the weights of multiple new models in order to further reduce negative flips.", "authors": ["Raphael Schumann", "Elman Mansimov", "Yi-An Lai", "Nikolaos Pappas", "Xibin Gao", "Yi Zhang"], "categories": ["cs.LG", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-01-25", "url": "https://arxiv.org/abs/2301.10546", "pdf_url": "https://arxiv.org/pdf/2301.10546v1", "arxiv_id": "2301.10546", "doi": "10.48550/arXiv.2301.10546", "citation_count": 7, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Conference of the European Chapter of the Association for Computational Linguistics", "quality_score": 0.2258}
{"id": "5612f4b4ebf284d6b2ac1bbcde9390a4ecf018a47ca2bb4b11dc879d552f0dd2", "sources": ["arxiv", "semantic_scholar"], "title": "Dataless Knowledge Fusion by Merging Weights of Language Models", "abstract": "Fine-tuning pre-trained language models has become the prevalent paradigm for building downstream NLP models. Oftentimes fine-tuned models are readily available but their training data is not, due to data privacy or intellectual property concerns. This creates a barrier to fusing knowledge across individual models to yield a better single model. In this paper, we study the problem of merging individual models built on different training data sets to obtain a single model that performs well both across all data set domains and can generalize on out-of-domain data. We propose a dataless knowledge fusion method that merges models in their parameter space, guided by weights that minimize prediction differences between the merged model and the individual models. Over a battery of evaluation settings, we show that the proposed method significantly outperforms baselines such as Fisher-weighted averaging or model ensembling. Further, we find that our method is a promising alternative to multi-task learning that can preserve or sometimes improve over the individual models without access to the training data. Finally, model merging is more efficient than training a multi-task model, thus making it applicable to a wider set of scenarios.", "authors": ["Xisen Jin", "Xiang Ren", "Daniel Preotiuc-Pietro", "Pengxiang Cheng"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2022-12-19", "url": "https://arxiv.org/abs/2212.09849", "pdf_url": "https://arxiv.org/pdf/2212.09849v6", "arxiv_id": "2212.09849", "doi": "10.48550/arXiv.2212.09849", "citation_count": 388, "influential_citation_count": 69, "has_code": true, "code_url": "https://github.com/bloomberg/dataless-model-merging", "venue": "International Conference on Learning Representations", "quality_score": 0.9225}
{"id": "385817bdb1c35abf723aac3315f39e9ebe64c685ee7fb704299e9f4693357e6b", "sources": ["arxiv", "semantic_scholar"], "title": "Editing Models with Task Arithmetic", "abstract": "Changing how pre-trained models behave -- e.g., improving their performance on a downstream task or mitigating biases learned during pre-training -- is a common practice when developing machine learning systems. In this work, we propose a new paradigm for steering the behavior of neural networks, centered around \\textit{task vectors}. A task vector specifies a direction in the weight space of a pre-trained model, such that movement in that direction improves performance on the task. We build task vectors by subtracting the weights of a pre-trained model from the weights of the same model after fine-tuning on a task. We show that these task vectors can be modified and combined together through arithmetic operations such as negation and addition, and the behavior of the resulting model is steered accordingly. Negating a task vector decreases performance on the target task, with little change in model behavior on control tasks. Moreover, adding task vectors together can improve performance on multiple tasks at once. Finally, when tasks are linked by an analogy relationship of the form ``A is to B as C is to D\", combining task vectors from three of the tasks can improve performance on the fourth, even when no data from the fourth task is used for training. Overall, our experiments with several models, modalities and tasks show that task arithmetic is a simple, efficient and effective way of editing models.", "authors": ["Gabriel Ilharco", "Marco Tulio Ribeiro", "Mitchell Wortsman", "Suchin Gururangan", "Ludwig Schmidt", "Hannaneh Hajishirzi", "Ali Farhadi"], "categories": ["cs.LG", "cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2022-12-08", "url": "https://arxiv.org/abs/2212.04089", "pdf_url": "https://arxiv.org/pdf/2212.04089v3", "arxiv_id": "2212.04089", "doi": "10.48550/arXiv.2212.04089", "citation_count": 1041, "influential_citation_count": 279, "has_code": false, "code_url": null, "venue": "International Conference on Learning Representations", "quality_score": 1.0}
{"id": "b00efef6a6fc88cfc94f229aca06d4c66cf649f4e3e3fa544d8a51d2a1e85044", "sources": ["arxiv", "semantic_scholar"], "title": "Task-Driven Hybrid Model Reduction for Dexterous Manipulation", "abstract": "In contact-rich tasks, like dexterous manipulation, the hybrid nature of making and breaking contact creates challenges for model representation and control. For example, choosing and sequencing contact locations for in-hand manipulation, where there are thousands of potential hybrid modes, is not generally tractable. In this paper, we are inspired by the observation that far fewer modes are actually necessary to accomplish many tasks. Building on our prior work learning hybrid models, represented as linear complementarity systems, we find a reduced-order hybrid model requiring only a limited number of task-relevant modes. This simplified representation, in combination with model predictive control, enables real-time control yet is sufficient for achieving high performance. We demonstrate the proposed method first on synthetic hybrid systems, reducing the mode count by multiple orders of magnitude while achieving task performance loss of less than 5%. We also apply the proposed method to a three-fingered robotic hand manipulating a previously unknown object. With no prior knowledge, we achieve state-of-the-art closed-loop performance within a few minutes of online learning, by collecting only a few thousand environment samples.", "authors": ["Wanxin Jin", "Michael Posa"], "categories": ["cs.RO", "eess.SY"], "fields_of_study": ["Computer Science", "Engineering"], "published_date": "2022-11-30", "url": "https://arxiv.org/abs/2211.16657", "pdf_url": "https://arxiv.org/pdf/2211.16657v2", "arxiv_id": "2211.16657", "doi": "10.1109/TRO.2024.3359531", "citation_count": 21, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/wanxinjin/Task-Driven-Hybrid-Reduction", "venue": "IEEE Transactions on robotics", "quality_score": 0.3356}
{"id": "664a87f57af3cd5d180d4ee81431f7a09d03fae9adf2b4f1c1c7b86e66ce3aef", "sources": ["arxiv", "semantic_scholar"], "title": "Modelling COVID-19-III: endemic spread in India", "abstract": "A disease in a given population is termed endemic when it exhibits a steady prevalence. We address the pertinent question as to what extent COVID-19 has turned endemic in India. There are several existing models for studying endemic behaviour, such as the extensions of the traditional temporal SIR model or the spatio-temporal endemic-epidemic model of Held et al. (2005) and its extensions. We propose a \"spatio-temporal Gravity model\" in a state of the art generalised linear model set up that can be deployed at various spatial resolutions. In absence of routine and quality covariates in the context of COVID-19 at finer spatial scales, we make use of extraneous covariates like air-traffic passenger count that enables us to capture the local mobility and social interactions effectively. This makes the proposed model different from the existing models. The proposed gravity model not only produces consistent estimators, but also outperforms the other models when applied to Indian COVID-19 data.", "authors": ["Madhuchhanda Bhattacharjee", "Arup Bose"], "categories": ["stat.AP"], "fields_of_study": ["Mathematics"], "published_date": "2022-11-11", "url": "https://arxiv.org/abs/2211.06215", "pdf_url": "https://arxiv.org/pdf/2211.06215v1", "arxiv_id": "2211.06215", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.0}
{"id": "8248004dd3f126e04246c8c4e128c2453f2940ee29c04b0224804fb896241438", "sources": ["arxiv", "semantic_scholar"], "title": "Momentum-based Weight Interpolation of Strong Zero-Shot Models for Continual Learning", "abstract": "Large pre-trained, zero-shot capable models have shown considerable success both for standard transfer and adaptation tasks, with particular robustness towards distribution shifts. In addition, subsequent fine-tuning can considerably improve performance on a selected downstream task. However, through naive fine-tuning, these zero-shot models lose their generalizability and robustness towards distribution shifts. This is a particular problem for tasks such as Continual Learning (CL), where continuous adaptation has to be performed as new task distributions are introduced sequentially. In this work, we showcase that where fine-tuning falls short to adapt such zero-shot capable models, simple momentum-based weight interpolation can provide consistent improvements for CL tasks in both memory-free and memory-based settings. In particular, we find improvements of over $+4\\%$ on standard CL benchmarks, while reducing the error to the upper limit of jointly training on all tasks at once in parts by more than half, allowing the continual learner to inch closer to the joint training limits.", "authors": ["Zafir Stojanovski", "Karsten Roth", "Zeynep Akata"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2022-11-06", "url": "https://arxiv.org/abs/2211.03186", "pdf_url": "https://arxiv.org/pdf/2211.03186v1", "arxiv_id": "2211.03186", "doi": "10.48550/arXiv.2211.03186", "citation_count": 19, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3253}
{"id": "2e7aefcc61141b7d9ec6d2b493c3fb26f50645a1a99342c1abac51eff834fa28", "sources": ["arxiv", "semantic_scholar"], "title": "Where to start? Analyzing the potential value of intermediate models", "abstract": "Previous studies observed that finetuned models may be better base models than the vanilla pretrained model. Such a model, finetuned on some source dataset, may provide a better starting point for a new finetuning process on a desired target dataset. Here, we perform a systematic analysis of this intertraining scheme, over a wide range of English classification tasks. Surprisingly, our analysis suggests that the potential intertraining gain can be analyzed independently for the target dataset under consideration, and for a base model being considered as a starting point. This is in contrast to current perception that the alignment between the target dataset and the source dataset used to generate the base model is a major factor in determining intertraining success. We analyze different aspects that contribute to each. Furthermore, we leverage our analysis to propose a practical and efficient approach to determine if and how to select a base model in real-world settings. Last, we release an updating ranking of best models in the HuggingFace hub per architecture https://ibm.github.io/model-recycling/.", "authors": ["Leshem Choshen", "Elad Venezian", "Shachar Don-Yehia", "Noam Slonim", "Yoav Katz"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2022-10-31", "url": "https://arxiv.org/abs/2211.00107", "pdf_url": "https://arxiv.org/pdf/2211.00107v3", "arxiv_id": "2211.00107", "doi": "10.48550/arXiv.2211.00107", "citation_count": 30, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.3728}
{"id": "dc86c27cffd75f86396f7e7c59ab719309678c0a3cf9c4c049fe04002b0ae906", "sources": ["arxiv", "semantic_scholar"], "title": "Higher internal covers", "abstract": "We define and study a higher-dimensional version of model theoretic internality, and relate it to higher-dimensional definable groupoids in the base theory.", "authors": ["Moshe Kamensky"], "categories": ["math.LO", "math.CT"], "fields_of_study": ["Mathematics"], "published_date": "2022-10-06", "url": "https://arxiv.org/abs/2210.02699", "pdf_url": "https://arxiv.org/pdf/2210.02699v2", "arxiv_id": "2210.02699", "doi": "10.2140/mt.2023.2.449", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Model Th. 2 (2023) 449-479", "quality_score": 0.0}
{"id": "b0b121dda62409ac21c90c941532e2c08311f9e0f0f701d2140e50254f2ec97f", "sources": ["arxiv", "semantic_scholar"], "title": "Fractional Gagliardo-Nirenberg interpolation inequality and bounded mean oscillation", "abstract": "We prove Gagliardo-Nirenberg interpolation inequalities estimating the Sobolev semi-norm in terms of the bounded mean oscillation semi-norm and a Sobolev semi-norm, with some of the Sobolev semi-norms having fractional order.", "authors": ["Jean Van Schaftingen"], "categories": ["math.CA"], "fields_of_study": ["Mathematics"], "published_date": "2022-08-31", "url": "https://arxiv.org/abs/2208.14691", "pdf_url": "https://arxiv.org/pdf/2208.14691v3", "arxiv_id": "2208.14691", "doi": "10.5802/crmath.463", "citation_count": 7, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Comptes rendus. Mathematique", "quality_score": 0.2258}
{"id": "209a14ff922a2c84ba3ac9661679d1c0312bedc75f0f6c286cc050a8d4ac810b", "sources": ["arxiv", "semantic_scholar"], "title": "Patching open-vocabulary models by interpolating weights", "abstract": "Open-vocabulary models like CLIP achieve high accuracy across many image classification tasks. However, there are still settings where their zero-shot performance is far from optimal. We study model patching, where the goal is to improve accuracy on specific tasks without degrading accuracy on tasks where performance is already adequate. Towards this goal, we introduce PAINT, a patching method that uses interpolations between the weights of a model before fine-tuning and the weights after fine-tuning on a task to be patched. On nine tasks where zero-shot CLIP performs poorly, PAINT increases accuracy by 15 to 60 percentage points while preserving accuracy on ImageNet within one percentage point of the zero-shot model. PAINT also allows a single model to be patched on multiple tasks and improves with model scale. Furthermore, we identify cases of broad transfer, where patching on one task increases accuracy on other tasks even when the tasks have disjoint classes. Finally, we investigate applications beyond common benchmarks such as counting or reducing the impact of typographic attacks on CLIP. Our findings demonstrate that it is possible to expand the set of tasks on which open-vocabulary models achieve high accuracy without re-training them from scratch.", "authors": ["Gabriel Ilharco", "Mitchell Wortsman", "Samir Yitzhak Gadre", "Shuran Song", "Hannaneh Hajishirzi", "Simon Kornblith", "Ali Farhadi", "Ludwig Schmidt"], "categories": ["cs.CV", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2022-08-10", "url": "https://arxiv.org/abs/2208.05592", "pdf_url": "https://arxiv.org/pdf/2208.05592v2", "arxiv_id": "2208.05592", "doi": "10.48550/arXiv.2208.05592", "citation_count": 215, "influential_citation_count": 38, "has_code": false, "code_url": null, "venue": "Neural Information Processing Systems", "quality_score": 0.7955}
{"id": "ff130d892375d56be05f79909fc7e00952cfd2a5c0922e2b053f35c4bf061a8b", "sources": ["arxiv", "semantic_scholar"], "title": "Understanding Weight Similarity of Neural Networks via Chain Normalization Rule and Hypothesis-Training-Testing", "abstract": "We present a weight similarity measure method that can quantify the weight similarity of non-convex neural networks. To understand the weight similarity of different trained models, we propose to extract the feature representation from the weights of neural networks. We first normalize the weights of neural networks by introducing a chain normalization rule, which is used for weight representation learning and weight similarity measure. We extend the traditional hypothesis-testing method to a hypothesis-training-testing statistical inference method to validate the hypothesis on the weight similarity of neural networks. With the chain normalization rule and the new statistical inference, we study the weight similarity measure on Multi-Layer Perceptron (MLP), Convolutional Neural Network (CNN), and Recurrent Neural Network (RNN), and find that the weights of an identical neural network optimized with the Stochastic Gradient Descent (SGD) algorithm converge to a similar local solution in a metric space. The weight similarity measure provides more insight into the local solutions of neural networks. Experiments on several datasets consistently validate the hypothesis of weight similarity measure.", "authors": ["Guangcong Wang", "Guangrun Wang", "Wenqi Liang", "Jianhuang Lai"], "categories": ["cs.LG", "cs.CV", "math.ST", "stat.ML"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2022-08-08", "url": "https://arxiv.org/abs/2208.04369", "pdf_url": "https://arxiv.org/pdf/2208.04369v1", "arxiv_id": "2208.04369", "doi": "10.48550/arXiv.2208.04369", "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1945}
{"id": "8a79b13f689f78e400c93031b3d7e6719043e21a2972269548090db20c3b28c6", "sources": ["arxiv", "semantic_scholar"], "title": "On the Usability of Transformers-based models for a French Question-Answering task", "abstract": "For many tasks, state-of-the-art results have been achieved with Transformer-based architectures, resulting in a paradigmatic shift in practices from the use of task-specific architectures to the fine-tuning of pre-trained language models. The ongoing trend consists in training models with an ever-increasing amount of data and parameters, which requires considerable resources. It leads to a strong search to improve resource efficiency based on algorithmic and hardware improvements evaluated only for English. This raises questions about their usability when applied to small-scale learning problems, for which a limited amount of training data is available, especially for under-resourced languages tasks. The lack of appropriately sized corpora is a hindrance to applying data-driven and transfer learning-based approaches with strong instability cases. In this paper, we establish a state-of-the-art of the efforts dedicated to the usability of Transformer-based models and propose to evaluate these improvements on the question-answering performances of French language which have few resources. We address the instability relating to data scarcity by investigating various training strategies with data augmentation, hyperparameters optimization and cross-lingual transfer. We also introduce a new compact model for French FrALBERT which proves to be competitive in low-resource settings.", "authors": ["Oralie Cattan", "Christophe Servan", "Sophie Rosset"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2022-07-19", "url": "https://arxiv.org/abs/2207.09150", "pdf_url": "https://arxiv.org/pdf/2207.09150v1", "arxiv_id": "2207.09150", "doi": "10.26615/978-954-452-072-4_029", "citation_count": 15, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Recent Advances in Natural Language Processing", "quality_score": 0.301}
{"id": "fa6a6521d44fe13d6307d030e24a9a86f209823f2a589f2ed59826e6ab4229af", "sources": ["arxiv", "semantic_scholar"], "title": "A SIQRB delayed model for cholera and optimal control treatment", "abstract": "We improve a recent mathematical model for cholera by adding a time delay that represents the time between the instant at which an individual becomes infected and the instant at which he begins to have symptoms of cholera disease. We prove that the delayed cholera model is biologically meaningful and analyze the local asymptotic stability of the equilibrium points for positive time delays. An optimal control problem is proposed and analyzed, where the goal is to obtain optimal treatment strategies, through quarantine, that minimize the number of infective individuals and the bacterial concentration, as well as treatment costs. Necessary optimality conditions are applied to the delayed optimal control problem, with a $L^1$ type cost functional. We show that the delayed cholera model fits better the cholera outbreak that occurred in the Department of Artibonite -- Haiti, from 1 November 2010 to 1 May 2011, than the non-delayed model. Considering the data of the cholera outbreak in Haiti, we solve numerically the delayed optimal control problem and propose solutions for the outbreak control and eradication.", "authors": ["Ana P. Lemos-Paiao", "Helmut Maurer", "Cristiana J. Silva", "Delfim F. M. Torres"], "categories": ["math.OC", "q-bio.PE"], "fields_of_study": ["Mathematics", "Biology"], "published_date": "2022-06-25", "url": "https://arxiv.org/abs/2206.12688", "pdf_url": "https://arxiv.org/pdf/2206.12688v1", "arxiv_id": "2206.12688", "doi": "10.1051/mmnp/2022027", "citation_count": 14, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Mathematical Modelling of Natural Phenomena", "quality_score": 0.294}
{"id": "fb3be48833eea97af148d4982b502c8b89c5f1d155f2254c30816a1e37c07a7e", "sources": ["arxiv", "semantic_scholar"], "title": "Classical Aspects of a Distributional 3+1 Foam Model", "abstract": "A 3+1 spacetime, with a shift vector that is the unique fundamental solution to the linearized wave operator, is introduced to model an interpretation of Wheeler's layman's analogy of the Quantum foam. To understand the distributional aspects of this model is the guaranteed existence of a sequence of compactly supported shift vectors that converge to the fundamental solution used to introduce a sequence of 3+1 globally hyperbolic spacetimes. Using the sequence of these causally stable spacetimes it is shown that there exists a positive integer such that for all elements in the sequence with a greater index value than this integer and for any Eulerian observer will the shift vector increase more rapidly than any polynomial and the volume expansion is more rapid than a polynomial in all directions. The same conclusion remains valid for the trace of the extrinsic curvature. Nonetheless, it is shown, no matter how volatile the extrinsic curvature is for these elements there also exists elements in the other end of the sequence where the extrinsic curvature is negligible and the spacetime flat.", "authors": ["Claes Cramer"], "categories": ["gr-qc", "math-ph", "quant-ph"], "fields_of_study": ["Physics", "Mathematics"], "published_date": "2022-06-21", "url": "https://arxiv.org/abs/2206.10417", "pdf_url": "https://arxiv.org/pdf/2206.10417v11", "arxiv_id": "2206.10417", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.0}
{"id": "e8dc77fa2f412f46222a98d4cad80854faf2f6c7cc6941e94f3e8cf7537af00c", "sources": ["arxiv", "semantic_scholar"], "title": "UMass PCL at SemEval-2022 Task 4: Pre-trained Language Model Ensembles for Detecting Patronizing and Condescending Language", "abstract": "Patronizing and condescending language (PCL) is everywhere, but rarely is the focus on its use by media towards vulnerable communities. Accurately detecting PCL of this form is a difficult task due to limited labeled data and how subtle it can be. In this paper, we describe our system for detecting such language which was submitted to SemEval 2022 Task 4: Patronizing and Condescending Language Detection. Our approach uses an ensemble of pre-trained language models, data augmentation, and optimizing the threshold for detection. Experimental results on the evaluation dataset released by the competition hosts show that our work is reliably able to detect PCL, achieving an F1 score of 55.47% on the binary classification task and a macro F1 score of 36.25% on the fine-grained, multi-label detection task.", "authors": ["David Koleczek", "Alex Scarlatos", "Siddha Karakare", "Preshma Linet Pereira"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2022-04-18", "url": "https://arxiv.org/abs/2204.08304", "pdf_url": "https://arxiv.org/pdf/2204.08304v1", "arxiv_id": "2204.08304", "doi": "10.48550/arXiv.2204.08304", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Workshop on Semantic Evaluation", "quality_score": 0.0753}
{"id": "c17294dfc20d954ae895fc434ed6cdff85067bdba21dd67ff80f9351543a79ae", "sources": ["arxiv", "semantic_scholar"], "title": "HFL at SemEval-2022 Task 8: A Linguistics-inspired Regression Model with Data Augmentation for Multilingual News Similarity", "abstract": "This paper describes our system designed for SemEval-2022 Task 8: Multilingual News Article Similarity. We proposed a linguistics-inspired model trained with a few task-specific strategies. The main techniques of our system are: 1) data augmentation, 2) multi-label loss, 3) adapted R-Drop, 4) samples reconstruction with the head-tail combination. We also present a brief analysis of some negative methods like two-tower architecture. Our system ranked 1st on the leaderboard while achieving a Pearson's Correlation Coefficient of 0.818 on the official evaluation set.", "authors": ["Zihang Xu", "Ziqing Yang", "Yiming Cui", "Zhigang Chen"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2022-04-11", "url": "https://arxiv.org/abs/2204.04844", "pdf_url": "https://arxiv.org/pdf/2204.04844v1", "arxiv_id": "2204.04844", "doi": "10.48550/arXiv.2204.04844", "citation_count": 8, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "International Workshop on Semantic Evaluation", "quality_score": 0.2386}
{"id": "723e6d1888f5f93fdaf9d4bf96f8e528734ea47fb4d5755c357480de95f6abc7", "sources": ["arxiv", "semantic_scholar"], "title": "Long-Tailed Recognition via Weight Balancing", "abstract": "In the real open world, data tends to follow long-tailed class distributions, motivating the well-studied long-tailed recognition (LTR) problem. Naive training produces models that are biased toward common classes in terms of higher accuracy. The key to addressing LTR is to balance various aspects including data distribution, training losses, and gradients in learning. We explore an orthogonal direction, weight balancing, motivated by the empirical observation that the naively trained classifier has \"artificially\" larger weights in norm for common classes (because there exists abundant data to train them, unlike the rare classes). We investigate three techniques to balance weights, L2-normalization, weight decay, and MaxNorm. We first point out that L2-normalization \"perfectly\" balances per-class weights to be unit norm, but such a hard constraint might prevent classes from learning better classifiers. In contrast, weight decay penalizes larger weights more heavily and so learns small balanced weights; the MaxNorm constraint encourages growing small weights within a norm ball but caps all the weights by the radius. Our extensive study shows that both help learn balanced weights and greatly improve the LTR accuracy. Surprisingly, weight decay, although underexplored in LTR, significantly improves over prior work. Therefore, we adopt a two-stage training paradigm and propose a simple approach to LTR: (1) learning features using the cross-entropy loss by tuning weight decay, and (2) learning classifiers using class-balanced loss by tuning weight decay and MaxNorm. Our approach achieves the state-of-the-art accuracy on five standard benchmarks, serving as a future baseline for long-tailed recognition.", "authors": ["Shaden Alshammari", "Yu-Xiong Wang", "Deva Ramanan", "Shu Kong"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2022-03-27", "url": "https://arxiv.org/abs/2203.14197", "pdf_url": "https://arxiv.org/pdf/2203.14197v1", "arxiv_id": "2203.14197", "doi": "10.1109/CVPR52688.2022.00677", "citation_count": 196, "influential_citation_count": 28, "has_code": true, "code_url": "https://github.com/ShadeAlsha/LTR-weight-balancing", "venue": "Computer Vision and Pattern Recognition", "quality_score": 0.7312}
{"id": "ee63389abbb5e6afd52ce1bcc42f2ccb1eeac58164774d6395e3c69feced84ce", "sources": ["arxiv", "semantic_scholar"], "title": "Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time", "abstract": "The conventional recipe for maximizing model accuracy is to (1) train multiple models with various hyperparameters and (2) pick the individual model which performs best on a held-out validation set, discarding the remainder. In this paper, we revisit the second step of this procedure in the context of fine-tuning large pre-trained models, where fine-tuned models often appear to lie in a single low error basin. We show that averaging the weights of multiple models fine-tuned with different hyperparameter configurations often improves accuracy and robustness. Unlike a conventional ensemble, we may average many models without incurring any additional inference or memory costs -- we call the results \"model soups.\" When fine-tuning large pre-trained models such as CLIP, ALIGN, and a ViT-G pre-trained on JFT, our soup recipe provides significant improvements over the best model in a hyperparameter sweep on ImageNet. The resulting ViT-G model, which attains 90.94% top-1 accuracy on ImageNet, achieved a new state of the art. Furthermore, we show that the model soup approach extends to multiple image classification and natural language processing tasks, improves out-of-distribution performance, and improves zero-shot performance on new downstream tasks. Finally, we analytically relate the performance similarity of weight-averaging and logit-ensembling to flatness of the loss and confidence of the predictions, and validate this relation empirically. Code is available at https://github.com/mlfoundations/model-soups.", "authors": ["Mitchell Wortsman", "Gabriel Ilharco", "Samir Yitzhak Gadre", "Rebecca Roelofs", "Raphael Gontijo-Lopes", "Ari S. Morcos", "Hongseok Namkoong", "Ali Farhadi", "Yair Carmon", "Simon Kornblith", "Ludwig Schmidt"], "categories": ["cs.LG", "cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2022-03-10", "url": "https://arxiv.org/abs/2203.05482", "pdf_url": "https://arxiv.org/pdf/2203.05482v3", "arxiv_id": "2203.05482", "doi": "10.48550/arXiv.2203.05482", "citation_count": 1565, "influential_citation_count": 192, "has_code": true, "code_url": "https://github.com/mlfoundations/model-soups", "venue": "International Conference on Machine Learning", "quality_score": 1.0}
{"id": "e8f60c07e4b269ece44a7009a2e7e90aaffd5c4668b2c6b687324b511ef284fe", "sources": ["arxiv", "semantic_scholar"], "title": "Triple Motion Estimation and Frame Interpolation based on Adaptive Threshold for Frame Rate Up-Conversion", "abstract": "In this paper, we propose a novel motion-compensated frame rate up-conversion (MC-FRUC) algorithm. The proposed algorithm creates interpolated frames by first estimating motion vectors using unilateral (jointing forward and backward) and bilateral motion estimation. Then motion vectors are combined based on adaptive threshold, in order to creates high-quality interpolated frames and reduce block artifacts. Since motion-compensated frame interpolation along unilateral motion trajectories yields holes, a new algorithm is introduced to resolve this problem. The experimental results show that the quality of the interpolated frames using the proposed algorithm is much higher than the existing algorithms.", "authors": ["Hanieh Naderi", "Mohammad Rahmati"], "categories": ["eess.IV", "cs.AI", "cs.CV", "cs.MM"], "fields_of_study": ["Engineering", "Computer Science"], "published_date": "2022-03-05", "url": "https://arxiv.org/abs/2203.03621", "pdf_url": "https://arxiv.org/pdf/2203.03621v1", "arxiv_id": "2203.03621", "doi": "10.48550/arXiv.2203.03621", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0}
{"id": "4776b7a754e2e630469baf90d21410631a79bfc25ece756c1cad1e4081149ee8", "sources": ["arxiv", "semantic_scholar"], "title": "Model Reprogramming: Resource-Efficient Cross-Domain Machine Learning", "abstract": "In data-rich domains such as vision, language, and speech, deep learning prevails to deliver high-performance task-specific models and can even learn general task-agnostic representations for efficient finetuning to downstream tasks. However, deep learning in resource-limited domains still faces multiple challenges including (i) limited data, (ii) constrained model development cost, and (iii) lack of adequate pre-trained models for effective finetuning. This paper provides an overview of model reprogramming to bridge this gap. Model reprogramming enables resource-efficient cross-domain machine learning by repurposing and reusing a well-developed pre-trained model from a source domain to solve tasks in a target domain without model finetuning, where the source and target domains can be vastly different. In many applications, model reprogramming outperforms transfer learning and training from scratch. This paper elucidates the methodology of model reprogramming, summarizes existing use cases, provides a theoretical explanation of the success of model reprogramming, and concludes with a discussion on open-ended research questions and opportunities. A list of model reprogramming studies is actively maintained and updated at https://github.com/IBM/model-reprogramming.", "authors": ["Pin-Yu Chen"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2022-02-22", "url": "https://arxiv.org/abs/2202.10629", "pdf_url": "https://arxiv.org/pdf/2202.10629v4", "arxiv_id": "2202.10629", "doi": "10.1609/aaai.v38i20.30267", "citation_count": 87, "influential_citation_count": 4, "has_code": true, "code_url": "https://github.com/IBM/model-reprogramming", "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.4861}
{"id": "b7c17ad48ab990be9f90f51ebb8835b89e8b79dcda5a32e0120b067c43e87103", "sources": ["arxiv", "semantic_scholar"], "title": "FILM: Frame Interpolation for Large Motion", "abstract": "We present a frame interpolation algorithm that synthesizes multiple intermediate frames from two input images with large in-between motion. Recent methods use multiple networks to estimate optical flow or depth and a separate network dedicated to frame synthesis. This is often complex and requires scarce optical flow or depth ground-truth. In this work, we present a single unified network, distinguished by a multi-scale feature extractor that shares weights at all scales, and is trainable from frames alone. To synthesize crisp and pleasing frames, we propose to optimize our network with the Gram matrix loss that measures the correlation difference between feature maps. Our approach outperforms state-of-the-art methods on the Xiph large motion benchmark. We also achieve higher scores on Vimeo-90K, Middlebury and UCF101, when comparing to methods that use perceptual losses. We study the effect of weight sharing and of training with datasets of increasing motion range. Finally, we demonstrate our model's effectiveness in synthesizing high quality and temporally coherent videos on a challenging near-duplicate photos dataset. Codes and pre-trained models are available at https://film-net.github.io.", "authors": ["Fitsum Reda", "Janne Kontkanen", "Eric Tabellion", "Deqing Sun", "Caroline Pantofaru", "Brian Curless"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2022-02-10", "url": "https://arxiv.org/abs/2202.04901", "pdf_url": "https://arxiv.org/pdf/2202.04901v4", "arxiv_id": "2202.04901", "doi": "10.1007/978-3-031-20071-7_15", "citation_count": 238, "influential_citation_count": 47, "has_code": true, "code_url": "https://github.com/google-research/frame-interpolation", "venue": "European Conference on Computer Vision", "quality_score": 0.8406}
{"id": "ffee7bd98a6878025ec36e2fb397100cbcd6a9f86f689b16b752ad52a5147de7", "sources": ["arxiv", "semantic_scholar"], "title": "Global-Reasoned Multi-Task Learning Model for Surgical Scene Understanding", "abstract": "Global and local relational reasoning enable scene understanding models to perform human-like scene analysis and understanding. Scene understanding enables better semantic segmentation and object-to-object interaction detection. In the medical domain, a robust surgical scene understanding model allows the automation of surgical skill evaluation, real-time monitoring of surgeon's performance and post-surgical analysis. This paper introduces a globally-reasoned multi-task surgical scene understanding model capable of performing instrument segmentation and tool-tissue interaction detection. Here, we incorporate global relational reasoning in the latent interaction space and introduce multi-scale local (neighborhood) reasoning in the coordinate space to improve segmentation. Utilizing the multi-task model setup, the performance of the visual-semantic graph attention network in interaction detection is further enhanced through global reasoning. The global interaction space features from the segmentation module are introduced into the graph network, allowing it to detect interactions based on both node-to-node and global interaction reasoning. Our model reduces the computation cost compared to running two independent single-task models by sharing common modules, which is indispensable for practical applications. Using a sequential optimization technique, the proposed multi-task model outperforms other state-of-the-art single-task models on the MICCAI endoscopic vision challenge 2018 dataset. Additionally, we also observe the performance of the multi-task model when trained using the knowledge distillation technique. The official code implementation is made available in GitHub.", "authors": ["Lalithkumar Seenivasan", "Sai Mitheran", "Mobarakol Islam", "Hongliang Ren"], "categories": ["eess.IV", "cs.RO"], "fields_of_study": ["Engineering", "Computer Science"], "published_date": "2022-01-28", "url": "https://arxiv.org/abs/2201.11957", "pdf_url": "https://arxiv.org/pdf/2201.11957v1", "arxiv_id": "2201.11957", "doi": "10.1109/LRA.2022.3146544", "citation_count": 46, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/lalithjets/Global-reasoned-multi-task-model", "venue": "IEEE Robotics and Automation Letters", "quality_score": 0.418}
{"id": "aa0ffa9a8de37e85c387720b4480db74a94c91e5ffee730533ebcd63fa765b61", "sources": ["arxiv", "semantic_scholar"], "title": "Arithmetic Monodromy Groups of Dynamical Belyi maps", "abstract": "We consider a large family of dynamical Belyi maps of arbitrary degree and study the arithmetic monodromy groups attached to the iterates of such maps. Building on the results of Bouw-Ejder-Karemaker on the geometric monodromy groups of these maps, we show that the quotient of the arithmetic monodromy group by the geometric monodromy group has order either $1$ or $2$. Prior to this article, a result of this kind was only known for quadratic maps (Pink) and a few examples in degree $3$.", "authors": ["Ozlem Ejder"], "categories": ["math.NT"], "fields_of_study": ["Mathematics"], "published_date": "2022-01-22", "url": "https://arxiv.org/abs/2201.09005", "pdf_url": "https://arxiv.org/pdf/2201.09005v1", "arxiv_id": "2201.09005", "doi": null, "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1945}
{"id": "f94c08bf8846279809568f12e503970be7f6415d35dbb03100e3a47f56a06e39", "sources": ["arxiv", "semantic_scholar"], "title": "Arithmetic geometric model for the renormalisation of irrationally indifferent attractors", "abstract": "In this paper we build a geometric model for the renormalisation of irrationally indifferent fixed points. The geometric model incorporates the fine arithmetic properties of the rotation number at the fixed point. Using this model for the renormalisation, we build a topological model for the dynamics of a holomorphic map near an irrationally indifferent fixed point. Then, we explain the topology of the maximal invariant set for the model, and also explain the dynamics of the map on the maximal invariant set.", "authors": ["Davoud Cheraghi"], "categories": ["math.DS", "math.FA"], "fields_of_study": ["Mathematics", "Physics"], "published_date": "2021-12-29", "url": "https://arxiv.org/abs/2112.14557", "pdf_url": "https://arxiv.org/pdf/2112.14557v4", "arxiv_id": "2112.14557", "doi": "10.1088/1361-6544/ad0279", "citation_count": 2, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Nonlinearity", "quality_score": 0.1505}
{"id": "4cbd23bb997c6caeb87eb7f799117f9ac5662a2a3d8c76dd6682534e2da4ed00", "sources": ["arxiv", "semantic_scholar"], "title": "Modeling the debonding process of osseointegrated implants due to coupled adhesion and friction", "abstract": "Cementless implants have become widely used for total hip replacement surgery. The long-term stability of these implants is achieved by bone growing around and into the porous surface of the implant, a process called osseointegration. However, debonding of the bone-implant interface can still occur due to aseptic implant loosening and insufficient osseointegration, which may have dramatic consequences. The aim of this work is to describe a new 3D finite element frictional contact formulation for the debonding of partially osseointegrated implants. The contact model is based on a modified Coulomb's friction law (Immel et al. 2020, Biomech. Model. Mechanobiol.) that takes into account the tangential debonding of the bone-implant interface. This model is extended in the direction normal to the bone-implant interface by considering a cohesive zone model, to account for adhesion phenomena in the normal direction and for adhesive friction of partially bonded interfaces. The model is applied to simulate the debonding of an acetabular cup implant. The influence of partial osseointegration and adhesive effects on the long-term stability of the implant is assessed. The influence of different patient- and implant-specific parameters such as the friction coefficient, the trabecular Young's modulus and the interference fit is also analyzed, in order to determine the optimal stability for different configurations. Furthermore, this work provides guidelines for future experimental and computational studies, that are necessary for further parameter calibration.", "authors": ["Katharina Immel", "Vu-Hieu Nguyen", "Guillaume Haiat", "Roger A. Sauer"], "categories": ["physics.med-ph", "cs.CE"], "fields_of_study": ["Computer Science", "Physics", "Medicine"], "published_date": "2021-12-13", "url": "https://arxiv.org/abs/2112.06793", "pdf_url": "https://arxiv.org/pdf/2112.06793v2", "arxiv_id": "2112.06793", "doi": "10.1007/s10237-022-01637-7", "citation_count": 7, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Biomechanics and Modeling in Mechanobiology", "quality_score": 0.2258}
{"id": "f07c63b47d071dd7944334bf04e7189603d597538a3a9eb11805e9f6bf410087", "sources": ["arxiv", "semantic_scholar"], "title": "Merging Models with Fisher-Weighted Averaging", "abstract": "Averaging the parameters of models that have the same architecture and initialization can provide a means of combining their respective capabilities. In this paper, we take the perspective that this \"merging\" operation can be seen as choosing parameters that approximately maximize the joint likelihood of the posteriors of the models' parameters. Computing a simple average of the models' parameters therefore corresponds to making an isotropic Gaussian approximation to their posteriors. We develop an alternative merging procedure based on the Laplace approximation where we approximate each model's posterior as a Gaussian distribution whose precision matrix corresponds to its Fisher information. We first show that our \"Fisher merging\" technique provides a performance boost in settings where simple parameter averaging is currently used -- specifically, robust fine-tuning and model ensembling. Then, we compare merging to standard gradient-based transfer learning and demonstrate that merging enables a fundamentally different method for transferring capabilities across models. Specifically, we show that Fisher merging is competitive with gradient-based transfer learning approaches (while being significantly cheaper) in intermediate-task training and domain-adaptive pre-training. We also show that our merging procedure makes it possible to combine models in previously unexplored ways. We release our code to facilitate future research into methods for merging models.", "authors": ["Michael Matena", "Colin Raffel"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2021-11-18", "url": "https://arxiv.org/abs/2111.09832", "pdf_url": "https://arxiv.org/pdf/2111.09832v2", "arxiv_id": "2111.09832", "doi": "10.52202/068431-1287", "citation_count": 635, "influential_citation_count": 84, "has_code": false, "code_url": null, "venue": "Neural Information Processing Systems", "quality_score": 0.9647}
{"id": "5378acbe20da0b8fdbdbc06899f488c9ffece815ea92f3a1302c0cff00935ef8", "sources": ["arxiv", "semantic_scholar"], "title": "Worst case expansions of complete theories", "abstract": "Given a complete theory $T$ and a subset $Y \\subseteq X^k$, we precisely determine the {\\em worst case complexity}, with respect to further monadic expansions, of an expansion $(M,Y)$ by $Y$ of a model $M$ of $T$ with universe $X$. In particular, although by definition monadically stable/NIP theories are robust under arbitrary monadic expansions, we show that monadically NFCP (equivalently, mutually algebraic) theories are the largest class that is robust under anything beyond monadic expansions. We also exhibit a paradigmatic structure for the failure of each of monadic NFCP/stable/NIP and prove each of these paradigms definably embeds into a monadic expansion of a sufficiently saturated model of any theory without the corresponding property.", "authors": ["Samuel Braunfeld", "Michael C. Laskowski"], "categories": ["math.LO"], "fields_of_study": ["Mathematics"], "published_date": "2021-07-22", "url": "https://arxiv.org/abs/2107.10920", "pdf_url": "https://arxiv.org/pdf/2107.10920v2", "arxiv_id": "2107.10920", "doi": "10.2140/mt.2022.1.15", "citation_count": 5, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Model Th. 1 (2022) 15-30", "quality_score": 0.1945}
{"id": "470fd8abb2e9af7fe04279721da06a545a1cfe1a19cf483531d0d4d21ecbede1", "sources": ["arxiv", "semantic_scholar"], "title": "Self-training with noisy student model and semi-supervised loss function for dcase 2021 challenge task 4", "abstract": "This report proposes a polyphonic sound event detection (SED) method for the DCASE 2021 Challenge Task 4. The proposed SED model consists of two stages: a mean-teacher model for providing target labels regarding weakly labeled or unlabeled data and a self-training-based noisy student model for predicting strong labels for sound events. The mean-teacher model, which is based on the residual convolutional recurrent neural network (RCRNN) for the teacher and student model, is first trained using all the training data from a weakly labeled dataset, an unlabeled dataset, and a strongly labeled synthetic dataset. Then, the trained mean-teacher model predicts the strong label to each of the weakly labeled and unlabeled datasets, which is brought to the noisy student model in the second stage of the proposed SED model. Here, the structure of the noisy student model is identical to the RCRNN-based student model of the mean-teacher model in the first stage. Then, it is self-trained by adding feature noises, such as time-frequency shift, mixup, SpecAugment, and dropout-based model noise. In addition, a semi-supervised loss function is applied to train the noisy student model, which acts as label noise injection. The performance of the proposed SED model is evaluated on the validation set of the DCASE 2021 Challenge Task 4, and then, several ensemble models that combine five-fold validation models with different hyperparameters of the semi-supervised loss function are finally selected as our final models.", "authors": ["Nam Kyun Kim", "Hong Kook Kim"], "categories": ["cs.SD", "cs.LG", "eess.AS"], "fields_of_study": ["Computer Science", "Engineering"], "published_date": "2021-07-06", "url": "https://arxiv.org/abs/2107.02569", "pdf_url": "https://arxiv.org/pdf/2107.02569v1", "arxiv_id": "2107.02569", "doi": null, "citation_count": 14, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.294}
{"id": "02b0495b4938a34a44958de4b5cf6d7f058f0fbdb74917c25a086958dc7673ca", "sources": ["arxiv", "semantic_scholar"], "title": "Interpolation and Model Checking for Nonlinear Arithmetic", "abstract": "We present a new model-based interpolation procedure for satisfiability modulo theories (SMT). The procedure uses a new mode of interaction with the SMT solver that we call solving modulo a model. This either extends a given partial model into a full model for a set of assertions or returns an explanation (a model interpolant) when no solution exists. This mode of interaction fits well into the model-constructing satisfiability (MCSAT) framework of SMT. We use it to develop an interpolation procedure for any MCSAT-supported theory. In particular, this method leads to an effective interpolation procedure for nonlinear real arithmetic. We evaluate the new procedure by integrating it into a model checker and comparing it with state-of-art model-checking tools for nonlinear arithmetic.", "authors": ["Dejan Jovanović", "Bruno Dutertre"], "categories": ["cs.LO", "cs.PL", "cs.SC"], "fields_of_study": ["Computer Science"], "published_date": "2021-06-08", "url": "https://arxiv.org/abs/2106.04340", "pdf_url": "https://arxiv.org/pdf/2106.04340v1", "arxiv_id": "2106.04340", "doi": "10.1007/978-3-030-81688-9_13", "citation_count": 4, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "International Conference on Computer Aided Verification", "quality_score": 0.1747}
{"id": "36de292de3329dc39d77c23462908988c91a54df4b1b1d57bf3109a6fac7f19f", "sources": ["arxiv", "semantic_scholar"], "title": "Musical Prosody-Driven Emotion Classification: Interpreting Vocalists Portrayal of Emotions Through Machine Learning", "abstract": "The task of classifying emotions within a musical track has received widespread attention within the Music Information Retrieval (MIR) community. Music emotion recognition has traditionally relied on the use of acoustic features, verbal features, and metadata-based filtering. The role of musical prosody remains under-explored despite several studies demonstrating a strong connection between prosody and emotion. In this study, we restrict the input of traditional machine learning algorithms to the features of musical prosody. Furthermore, our proposed approach builds upon the prior by classifying emotions under an expanded emotional taxonomy, using the Geneva Wheel of Emotion. We utilize a methodology for individual data collection from vocalists, and personal ground truth labeling by the artist themselves. We found that traditional machine learning algorithms when limited to the features of musical prosody (1) achieve high accuracies for a single singer, (2) maintain high accuracy when the dataset is expanded to multiple singers, and (3) achieve high accuracies when trained on a reduced subset of the total features.", "authors": ["Nicholas Farris", "Brian Model", "Richard Savery", "Gil Weinberg"], "categories": ["cs.SD", "cs.LG", "eess.AS"], "fields_of_study": ["Computer Science", "Engineering"], "published_date": "2021-06-04", "url": "https://arxiv.org/abs/2106.02556", "pdf_url": "https://arxiv.org/pdf/2106.02556v2", "arxiv_id": "2106.02556", "doi": null, "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0753}
{"id": "b8321bd11df614dc10182956d8c23e9d0889bc68415c7e40987e515027383f2a", "sources": ["arxiv", "semantic_scholar"], "title": "Towards physically consistent data-driven weather forecasting: Integrating data assimilation with equivariance-preserving deep spatial transformers", "abstract": "There is growing interest in data-driven weather prediction (DDWP), for example using convolutional neural networks such as U-NETs that are trained on data from models or reanalysis. Here, we propose 3 components to integrate with commonly used DDWP models in order to improve their physical consistency and forecast accuracy. These components are 1) a deep spatial transformer added to the latent space of the U-NETs to preserve a property called equivariance, which is related to correctly capturing rotations and scalings of features in spatio-temporal data, 2) a data-assimilation (DA) algorithm to ingest noisy observations and improve the initial conditions for next forecasts, and 3) a multi-time-step algorithm, which combines forecasts from DDWP models with different time steps through DA, improving the accuracy of forecasts at short intervals. To show the benefit/feasibility of each component, we use geopotential height at 500~hPa (Z500) from ERA5 reanalysis and examine the short-term forecast accuracy of specific setups of the DDWP framework. Results show that the equivariance-preserving networks (U-STNs) clearly outperform the U-NETs, for example improving the forecast skill by $45\\%$. Using a sigma-point ensemble Kalman (SPEnKF) algorithm for DA and U-STN as the forward model, we show that stable, accurate DA cycles are achieved even with high observation noise. The DDWP+DA framework substantially benefits from large ($O(1000)$) ensembles that are inexpensively generated with the data-driven forward model in each DA cycle. The multi-time-step DDWP+DA framework also shows promises, e.g., it reduces the average error by factors of 2-3.", "authors": ["Ashesh Chattopadhyay", "Mustafa Mustafa", "Pedram Hassanzadeh", "Eviatar Bach", "Karthik Kashinath"], "categories": ["physics.ao-ph", "cs.AI", "cs.LG", "physics.comp-ph"], "fields_of_study": ["Computer Science", "Physics"], "published_date": "2021-03-16", "url": "https://arxiv.org/abs/2103.09360", "pdf_url": "https://arxiv.org/pdf/2103.09360v1", "arxiv_id": "2103.09360", "doi": "10.5194/GMD-2021-71", "citation_count": 42, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4084}
{"id": "695ab4feac12ddcff9312b75eca1cbbb9e27fbb960762249f2afe8577c651ecf", "sources": ["arxiv", "semantic_scholar"], "title": "Analysis of Interpolation based Image In-painting Approaches", "abstract": "Interpolation and internal painting are one of the basic approaches in image internal painting, which is used to eliminate undesirable parts that occur in digital images or to enhance faulty parts. This study was designed to compare the interpolation algorithms used in image in-painting in the literature. Errors and noise generated on the colour and grayscale formats of some of the commonly used standard images in the literature were corrected by using Cubic, Kriging, Radial based function and High dimensional model representation approaches and the results were compared using standard image comparison criteria, namely, PSNR (peak signal-to-noise ratio), SSIM (Structural SIMilarity), Mean Square Error (MSE). According to the results obtained from the study, the absolute superiority of the methods against each other was not observed. However, Kriging and RBF interpolation give better results both for numerical data and visual evaluation for image in-painting problems with large area losses.", "authors": ["Mustafa Zor", "Erkan Bostanci", "Mehmet Serdar Guzel", "Erinc Karatas"], "categories": ["cs.CV", "eess.IV"], "fields_of_study": ["Computer Science", "Engineering"], "published_date": "2021-02-12", "url": "https://arxiv.org/abs/2102.06564", "pdf_url": "https://arxiv.org/pdf/2102.06564v1", "arxiv_id": "2102.06564", "doi": "10.1201/9781003221333-8", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.0753}