Datasets:

fineset-io
/

model-merging-papers

id string	sources list	title string	abstract string	authors list	categories list	fields_of_study list	published_date timestamp[s]	url string	pdf_url string	arxiv_id string	doi string	citation_count int64	influential_citation_count int64	has_code bool	code_url string	venue string	quality_score float64
49c489b00d67a57a3f147cb9877b0e2d5beac0528ca0ba0557188ef83f34424c	[ "arxiv" ]	PACT: Preserving Anchored Cores in Task-vectors for Model Merging	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...	[ "Ningyuan Shi", "Zhipeng Zhou", "Hao Wang", "Chunyan Miao", "Peilin Zhao" ]	[ "cs.LG" ]	[]	2026-06-17T00:00:00	https://arxiv.org/abs/2606.18627	https://arxiv.org/pdf/2606.18627v1	2606.18627	null	0	0	false	null	null	0.35
395f544d427cb59b539ac8d673e2264256ff312294d4e3e6ad6012bd549c970d	[ "arxiv" ]	Essential Subspace Merging for Multi-Task Learning	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 update...	[ "Longhua Li", "Lei Qi", "Xin Geng", "Qi Tian" ]	[ "cs.LG", "cs.AI" ]	[]	2026-06-17T00:00:00	https://arxiv.org/abs/2606.19164	https://arxiv.org/pdf/2606.19164v1	2606.19164	null	0	0	false	null	null	0.35
2df46214da6c29caf600f0569b82aabac9c2df46536dadd091d8a22ac969162d	[ "arxiv" ]	Enhancing Multilingual Reasoning via Steerable Model Merging	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 sou...	[ "Zhuoran Li", "Rui Xu", "Jian Yang", "Junnan Liu", "Zhijun Chen", "Qianren Mao", "Hongcheng Guo", "Jiaheng Liu", "Likang Xiao", "Ming Li", "Xiaojie Wang" ]	[ "cs.CL" ]	[]	2026-06-17T00:00:00	https://arxiv.org/abs/2606.19002	https://arxiv.org/pdf/2606.19002v1	2606.19002	null	0	0	false	null	null	0.35
28890da352d2d2348f1e45e05f8bb8bd71ab5fd0755926ee200db6c818f73759	[ "arxiv" ]	Task-Restricted Symmetries in Recurrent Weight Space	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 co...	[ "Simon Dräger" ]	[ "cs.LG" ]	[]	2026-06-16T00:00:00	https://arxiv.org/abs/2606.18457	https://arxiv.org/pdf/2606.18457v1	2606.18457	null	0	0	false	null	null	0.35
1b591adcb860ed122ac1b487df99fba218e8b7b048bbe8543967d98e2280a855	[ "arxiv", "semantic_scholar" ]	Post-Hoc Merging is Not Enough: Many-Shot Model Merging with Loss-Gap Balancing	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 ...	[ "Kyungjin Im", "Miru Kim", "Chanin Eom", "Minhae Kwon" ]	[ "cs.AI" ]	[ "Computer Science" ]	2026-06-15T00:00:00	https://arxiv.org/abs/2606.16501	https://arxiv.org/pdf/2606.16501v1	2606.16501	null	0	0	false	null	null	0.35
cf332f398fe5ffabea6f8897e661103c111d4ddbfcf5a148d042ce3f50248124	[ "arxiv", "semantic_scholar" ]	From Parameters to Feature Space: Task Arithmetic for Backdoor Mitigation in Model Merging	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 substanti...	[ "Zhenqian Zhu", "Yamin Hu", "Yiya Diao", "Weixiang Li", "Haodong Li", "Wenjian Luo" ]	[ "cs.CR", "cs.LG" ]	[ "Computer Science" ]	2026-06-10T00:00:00	https://arxiv.org/abs/2606.12498	https://arxiv.org/pdf/2606.12498v1	2606.12498	null	0	0	false	null	null	0.35
bd802d69ca3ac24ad1e580769fb44be3f2888cf6f4e649521cb82710bdb8a7a5	[ "arxiv", "semantic_scholar" ]	Closed-Form Spectral Regularization for Multi-Task Model Merging	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 ...	[ "Yongxian Wei", "Runxi Cheng", "Xingxuan Zhang", "Li Shen", "Chun Yuan", "Peng Cui", "Dacheng Tao" ]	[ "cs.LG", "cs.CV" ]	[ "Computer Science" ]	2026-06-05T00:00:00	https://arxiv.org/abs/2606.07289	https://arxiv.org/pdf/2606.07289v1	2606.07289	null	0	0	false	null	null	0.35
147dd55918a7551461e87f20ec491db53a9fceb48bdc3a31e2173ad7b1ae29d9	[ "arxiv", "semantic_scholar" ]	TaDA: Calibrated Probe Gating for Task-Domain LoRA Merging	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...	[ "Huy Quoc To", "Fuyi Li", "Guangyan Huang", "Ming Liu" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-06-03T00:00:00	https://arxiv.org/abs/2606.05016	https://arxiv.org/pdf/2606.05016v1	2606.05016	null	0	0	false	null	null	0.35
4e49f8d87029823b017e56d722de2f7801937a2a0c1695f02e83aaed3e48dd2b	[ "arxiv", "semantic_scholar" ]	RogueMerge: Robust and Unified Attacks against LLM Model Merging	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 w...	[ "Jinghuai Zhang", "Yetian He", "Kunlin Cai", "Han Zhao", "Fnu Suya", "Yuan Tian" ]	[ "cs.CR", "cs.LG" ]	[ "Computer Science" ]	2026-06-02T00:00:00	https://arxiv.org/abs/2606.03344	https://arxiv.org/pdf/2606.03344v1	2606.03344	null	0	0	false	null	null	0.35
456376c706209431c598f624319e233b0fd495ede4262d4232c71457c3843746	[ "arxiv", "semantic_scholar" ]	Decentralized Instruction Tuning: Conflict-Aware Splitting and Weight Merging	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 independen...	[ "Minsik Choi", "Geewook Kim" ]	[ "cs.LG" ]	[ "Computer Science" ]	2026-06-01T00:00:00	https://arxiv.org/abs/2606.01717	https://arxiv.org/pdf/2606.01717v1	2606.01717	null	0	0	true	https://github.com/naver-ai/merit	null	0.65
5078b7222424a78bbcb406dd8820db45bbaf87f409cd09627c0404c5a277f551	[ "arxiv", "semantic_scholar" ]	ThinkSwitch: Context Distillation with LoRA and Weight Interpolation for Specific-Purpose Reasoning Tasks	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-tra...	[ "Dhruv Saini", "Rohan Pandey" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2026-05-31T00:00:00	https://arxiv.org/abs/2606.01080	https://arxiv.org/pdf/2606.01080v1	2606.01080	null	0	0	false	null	null	0.35
aee8014d45f7f9672028b1ebd4be4ef199a587f45b75e86b1b226c6390100878	[ "arxiv", "semantic_scholar" ]	Saliency-Aware Model Merging	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 ...	[ "Jungin Park", "Jiyoung Lee", "Kwanghoon Sohn" ]	[ "cs.LG", "cs.CV" ]	[ "Computer Science" ]	2026-05-30T00:00:00	https://arxiv.org/abs/2606.00511	https://arxiv.org/pdf/2606.00511v1	2606.00511	null	0	0	false	null	null	0.35
412e8b6e48974ff122c31dc2b22eef04d34eb995c98ab6d0a5cbc6399dcc2ec5	[ "arxiv", "semantic_scholar" ]	Access Sets Matter: Budgeting Expert Reads for Scalable Weight-Space Model Merging	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 oper...	[ "Yuanyi Wang", "Yanggan Gu", "Su Lu", "Yifan Yang", "Zhaoyi Yan", "Congkai Xie", "Jianmin Wu", "Hongxia Yang" ]	[ "cs.LG", "eess.SY" ]	[ "Computer Science", "Engineering" ]	2026-05-28T00:00:00	https://arxiv.org/abs/2605.29489	https://arxiv.org/pdf/2605.29489v1	2605.29489	null	0	0	false	null	null	0.35
12df446b2647a3b452f7fe785de9055821053696a351e665f0c00b57c56a720d	[ "arxiv", "semantic_scholar" ]	Model Merging by Output-Space Projection	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 me...	[ "Bethan Evans", "Benjamin Etheridge", "Stephen Roberts", "Jared Tanner" ]	[ "cs.LG", "cs.IT" ]	[ "Computer Science", "Mathematics" ]	2026-05-27T00:00:00	https://arxiv.org/abs/2605.29101	https://arxiv.org/pdf/2605.29101v1	2605.29101	null	0	0	false	null	null	0.35
67b9cba3834464d319f7ef9b44251449cc6198e7fb911bab4dc7d715ce1a9991	[ "arxiv", "semantic_scholar" ]	What-If World: A Causal Benchmark for General World Models in Embodied Scenarios	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...	[ "Kunlin Cai", "Rui Song", "Jinghuai Zhang", "Kaiyuan Zhang", "Pranav Bodapati", "Alicia Yu", "Fnu Suya", "Mohammad Rostami", "Jiaqi Ma", "Yuan Tian" ]	[ "cs.CV" ]	[ "Computer Science" ]	2026-05-26T00:00:00	https://arxiv.org/abs/2605.27589	https://arxiv.org/pdf/2605.27589v1	2605.27589	null	0	0	true	null	null	0.65
a54dc89832d4feea79b4c2458de043b65a0686836f484fe0362be31cd84b4e92	[ "arxiv", "semantic_scholar" ]	Model Merging on Loss Landscape: A Geometry Perspective	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 m...	[ "Juanwu Lu", "Anand Bhaskar", "Brian Axelrod", "Ekaterina Tolstaya", "Tristan Emrich" ]	[ "cs.LG", "cs.AI", "stat.ML" ]	[ "Computer Science", "Mathematics" ]	2026-05-26T00:00:00	https://arxiv.org/abs/2605.26693	https://arxiv.org/pdf/2605.26693v1	2605.26693	null	0	0	false	null	null	0.35
2620bf3007c1fffa0d289e418af177aba55134e727d5a7148ef8f40fbd958dc4	[ "arxiv", "semantic_scholar" ]	On the Limits of Model Merging for Multilinguality in Pre-Training	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, mer...	[ "Seth Aycock", "Fedor Vitiugin", "Aleksandr Umnov", "Christof Monz", "Khalil Sima'an" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-05-25T00:00:00	https://arxiv.org/abs/2605.25846	https://arxiv.org/pdf/2605.25846v1	2605.25846	null	0	0	false	null	null	0.35
05faab4447d90c658ab877b93dc73703d4326f5591fd5014731e03717ac3696c	[ "arxiv", "semantic_scholar" ]	Causal Physics Steering in Video World Models via Concept Activation Vectors	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 separa...	[ "Nahid Alam" ]	[ "cs.CV" ]	[ "Computer Science" ]	2026-05-23T00:00:00	https://arxiv.org/abs/2605.24322	https://arxiv.org/pdf/2605.24322v1	2605.24322	null	0	0	false	null	null	0.35
2d9ae35b0561209cbb693daaff9e23b48160e3d2adf8b2dc4d05cc9cb6bf4984	[ "arxiv", "semantic_scholar" ]	GEM-4D: Geometry-Enhanced Video World Models for Robot Manipulation	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 ...	[ "Kaichen Zhou", "Yuzhen Chen", "Fangneng Zhan", "Hang Hua", "Grace Chen", "Xinhai Chang", "Ao Qu", "Yilun Du", "Zhuang Liu", "Paul Pu Liang", "Mengyu Wang" ]	[ "cs.CV", "cs.RO" ]	[ "Computer Science" ]	2026-05-20T00:00:00	https://arxiv.org/abs/2605.22882	https://arxiv.org/pdf/2605.22882v3	2605.22882	null	1	0	false	null	null	0.35
1dc8d26a5e3de54d49e2b766ae1778c983910bfa265dcdb3482e7e1403e251fe	[ "arxiv", "semantic_scholar" ]	Tunable MAGMAX: Preference-Aware Model Merging for Continual Learning	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 par...	[ "Kei Hiroshima", "Kento Uchida", "Shinichi Shirakawa" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2026-05-20T00:00:00	https://arxiv.org/abs/2605.20803	https://arxiv.org/pdf/2605.20803v1	2605.20803	null	0	0	false	null	null	0.35
e7b8fedf7f260f82ca10f1f5cac674f55430d72fa9565399e574385de9cf7939	[ "arxiv", "semantic_scholar" ]	Unlocking the Potential of Continual Model Merging: An ODE Perspective	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 ...	[ "Lihong Lin", "Haidong Kang" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2026-05-19T00:00:00	https://arxiv.org/abs/2605.19409	https://arxiv.org/pdf/2605.19409v3	2605.19409	null	0	0	false	null	null	0.35
aa6c8164da3fdde4a93cb23ca38cc6dc81f2015e9e31df2f32fcbd76fb0ae033	[ "arxiv", "semantic_scholar" ]	Distilling Linearized Behavior into Non-Linear Fine-Tuning for Effective Task Arithmetic	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 disent...	[ "Thomas Sommariva", "Francesca Morandi", "Simone Calderara", "Angelo Porrello" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2026-05-18T00:00:00	https://arxiv.org/abs/2605.18993	https://arxiv.org/pdf/2605.18993v2	2605.18993	null	0	0	false	null	null	0.35
d17440f0bf7bba9cbaf5113028c821a5f47ce81a85eb4f2ee19a2a437b20ca3d	[ "arxiv", "semantic_scholar" ]	Dynamic Model Merging Made Slim	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...	[ "Guodong Du", "Wanyu Lin" ]	[ "cs.LG", "cs.AI", "cs.CL" ]	[ "Computer Science" ]	2026-05-17T00:00:00	https://arxiv.org/abs/2605.18904	https://arxiv.org/pdf/2605.18904v1	2605.18904	null	2	0	false	null	null	0.35
6613298ee5f564f65b60c55ffbcb7d084fe4918cb2546935c9bc0a29ebd0569b	[ "arxiv", "semantic_scholar" ]	E-PMQ: Expert-Guided Post-Merge Quantization with Merged-Weight Anchoring	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 traini...	[ "Wenjun Wang", "Yanggan Gu", "Shuo Cai", "Yuanyi Wang", "Pengkai Wang", "Jianmin Wu", "Hongxia Yang" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-05-16T00:00:00	https://arxiv.org/abs/2605.16882	https://arxiv.org/pdf/2605.16882v1	2605.16882	null	2	0	false	null	null	0.35
c839223e80b83e549e5d2ea221266f31441717b2990ec107efe364fb1e82d2de	[ "arxiv", "semantic_scholar" ]	Bayesian Model Merging	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...	[ "Kaiyang Li", "Shaobo Han", "Qing Su", "Shihao Ji" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2026-05-13T00:00:00	https://arxiv.org/abs/2605.12843	https://arxiv.org/pdf/2605.12843v1	2605.12843	null	0	0	false	null	null	0.35
4930a71f466f4ebff3edc2a1306f6215235eb5ef4828386386f16ac96a8fe3cb	[ "arxiv", "semantic_scholar" ]	FeatCal: Feature Calibration for Post-Merging Models	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 t...	[ "Yanggan Gu", "Shuo Cai", "Zihao Wang", "Wenjun Wang", "Yuanyi Wang", "Pengkai Wang", "Sirui Huang", "Su Lu", "Jianmin Wu", "Hongxia Yang" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2026-05-13T00:00:00	https://arxiv.org/abs/2605.13030	https://arxiv.org/pdf/2605.13030v1	2605.13030	null	2	0	false	null	null	0.35
4d47028b83a40d304b240a66abfd251dc54660ac5f485f8f02dcbda8592f01b8	[ "arxiv", "semantic_scholar" ]	EHR-RAGp: Retrieval-Augmented Prototype-Guided Foundation Model for Electronic Health Records	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 con...	[ "Saeed Shurrab", "Mariam Al-Omari", "Dana El Samad", "Farah E. Shamout" ]	[ "cs.IR", "cs.AI", "cs.LG" ]	[ "Computer Science" ]	2026-05-12T00:00:00	https://arxiv.org/abs/2605.12335	https://arxiv.org/pdf/2605.12335v1	2605.12335	null	0	0	false	null	null	0.35
49e1be355fac19775b08c9c2a98481aab55608d95b3781251826db4ae4199ab4	[ "arxiv", "semantic_scholar" ]	Revitalizing the Beginning: Avoiding Storage Dependency for Model Merging in Continual Learning	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 li...	[ "Xi Wang", "Cheng Deng" ]	[ "cs.LG", "cs.CV" ]	[ "Computer Science" ]	2026-05-08T00:00:00	https://arxiv.org/abs/2605.08311	https://arxiv.org/pdf/2605.08311v1	2605.08311	null	0	0	false	null	null	0.35
c4c7e408fe306af6d068f85164a09e6c07113b5c3bf2991a68803b3725a67dd7	[ "arxiv", "semantic_scholar" ]	A Tutorial for Evaluating Cure Model Appropriateness	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 de...	[ "A Tutorial for Evaluating Cure Model Appropriateness Geethanjalee Mudunkotuwa", "Durbadal Ghosh", "Subodh Selukar" ]	[ "stat.ME" ]	[ "Mathematics" ]	2026-05-06T00:00:00	https://arxiv.org/abs/2605.04999	https://arxiv.org/pdf/2605.04999v2	2605.04999	null	0	0	false	null	null	0.35
b0888e2f3515294cf1c12a315703628619062902b93a6bf304294b87f36a687b	[ "arxiv", "semantic_scholar" ]	Model Merging: Foundations and Algorithms	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...	[ "Donato Crisostomi" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2026-05-02T00:00:00	https://arxiv.org/abs/2605.01580	https://arxiv.org/pdf/2605.01580v1	2605.01580	null	0	0	false	null	null	0.35
39cc3cc4d4dcdcbf8a4b9a2e71005daba831fe658bb604e0a881471aa03607b0	[ "arxiv", "semantic_scholar" ]	Auto-FlexSwitch: Efficient Dynamic Model Merging via Learnable Task Vector Compression	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 param...	[ "Junqi Gao", "Dazhi Zhang", "Zhichang Guo", "Biqing Qi", "Yi Ran", "Wangmeng Zuo" ]	[ "cs.LG" ]	[ "Computer Science" ]	2026-04-30T00:00:00	https://arxiv.org/abs/2604.28109	https://arxiv.org/pdf/2604.28109v1	2604.28109	10.48550/arXiv.2604.28109	0	0	false	null	arXiv.org	0.55
ad97dee1a4e25cf7bd714d775cb31ba4027eec524dd379dbade962f7e3bad9ac	[ "arxiv", "semantic_scholar" ]	Embedding Arithmetic: A Lightweight, Tuning-Free Framework for Post-hoc Bias Mitigation in Text-to-Image Models	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 con...	[ "Venkatesh Thirugnana Sambandham", "Torsten Schön" ]	[ "cs.CV" ]	[ "Computer Science" ]	2026-04-20T00:00:00	https://arxiv.org/abs/2604.18167	https://arxiv.org/pdf/2604.18167v1	2604.18167	10.48550/arXiv.2604.18167	0	0	true	https://github.com/cvims/EMBEDDING-ARITHMETIC}	arXiv.org	0.85
983090ef75a7f9605b4908be64c04f9e7cd9d79d0f98c584f65806a7d147ad40	[ "arxiv", "semantic_scholar" ]	Understanding and Enforcing Weight Disentanglement in Task Arithmetic	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, wha...	[ "Shangge Liu", "Yuehan Yin", "Lei Wang", "Qi Fan", "Yinghuan Shi", "Wenbin Li", "Yang Gao", "Dacheng Tao" ]	[ "cs.AI" ]	[ "Computer Science" ]	2026-04-18T00:00:00	https://arxiv.org/abs/2604.17078	https://arxiv.org/pdf/2604.17078v1	2604.17078	10.48550/arXiv.2604.17078	0	0	true	https://github.com/RL-MIND/OrthoReg}{https://github.com/RL-MIND/OrthoReg}	arXiv.org	0.8465
810634cf4ae51a3ac5284f6fe396bc40a5580303ae4037458ae13dbcb06af677	[ "arxiv", "semantic_scholar" ]	Task Alignment: A simple and effective proxy for model merging in computer vision	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 datase...	[ "Pau de Jorge", "César Roberto de Souza", "Björn Michele", "Mert Bülent Sarıyıldız", "Philippe Weinzaepfel", "Florent Perronnin", "Diane Larlus", "Yannis Kalantidis" ]	[ "cs.CV" ]	[ "Computer Science" ]	2026-04-14T00:00:00	https://arxiv.org/abs/2604.12935	https://arxiv.org/pdf/2604.12935v1	2604.12935	10.48550/arXiv.2604.12935	1	0	false	null	arXiv.org	0.5431
92d17abab81ff5ac8976c3d7e6d482b91e5f8fff55e74a3f803824cba484ff75	[ "arxiv", "semantic_scholar" ]	One Model to Translate Them All? A Journey to Mount Doom for Multilingual Model Merging	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 fo...	[ "Baban Gain", "Asif Ekbal", "Trilok Nath Singh" ]	[ "cs.CL", "cs.AI" ]	[ "Computer Science" ]	2026-04-03T00:00:00	https://arxiv.org/abs/2604.02881	https://arxiv.org/pdf/2604.02881v1	2604.02881	10.48550/arXiv.2604.02881	2	0	false	null	arXiv.org	0.5305
e78f568ec4f4de1e66aacc1a647a9724d46e1db0e6ecd4de2a0c54c0df7018bf	[ "arxiv", "semantic_scholar" ]	Countering Catastrophic Forgetting of Large Language Models for Better Instruction Following via Weight-Space Model Merging	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 gener...	[ "Mengxian Lyu", "Cheng Peng", "Ziyi Chen", "Mengyuan Zhang", "Jieting Li Lu", "Yonghui Wu" ]	[ "cs.CL", "cs.AI" ]	[ "Computer Science" ]	2026-04-02T00:00:00	https://arxiv.org/abs/2604.01538	https://arxiv.org/pdf/2604.01538v1	2604.01538	10.48550/arXiv.2604.01538	0	0	true	null	arXiv.org	0.8181
f65ff7e2ce2f761f8bea3dfa8b6abb18a456892fd1c93b0cb11808af831dfcfc	[ "arxiv", "semantic_scholar" ]	Yet Even Less Is Even Better For Agentic, Reasoning, and Coding LLMs	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 ach...	[ " 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", "...	[ "cs.SE" ]	[ "Computer Science" ]	2026-04-01T00:00:00	https://arxiv.org/abs/2604.00824	https://arxiv.org/pdf/2604.00824v3	2604.00824	10.48550/arXiv.2604.00824	0	0	false	null	arXiv.org	0.5282
5c142e2cc97364a477a5751796bb3145d048a3a33452ce2c4487484b411bf719	[ "arxiv", "semantic_scholar" ]	Robust Language Identification for Romansh Varieties	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...	[ "Charlotte Model", "Sina Ahmadi", "Jannis Vamvas" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-03-16T00:00:00	https://arxiv.org/abs/2603.15969	https://arxiv.org/pdf/2603.15969v2	2603.15969	10.48550/arXiv.2603.15969	1	0	false	null	arXiv.org	0.5099
e6046277e3441301722c53b9153a9b30a0c0165164610b33386bc5a8e0aafc3f	[ "arxiv", "semantic_scholar" ]	Resolving Interference (RI): Disentangling Models for Improved Model Merging	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...	[ "Pratik Ramesh", "George Stoica", "Arun Iyer", "Leshem Choshen", "Judy Hoffman" ]	[ "cs.LG", "cs.AI", "cs.CL", "cs.CV" ]	[ "Computer Science" ]	2026-03-13T00:00:00	https://arxiv.org/abs/2603.13467	https://arxiv.org/pdf/2603.13467v1	2603.13467	10.48550/arXiv.2603.13467	0	0	true	https://github.com/pramesh39/resolving_interference	arXiv.org	0.7827
fe57823867561511731633bcaa9cc3339d991cdde48cc5784e2b4c35349c9f1b	[ "arxiv", "semantic_scholar" ]	Mortgage Burnout and Selection Effects in Heterogeneous Cox Hazard Models	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 ...	[ "Andrew Lesniewski" ]	[ "q-fin.MF", "econ.GN", "stat.ME" ]	[ "Economics", "Mathematics" ]	2026-03-12T00:00:00	https://arxiv.org/abs/2603.12422	https://arxiv.org/pdf/2603.12422v3	2603.12422	null	0	0	false	null	null	0.3216
a30c8672c6849d29fb693d80d07ae96594564ace209e679e5ce86faabace6313	[ "arxiv", "semantic_scholar" ]	An Empirical Study and Theoretical Explanation on Task-Level Model-Merging Collapse	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...	[ "Yuan Cao", "Dezhi Ran", "Yuzhe Guo", "Mengzhou Wu", "Simin Chen", "Linyi Li", "Wei Yang", "Tao Xie" ]	[ "cs.AI" ]	[ "Computer Science" ]	2026-03-10T00:00:00	https://arxiv.org/abs/2603.09463	https://arxiv.org/pdf/2603.09463v1	2603.09463	10.48550/arXiv.2603.09463	0	0	false	null	arXiv.org	0.503
37f13bce44fd265b6a320b12f9c050d116b89cc93a79a83687c452d9bb5a7b55	[ "arxiv", "semantic_scholar" ]	Model Merging in the Era of Large Language Models: Methods, Applications, and Future Directions	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 a...	[ "Mingyang Song", "Mao Zheng" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-03-10T00:00:00	https://arxiv.org/abs/2603.09938	https://arxiv.org/pdf/2603.09938v2	2603.09938	10.48550/arXiv.2603.09938	1	0	false	null	arXiv.org	0.503
db3afe44be8b623f6135a2f763b6005dca6f6d1e7fcf58df8df321f7d774096c	[ "arxiv", "semantic_scholar" ]	DC-Merge: Improving Model Merging with Directional Consistency	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...	[ "Han-Chen Zhang", "Zi-Hao Zhou", "Mao-Lin Luo", "Shimin Di", "Min-Ling Zhang", "Tong Wei" ]	[ "cs.LG", "cs.CV" ]	[ "Computer Science" ]	2026-03-06T00:00:00	https://arxiv.org/abs/2603.06242	https://arxiv.org/pdf/2603.06242v2	2603.06242	10.48550/arXiv.2603.06242	0	0	true	https://github.com/Tobeginwith/DC-Merge	arXiv.org	0.7703
016ba6b7f09a96c77fbcb5f32eb427ad06f939de0673edd78cb9f8225d5dec4f	[ "arxiv", "semantic_scholar" ]	Magic partition functions: Sign smoothing convolutions with Dirichlet invertible arithmetic functions	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 num...	[ "Maxie Dion Schmidt" ]	[ "math.NT" ]	[ "Mathematics" ]	2026-03-06T00:00:00	https://arxiv.org/abs/2603.06890	https://arxiv.org/pdf/2603.06890v1	2603.06890	null	0	0	false	null	null	0.3172
40b92de5d2aedd4870494e7f9ab1b1cfbbcc73e90e3db0aec6121eb9bd96a90a	[ "arxiv", "semantic_scholar" ]	Bridging Domains through Subspace-Aware Model Merging	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 distinc...	[ "Levy Chaves", "Chao Zhou", "Rebekka Burkholz", "Eduardo Valle", "Sandra Avila" ]	[ "cs.LG", "cs.AI", "cs.CV" ]	[ "Computer Science" ]	2026-03-06T00:00:00	https://arxiv.org/abs/2603.05768	https://arxiv.org/pdf/2603.05768v2	2603.05768	10.48550/arXiv.2603.05768	0	0	false	null	arXiv.org	0.4984
b581d341a74672acab3484e932bcbd2ece749a7a612d03a5f423f853eb112465	[ "arxiv", "semantic_scholar" ]	Model Medicine: A Clinical Framework for Understanding, Diagnosing, and Treating AI Models	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 pa...	[ "Jihoon Jeong" ]	[ "cs.AI", "cs.CL", "cs.LG" ]	[ "Computer Science" ]	2026-03-05T00:00:00	https://arxiv.org/abs/2603.04722	https://arxiv.org/pdf/2603.04722v2	2603.04722	10.48550/arXiv.2603.04722	2	2	true	null	arXiv.org	0.7685
e12060224d6c67308d7f2a09f5a6038d0ad1760d870936f88650fc5c6b868993	[ "arxiv", "semantic_scholar" ]	BD-Merging: Bias-Aware Dynamic Model Merging with Evidence-Guided Contrastive Learning	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 M...	[ "Yuhan Xie", "Chen Lyu" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2026-03-04T00:00:00	https://arxiv.org/abs/2603.03920	https://arxiv.org/pdf/2603.03920v2	2603.03920	10.48550/arXiv.2603.03920	0	0	false	null	arXiv.org	0.4961
9140caa560fbd31b5d80aa941eba5427a59e7abe3257ee8ebb13c1f8cf9ab61c	[ "arxiv", "semantic_scholar" ]	ACE-Merging: Data-Free Model Merging with Adaptive Covariance Estimation	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, resolvi...	[ "Bo Xu", "Haotian Wu", "Hehai Lin", "Weiquan Huang", "Beier Zhu", "Yao Shu", "Chengwei Qin" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-03-03T00:00:00	https://arxiv.org/abs/2603.02945	https://arxiv.org/pdf/2603.02945v2	2603.02945	10.48550/arXiv.2603.02945	0	0	false	null	arXiv.org	0.495
c283172d605ef6c0b753cd0c6548ef754c6c1c256bdbf23b9c2531bdbd219265	[ "arxiv", "semantic_scholar" ]	Leveraging Model Soups to Classify Intangible Cultural Heritage Images from the Mekong Delta	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 spurio...	[ "Quoc-Khang Tran", "Minh-Thien Nguyen", "Nguyen-Khang Pham" ]	[ "cs.CV", "cs.AI", "cs.LG" ]	[ "Computer Science" ]	2026-03-02T00:00:00	https://arxiv.org/abs/2603.02181	https://arxiv.org/pdf/2603.02181v1	2603.02181	10.32913/mic-ict-research.v2025.n3.1395	0	0	false	null	null	0.3143
efd5ab0e7601576782c61cce15281a78fc4b8ddee519aabb849d90cd60e549b4	[ "arxiv", "semantic_scholar" ]	Model Merging in the Essential Subspace	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, ...	[ "Longhua Li", "Lei Qi", "Qi Tian", "Xin Geng" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2026-02-23T00:00:00	https://arxiv.org/abs/2602.20208	https://arxiv.org/pdf/2602.20208v1	2602.20208	10.48550/arXiv.2602.20208	2	0	false	null	arXiv.org	0.4858
5117d328859e8ea7be7aa8a22eabc7717ef438946e98d7f9f29f0ed6a7e71cc3	[ "arxiv", "semantic_scholar" ]	Interpolation in Proof Theory	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 Pitt...	[ "Iris van der Giessen", "Raheleh Jalali", "Roman Kuznets" ]	[ "cs.LO" ]	[ "Computer Science" ]	2026-02-18T00:00:00	https://arxiv.org/abs/2602.16318	https://arxiv.org/pdf/2602.16318v1	2602.16318	10.48550/arXiv.2602.16318	3	1	false	null	arXiv.org	0.4801
9cccff0d7842f751feb9afa193c1a35a1620568eef4f41ce9df3cc39517ac554	[ "arxiv", "semantic_scholar" ]	Beyond Parameter Arithmetic: Sparse Complementary Fusion for Distribution-Aware Model Merging	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 s...	[ "Weihong Lin", "Lin Sun", "Qilong Shi", "Aomufei Yuan", "Yuxuan Tian", "Zhengyang Wang", "Guangxiang Zhao", "Xiangzheng Zhang", "Tong Yang" ]	[ "cs.AI" ]	[ "Computer Science" ]	2026-02-12T00:00:00	https://arxiv.org/abs/2602.11717	https://arxiv.org/pdf/2602.11717v1	2602.11717	10.48550/arXiv.2602.11717	0	0	false	null	arXiv.org	0.4732
7dd364e9eb6e0b67bf0d0306f9f9b865dda40acf9b091cc718b58b19c537f8c6	[ "arxiv", "semantic_scholar" ]	Craig Interpolation in Program Verification	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 derivati...	[ "Philipp Rümmer" ]	[ "cs.LO" ]	[ "Computer Science" ]	2026-02-09T00:00:00	https://arxiv.org/abs/2602.08532	https://arxiv.org/pdf/2602.08532v1	2602.08532	10.48550/arXiv.2602.08532	0	0	false	null	arXiv.org	0.4698
c21aeafc7629df476c4c2bb500d537f78feac59fd8c1430fa9a67d84ff3402cd	[ "arxiv", "semantic_scholar" ]	M-Loss: Quantifying Model Merging Compatibility with Limited Unlabeled Data	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 mer...	[ "Tiantong Wang", "Yiyang Duan", "Haoyu Chen", "Tiantong Wu", "Wei Yang Bryan Lim" ]	[ "cs.LG" ]	[ "Computer Science" ]	2026-02-09T00:00:00	https://arxiv.org/abs/2602.08564	https://arxiv.org/pdf/2602.08564v1	2602.08564	10.1609/aaai.v40i31.39854	1	0	true	https://github.com/languangduan/mLoss	AAAI Conference on Artificial Intelligence	0.726
0b258e7e0e3659d5a5ebd84a017f3459d1658a407b3bb58fbef072157e616fd7	[ "arxiv", "semantic_scholar" ]	Definability and Interpolation in Philosophy	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 generali...	[ "Johan van Benthem" ]	[ "cs.LO" ]	[ "Computer Science" ]	2026-02-08T00:00:00	https://arxiv.org/abs/2602.07907	https://arxiv.org/pdf/2602.07907v1	2602.07907	10.48550/arXiv.2602.07907	0	0	false	null	arXiv.org	0.4686
e7819db6fdbfdbee89da2a0c4e0238496b8c5b49064b31f30cea7737204e0c7c	[ "arxiv", "semantic_scholar" ]	Fine-Grained Model Merging via Modular Expert Recombination	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 instanc...	[ "Haiyun Qiu", "Xingyu Wu", "Liang Feng", "Kay Chen Tan" ]	[ "cs.LG" ]	[ "Computer Science" ]	2026-02-06T00:00:00	https://arxiv.org/abs/2602.06552	https://arxiv.org/pdf/2602.06552v1	2602.06552	10.48550/arXiv.2602.06552	0	0	false	null	arXiv.org	0.4664
bfe6a4b4d2b0af2b24edfa14388688ea651303824707410cb8f9892310326f57	[ "arxiv", "semantic_scholar" ]	Orthogonal Model Merging	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 a...	[ "Sihan Yang", "Kexuan Shi", "Weiyang Liu" ]	[ "cs.LG" ]	[ "Computer Science" ]	2026-02-05T00:00:00	https://arxiv.org/abs/2602.05943	https://arxiv.org/pdf/2602.05943v1	2602.05943	10.48550/arXiv.2602.05943	1	0	false	null	arXiv.org	0.4652
03cce1c5e4d2cd288c2ab05df72c16cdc45616146d85b9a95803a227ce8e4f89	[ "arxiv", "semantic_scholar" ]	When Shared Knowledge Hurts: Spectral Over-Accumulation in Model Merging	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 t...	[ "Yayuan Li", "Ze Peng", "Jian Zhang", "Jintao Guo", "Yue Duan", "Yinghuan Shi" ]	[ "cs.LG", "cs.AI", "cs.CL", "cs.CV" ]	[ "Computer Science" ]	2026-02-05T00:00:00	https://arxiv.org/abs/2602.05536	https://arxiv.org/pdf/2602.05536v2	2602.05536	10.48550/arXiv.2602.05536	3	0	true	https://github.com/lyymuwu/SVC	arXiv.org	0.719
c660043a721b5b119eaa4310f588f2050d1b82cfe0306898dea8c6d906875faf	[ "arxiv", "semantic_scholar" ]	Model-Dowser: Data-Free Importance Probing to Mitigate Catastrophic Forgetting in Multimodal Large Language Models	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 mitig...	[ "Hyeontaek Hwang", "Nguyen Dinh Son", "Daeyoung Kim" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-02-04T00:00:00	https://arxiv.org/abs/2602.04509	https://arxiv.org/pdf/2602.04509v7	2602.04509	10.48550/arXiv.2602.04509	0	0	false	null	arXiv.org	0.4641
64eef9dd34cb982c11b268d9a395f77d488a6be603f8ec523276114a5859e4e7	[ "arxiv", "semantic_scholar" ]	Self-Soupervision: Cooking Model Soups without Labels	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 lo...	[ "Anthony Fuller", "James R. Green", "Evan Shelhamer" ]	[ "cs.LG" ]	[ "Computer Science" ]	2026-02-02T00:00:00	https://arxiv.org/abs/2602.02890	https://arxiv.org/pdf/2602.02890v2	2602.02890	10.48550/arXiv.2602.02890	0	0	true	https://github.com/antofuller/self_soupervision	arXiv.org	0.7136
85b3ebd467ba4360f31d987da572e76812044e55777a1ab27ea7d40841b7f359	[ "arxiv", "semantic_scholar" ]	AutoMerge: Search-Based Model Merging Framework for Effective Model Reuse	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...	[ "You Lu", "Jiyang Zhang", "Bihuan Chen", "Chaofeng Sha", "Dingji Wang", "Xin Peng" ]	[ "cs.SE" ]	[ "Computer Science" ]	2026-01-30T00:00:00	https://arxiv.org/abs/2601.22748	https://arxiv.org/pdf/2601.22748v1	2601.22748	10.48550/arXiv.2601.22748	0	0	false	null	arXiv.org	0.4583
f2f794df11319183e552813d69f3ab08cb7c7e11c930aaffd41fdd33706ce502	[ "arxiv", "semantic_scholar" ]	Per-parameter Task Arithmetic for Unlearning in Large Language Models	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 ...	[ "Chengyi Cai", "Zesheng Ye", "Jiangchao Yao", "Jianzhong Qi", "Bo Han", "Xiaolu Zhang", "Feng Liu", "Jun Zhou" ]	[ "cs.LG" ]	[ "Computer Science" ]	2026-01-29T00:00:00	https://arxiv.org/abs/2601.22030	https://arxiv.org/pdf/2601.22030v1	2601.22030	10.48550/arXiv.2601.22030	0	0	false	null	arXiv.org	0.4572
73c76cc32d5ab353f4ed6b6d4281fea131e0071f25e5aa5c90962a36908692c2	[ "arxiv", "semantic_scholar" ]	SOUP: Token-level Single-sample Mix-policy Reinforcement Learning for Large Language Models	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 sign...	[ "Lei Yang", "Wei Bi", "Chenxi Sun", "Renren Jin", "Deyi Xiong" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-01-29T00:00:00	https://arxiv.org/abs/2601.21476	https://arxiv.org/pdf/2601.21476v1	2601.21476	10.48550/arXiv.2601.21476	1	0	false	null	arXiv.org	0.4572
1bbe87ba20db336e216b7930120005615a6c26b8f8d4fe98d7dca21b8e94cbc6	[ "arxiv", "semantic_scholar" ]	Multi-task Code LLMs: Data Mix or Model Merge?	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 m...	[ "Mingzhi Zhu", "Boris Sobolev", "Rahul Krishna", "Raju Pavuluri", "Stacy Patterson", "Michele Merler" ]	[ "cs.CL", "cs.AI" ]	[ "Computer Science" ]	2026-01-28T00:00:00	https://arxiv.org/abs/2601.21115	https://arxiv.org/pdf/2601.21115v1	2601.21115	10.48550/arXiv.2601.21115	1	0	false	null	arXiv.org	0.456
32e0c9120fa185cfe7cb9859c407bbcd3ff747aa423e8e2a584cc8e85d3b6556	[ "arxiv", "semantic_scholar" ]	Behavior Knowledge Merge in Reinforced Agentic Models	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 met...	[ "Xiangchi Yuan", "Dachuan Shi", "Chunhui Zhang", "Zheyuan Liu", "Shenglong Yao", "Soroush Vosoughi", "Wenke Lee" ]	[ "cs.LG" ]	[ "Computer Science" ]	2026-01-20T00:00:00	https://arxiv.org/abs/2601.13572	https://arxiv.org/pdf/2601.13572v1	2601.13572	10.48550/arXiv.2601.13572	6	0	false	null	arXiv.org	0.4469
8c344cd2626c159a853f5559148041fbd02b5e3a6786b03c83f5538680a06a44	[ "arxiv", "semantic_scholar" ]	Will it Merge? On The Causes of Model Mergeability	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....	[ "Adir Rahamim", "Asaf Yehudai", "Boaz Carmeli", "Leshem Choshen", "Yosi Mass", "Yonatan Belinkov" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-01-10T00:00:00	https://arxiv.org/abs/2601.06672	https://arxiv.org/pdf/2601.06672v1	2601.06672	10.48550/arXiv.2601.06672	2	0	false	null	arXiv.org	0.4354
60c611a8cf880669efa34be0297a7de91185e586e17895eee264693bf5f928e5	[ "arxiv", "semantic_scholar" ]	Model Merging via Multi-Teacher Knowledge Distillation	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 com...	[ "Seyed Arshan Dalili", "Mehrdad Mahdavi" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2025-12-24T00:00:00	https://arxiv.org/abs/2512.21288	https://arxiv.org/pdf/2512.21288v1	2512.21288	10.48550/arXiv.2512.21288	0	0	true	https://github.com/arshandalili/SAMerging	arXiv.org	0.6428
1b5d0401b1e0b3674ad18977ff1823de10087727f5f161c57ee398e7292ac6d2	[ "arxiv", "semantic_scholar" ]	MAGIC: Achieving Superior Model Merging via Magnitude Calibration	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 mo...	[ "Yayuan Li", "Jian Zhang", "Jintao Guo", "Zihan Cheng", "Lei Qi", "Yinghuan Shi", "Yang Gao" ]	[ "cs.LG", "cs.AI", "cs.CL", "cs.CV" ]	[ "Computer Science" ]	2025-12-22T00:00:00	https://arxiv.org/abs/2512.19320	https://arxiv.org/pdf/2512.19320v1	2512.19320	10.48550/arXiv.2512.19320	1	0	true	https://github.com/lyymuwu/MAGIC	arXiv.org	0.6393
9fd992bbeca9afbab3147779a5ff8af5ff92576ee4456e9f814bbcf15d758bbc	[ "arxiv", "semantic_scholar" ]	Uniform Interpolation	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 definabi...	[ "Sam van Gool" ]	[ "math.LO", "cs.LO" ]	[ "Computer Science", "Mathematics" ]	2025-12-17T00:00:00	https://arxiv.org/abs/2512.15391	https://arxiv.org/pdf/2512.15391v3	2512.15391	10.48550/arXiv.2512.15391	3	0	false	null	arXiv.org	0.4079
bf66ba2d409e7fa116abe1d3dd3dcee8a47426a45c77acd8bf42e007c4d00ee7	[ "arxiv", "semantic_scholar" ]	Per-Axis Weight Deltas for Frequent Model Updates	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 ...	[ "Stefan Kuyumdzhiev", "Radostin Cholakov" ]	[ "cs.LG" ]	[ "Computer Science" ]	2025-12-16T00:00:00	https://arxiv.org/abs/2512.19720	https://arxiv.org/pdf/2512.19720v1	2512.19720	10.48550/arXiv.2512.19720	0	0	true	https://github.com/kuiumdjiev/Per-Axis-Weight-Deltas-for-Frequent-Model-Updates	arXiv.org	0.6286
11b40cfbec786eafba8795f92876a45ffc4ac6f78049ede64d82dfe00782309d	[ "arxiv", "semantic_scholar" ]	Interpolation in Knowledge Representation	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 in...	[ "Jean Christoph Jung", "Patrick Koopmann", "Matthias Knorr" ]	[ "cs.AI", "cs.LO" ]	[ "Computer Science" ]	2025-12-09T00:00:00	https://arxiv.org/abs/2512.08833	https://arxiv.org/pdf/2512.08833v1	2512.08833	10.48550/arXiv.2512.08833	4	0	false	null	arXiv.org	0.3987
f9a84988676292f9bdeb0de68ce40ee993dad14ba5d8aa9cb77eee7f337c2297	[ "arxiv", "semantic_scholar" ]	An Empirical Survey of Model Merging Algorithms for Social Bias Mitigation	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 empirica...	[ "Daiki Shirafuji", "Tatsuhiko Saito", "Yasutomo Kimura" ]	[ "cs.CL", "cs.AI" ]	[ "Computer Science" ]	2025-12-02T00:00:00	https://arxiv.org/abs/2512.02689	https://arxiv.org/pdf/2512.02689v1	2512.02689	10.48550/arXiv.2512.02689	0	0	false	null	Pacific Asia Conference on Language, Information and Computation	0.3907
2f154303d35a995619f92c14c3a0c2fe90b167e7f511e42636cc057cb775ddeb	[ "arxiv", "semantic_scholar" ]	Interpolation in Non-Classical Logics	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...	[ "Wesley Fussner" ]	[ "math.LO", "cs.LO" ]	[ "Mathematics", "Computer Science" ]	2025-12-01T00:00:00	https://arxiv.org/abs/2512.01600	https://arxiv.org/pdf/2512.01600v1	2512.01600	10.48550/arXiv.2512.01600	0	0	false	null	arXiv.org	0.3896
0c5159109d17abf48e407b94f9ab445211d9b7ed1e2fd79040f42f292d2b45da	[ "arxiv", "semantic_scholar" ]	Stay Unique, Stay Efficient: Preserving Model Personality in Multi-Task Merging	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 informat...	[ "Kuangpu Guo", "Yuhe Ding", "Jian Liang", "Zilei Wang", "Ran He" ]	[ "cs.LG", "cs.CV" ]	[ "Computer Science" ]	2025-12-01T00:00:00	https://arxiv.org/abs/2512.01461	https://arxiv.org/pdf/2512.01461v1	2512.01461	10.48550/arXiv.2512.01461	1	0	true	https://github.com/krumpguo/DTS	arXiv.org	0.6021
6f68ef8a58ff5ee806cb91118686181058b1796afffc9f680e7f6c167524398b	[ "arxiv", "semantic_scholar" ]	Merge and Bound: Direct Manipulations on Weights for Class Incremental Learning	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...	[ "Taehoon Kim", "Donghwan Jang", "Bohyung Han" ]	[ "cs.CV", "cs.AI", "cs.LG" ]	[ "Computer Science" ]	2025-11-26T00:00:00	https://arxiv.org/abs/2511.21490	https://arxiv.org/pdf/2511.21490v1	2511.21490	10.48550/arXiv.2511.21490	0	0	false	null	arXiv.org	0.3839
1c5a372283c92d5cc4d8f0c5bc464528373ef5216f69c37f36f25e1a8465a520	[ "arxiv", "semantic_scholar" ]	A Systematic Study of In-the-Wild Model Merging for Large Language Models	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 t...	[ "Oğuz Kağan Hitit", "Leander Girrbach", "Zeynep Akata" ]	[ "cs.CL", "cs.LG" ]	[ "Computer Science" ]	2025-11-26T00:00:00	https://arxiv.org/abs/2511.21437	https://arxiv.org/pdf/2511.21437v2	2511.21437	null	2	0	false	null	Transactions on Machine Learning Research (03/2026)	0.3839
88c66ad8c9d2d8aac7aa6523e2ab20ae18e196dacefd68c80362764eb6bd514d	[ "arxiv", "semantic_scholar" ]	Merging without Forgetting: Continual Fusion of Task-Specific Models via Optimal Transport	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...	[ "Zecheng Pan", "Zhikang Chen", "Ding Li", "Min Zhang", "Sen Cui", "Hongshuo Jin", "Luqi Tao", "Yi Yang", "Deheng Ye", "Yu Zhang", "Tingting Zhu", "Tianling Ren" ]	[ "cs.LG", "cs.AI", "cs.CV" ]	[ "Computer Science" ]	2025-11-24T00:00:00	https://arxiv.org/abs/2511.19561	https://arxiv.org/pdf/2511.19561v1	2511.19561	10.48550/arXiv.2511.19561	0	0	false	null	arXiv.org	0.3816
22bccf603496fbbc62e67e95feede8f7aa8aad425ff061d8b955282a518080df	[ "arxiv", "semantic_scholar" ]	Escaping Optimization Stagnation: Taking Steps Beyond Task Arithmetic via Difference Vectors	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-tune...	[ "Jinping Wang", "Zhiqiang Gao", "Dinggen Zhang", "Zhiwu Xie" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2025-11-22T00:00:00	https://arxiv.org/abs/2511.17987	https://arxiv.org/pdf/2511.17987v1	2511.17987	10.48550/arXiv.2511.17987	0	0	false	null	AAAI Conference on Artificial Intelligence	0.3793
f7ece0a771f4beea649c8d8a2c8fac58c1def5e7d6fc5618e35828cd3666cd75	[ "arxiv", "semantic_scholar" ]	Task Addition and Weight Disentanglement in Closed-Vocabulary Models	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 t...	[ "Adam Hazimeh", "Alessandro Favero", "Pascal Frossard" ]	[ "cs.LG" ]	[ "Computer Science" ]	2025-11-18T00:00:00	https://arxiv.org/abs/2511.14569	https://arxiv.org/pdf/2511.14569v1	2511.14569	10.48550/arXiv.2511.14569	4	0	false	null	arXiv.org	0.3747
956fb0caa34a99fb85a2586c42a37fbf4c566a25a1ef054bd3d390aada0ea8cd	[ "arxiv", "semantic_scholar" ]	MergeSlide: Continual Model Merging and Task-to-Class Prompt-Aligned Inference for Lifelong Learning on Whole Slide Images	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 fr...	[ "Doanh C. Bui", "Ba Hung Ngo", "Hoai Luan Pham", "Khang Nguyen", "Maï K. Nguyen", "Yasuhiko Nakashima" ]	[ "cs.CV" ]	[ "Computer Science" ]	2025-11-17T00:00:00	https://arxiv.org/abs/2511.13099	https://arxiv.org/pdf/2511.13099v1	2511.13099	10.1109/WACV61042.2026.00472	0	0	true	https://github.com/caodoanh2001/MergeSlide	IEEE Workshop/Winter Conference on Applications of Computer Vision	0.5773
1f43f2ec73aaf6558699cfbe33e8e51dd0cb00f6c2239d0d043640eda523af40	[ "arxiv", "semantic_scholar" ]	A Novel Hierarchical Integration Method for Efficient Model Merging in Medical LLMs	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...	[ "Prakrit Timilsina", "Anuj Nepal", "Rajan Kadel", "Robin Doss" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2025-11-17T00:00:00	https://arxiv.org/abs/2511.13373	https://arxiv.org/pdf/2511.13373v1	2511.13373	10.1109/SmartIoT66867.2025.00031	0	0	false	null	International Conferences on Smart Internet of Things	0.3735
2bf4488fb9d4c127e6f476c66ff2a1d8a566046d75fbde200fb561814b74f6e2	[ "arxiv", "semantic_scholar" ]	Defending Unauthorized Model Merging via Dual-Stage Weight Protection	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 un...	[ "Wei-Jia Chen", "Min-Yen Tsai", "Cheng-Yi Lee", "Chia-Mu Yu" ]	[ "cs.CV", "cs.CR" ]	[ "Computer Science" ]	2025-11-14T00:00:00	https://arxiv.org/abs/2511.11851	https://arxiv.org/pdf/2511.11851v3	2511.11851	10.48550/arXiv.2511.11851	0	0	false	null	arXiv.org	0.3701
af45a91f5eea018489d72d95ab3226dd0d3c129796702912eba7598774ca41a1	[ "arxiv", "semantic_scholar" ]	Do Not Merge My Model! Safeguarding Open-Source LLMs Against Unauthorized Model Merging	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 ...	[ "Qinfeng Li", "Miao Pan", "Jintao Chen", "Fu Teng", "Zhiqiang Shen", "Ge Su", "Hao Peng", "Xuhong Zhang" ]	[ "cs.CR", "cs.AI" ]	[ "Computer Science" ]	2025-11-13T00:00:00	https://arxiv.org/abs/2511.10712	https://arxiv.org/pdf/2511.10712v2	2511.10712	10.48550/arXiv.2511.10712	1	0	true	null	AAAI Conference on Artificial Intelligence	0.5702
b22815202aa071278bc54ea9838a5d2caf6242a66e15042a8478e74146fe6059	[ "arxiv", "semantic_scholar" ]	LT-Soups: Bridging Head and Tail Classes via Subsampled Model Soups	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 ...	[ "Masih Aminbeidokhti", "Subhankar Roy", "Eric Granger", "Elisa Ricci", "Marco Pedersoli" ]	[ "cs.LG", "cs.CV" ]	[ "Computer Science" ]	2025-11-11T00:00:00	https://arxiv.org/abs/2511.10683	https://arxiv.org/pdf/2511.10683v2	2511.10683	10.48550/arXiv.2511.10683	1	0	false	null	arXiv.org	0.3667
d880ec144875557a21089ac082477e66e598253f377293425d17a849f77b811f	[ "arxiv", "semantic_scholar" ]	Steering Language Models with Weight Arithmetic	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...	[ "Constanza Fierro", "Fabien Roger" ]	[ "cs.CL", "cs.LG" ]	[ "Computer Science" ]	2025-11-07T00:00:00	https://arxiv.org/abs/2511.05408	https://arxiv.org/pdf/2511.05408v2	2511.05408	10.48550/arXiv.2511.05408	8	0	false	null	arXiv.org	0.3621
6baa54d76f87e54f2662c813a06ce971062d0e3752c096e81291d0fe3fe31fc5	[ "arxiv", "semantic_scholar" ]	Polarization-resolved imaging improves eye tracking	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...	[ "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" ]	[ "cs.CV", "physics.optics" ]	[ "Computer Science", "Physics" ]	2025-11-06T00:00:00	https://arxiv.org/abs/2511.04652	https://arxiv.org/pdf/2511.04652v1	2511.04652	10.48550/arXiv.2511.04652	2	0	false	null	arXiv.org	0.3609
2b8e739be17d954bee543be8b12ee72c414cf3d7de582af80fc3fea2784ea881	[ "arxiv", "semantic_scholar" ]	Human Mesh Modeling for Anny Body	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 knowl...	[ "Romain Brégier", "Guénolé Fiche", "Laura Bravo-Sánchez", "Thomas Lucas", "Matthieu Armando", "Philippe Weinzaepfel", "Grégory Rogez", "Fabien Baradel" ]	[ "cs.CV" ]	[ "Computer Science" ]	2025-11-05T00:00:00	https://arxiv.org/abs/2511.03589	https://arxiv.org/pdf/2511.03589v2	2511.03589	10.48550/arXiv.2511.03589	7	1	true	https://github.com/naver/anny	arXiv.org	0.556
ede1f48e9942d46264fb042bbf0f8ee397dab10fd9e5d6584387fb2688ea5290	[ "arxiv", "semantic_scholar" ]	Merging Continual Pretraining Models for Domain-Specialized LLMs: A Case Study in Finance	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 traini...	[ "Kentaro Ueda", "François Portet", "Hirohiko Suwa", "Keiichi Yasumoto" ]	[ "cs.CL" ]	[ "Computer Science" ]	2025-11-04T00:00:00	https://arxiv.org/abs/2511.02451	https://arxiv.org/pdf/2511.02451v1	2511.02451	10.48550/arXiv.2511.02451	1	0	false	null	arXiv.org	0.3586
47ac3b07fb2499582dfff7d95e28fd2e07aeca76184b577ccddb163070e425cd	[ "arxiv", "semantic_scholar" ]	Model Merging with Functional Dual Anchors	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 ...	[ "Kexuan Shi", "Yandong Wen", "Weiyang Liu" ]	[ "cs.LG" ]	[ "Computer Science" ]	2025-10-24T00:00:00	https://arxiv.org/abs/2510.21223	https://arxiv.org/pdf/2510.21223v1	2510.21223	10.48550/arXiv.2510.21223	3	0	false	null	arXiv.org	0.346
d64abf2f883e84edbd93e6ffcbb7004e901cb32b901d2419651bb932038ea02e	[ "arxiv", "semantic_scholar" ]	Capability Ceilings in Autoregressive Language Models: Empirical Evidence from Knowledge-Intensive Tasks	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 lo...	[ "Javier Marín" ]	[ "cs.AI" ]	[ "Computer Science" ]	2025-10-23T00:00:00	https://arxiv.org/abs/2510.21866	https://arxiv.org/pdf/2510.21866v1	2510.21866	10.48550/arXiv.2510.21866	1	0	false	null	arXiv.org	0.3449
a371c173fa7ec3b87f668b9288e7f8108ee524d52fa892cfb2a1314b6cc860b2	[ "arxiv", "semantic_scholar" ]	Adapting Multilingual Models to Code-Mixed Tasks via Model Merging	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 (...	[ "Prashant Kodali", "Vaishnavi Shivkumar", "Swarang Joshi", "Monojit Choudhary", "Ponnurangam Kumaraguru", "Manish Shrivastava" ]	[ "cs.CL" ]	[ "Computer Science" ]	2025-10-22T00:00:00	https://arxiv.org/abs/2510.19782	https://arxiv.org/pdf/2510.19782v2	2510.19782	10.1145/3799830.3799852	4	0	false	null	ACM IKDD Conference on Data Science	0.3438
9fb80a9fc4b1c2f256df22e7efd05db1e411dd68e6ca20e8419a143f9b10ee83	[ "arxiv", "semantic_scholar" ]	MIN-Merging: Merge the Important Neurons for Model Merging	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 fr...	[ "Yunfei Liang" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2025-10-18T00:00:00	https://arxiv.org/abs/2510.17890	https://arxiv.org/pdf/2510.17890v2	2510.17890	10.48550/arXiv.2510.17890	0	0	true	null	arXiv.org	0.5242
61d7f5e295921aede5c87def6017b2dbbf9707fc1eb37a9d371646715d099d3f	[ "arxiv", "semantic_scholar" ]	Six Proofs of Interpolation for the Modal Logic K	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.	[ "Nick Bezhanishvili", "Balder ten Cate", "Rosalie Iemhoff" ]	[ "cs.LO" ]	[ "Computer Science" ]	2025-10-18T00:00:00	https://arxiv.org/abs/2510.16398	https://arxiv.org/pdf/2510.16398v2	2510.16398	10.48550/arXiv.2510.16398	3	1	false	null	arXiv.org	0.3392
ea237e51da3028808d7c6564f34fc44f384387262adc7728917615cdd4572b17	[ "arxiv", "semantic_scholar" ]	Purifying Task Vectors in Knowledge-Aware Subspace for Model Merging	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 su...	[ "Bang An", "Yibo Yang", "Philip Torr", "Bernard Ghanem" ]	[ "cs.AI" ]	[ "Computer Science" ]	2025-10-16T00:00:00	https://arxiv.org/abs/2510.14697	https://arxiv.org/pdf/2510.14697v1	2510.14697	10.48550/arXiv.2510.14697	1	0	false	null	arXiv.org	0.3369
e93a218744a82719bfd5f42659b24f6e33d27f1db5eff5733af69c263fbfdf5e	[ "arxiv", "semantic_scholar" ]	Weight Weaving: Parameter Pooling for Data-Free Model Merging	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 we...	[ "Levy Chaves", "Eduardo Valle", "Sandra Avila" ]	[ "cs.LG", "cs.CV" ]	[ "Computer Science" ]	2025-10-15T00:00:00	https://arxiv.org/abs/2510.13921	https://arxiv.org/pdf/2510.13921v1	2510.13921	10.48550/arXiv.2510.13921	1	0	false	null	arXiv.org	0.3357
a051feaeee8ec3dc8cfce5862561ca5ab9ab10e9ee8eef1aef83f8a9e14bac46	[ "arxiv", "semantic_scholar" ]	Towards Reversible Model Merging For Low-rank Weights	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-ran...	[ "Mohammadsajad Alipour", "Mohammad Mohammadi Amiri" ]	[ "cs.LG", "cs.AI", "cs.CL", "cs.CV" ]	[ "Computer Science" ]	2025-10-15T00:00:00	https://arxiv.org/abs/2510.14163	https://arxiv.org/pdf/2510.14163v1	2510.14163	10.48550/arXiv.2510.14163	1	0	false	null	arXiv.org	0.3357
835188ea4527da006fce8aa6733da3907118dcab5104a73393936a9774a2740c	[ "arxiv", "semantic_scholar" ]	Revisiting Model Interpolation for Efficient Reasoning	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 w...	[ "Taiqiang Wu", "Runming Yang", "Tao Liu", "Jiahao Wang", "Ngai Wong" ]	[ "cs.AI", "cs.CL" ]	[ "Computer Science" ]	2025-10-13T00:00:00	https://arxiv.org/abs/2510.10977	https://arxiv.org/pdf/2510.10977v2	2510.10977	10.48550/arXiv.2510.10977	8	0	true	https://github.com/wutaiqiang/MI}{Github}	arXiv.org	0.5153
5c3a2e7ffe28bf88ce70621a4e881e8fabf60d2ccabc3f4141273d4e1fe55bd4	[ "arxiv", "semantic_scholar" ]	Generative World Modelling for Humanoids: 1X World Model Challenge Technical Report	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 forecas...	[ "Riccardo Mereu", "Aidan Scannell", "Yuxin Hou", "Yi Zhao", "Aditya Jitta", "Antonio Dominguez", "Luigi Acerbi", "Amos Storkey", "Paul Chang" ]	[ "cs.LG", "cs.AI", "cs.RO" ]	[ "Computer Science" ]	2025-10-08T00:00:00	https://arxiv.org/abs/2510.07092	https://arxiv.org/pdf/2510.07092v1	2510.07092	10.48550/arXiv.2510.07092	5	0	true	null	arXiv.org	0.5065

End of preview. Expand in Data Studio

Model Merging Papers — FineSet

A research-paper dataset on Model Merging Papers, assembled, deduplicated, and quality-scored by FineSet from arXiv and Semantic Scholar.

📸 This is a dated snapshot — generated 2026-06-19. It is not auto-updated. Research on Model Merging Papers moves fast — new papers land on arXiv every week. Want this same dataset refreshed daily, on a topic you choose? See the bottom. ↓

Why this dataset

Quality-scored: quality_score float (0–1), blends citations with recency + code/venue signals — filter out the noise
Papers with code: 109 flagged via has_code — find reproducible work fast
Deduplicated: arXiv + Semantic Scholar cross-referenced, duplicate records merged
Clean JSONL: 381 records, one per line, normalized fields — no encoding garbage

Dataset details

Records: 381
Date range: 2021–2026
Snapshot date: 2026-06-19 (frozen — see note above)
Sources: arXiv, Semantic Scholar (cross-referenced, duplicates merged)
arXiv categories: cs.LG, cs.CL
Quality scoring: citations + recency + code/venue blend, 0–1 (p50=0.32, p90=0.602)
Format: JSONL, one record per line

Fields

Field	Type	Description
id	string	Deterministic SHA256 record id
sources	list	Which sources contributed (`arxiv`, `semantic_scholar`)
title	string	Paper title
abstract	string	Full abstract
authors	list	Author names
categories	list	arXiv category codes
fields_of_study	list	Semantic Scholar field tags
published_date	string	ISO 8601 date
url	string	arXiv abstract URL
pdf_url	string\|null	Open-access PDF if available
arxiv_id	string\|null	arXiv identifier
doi	string\|null	DOI if available
citation_count	int	Citation count (Semantic Scholar)
influential_citation_count	int	Influential citations (Semantic Scholar)
has_code	bool	Code repo detected in the arXiv comment
code_url	string\|null	GitHub URL if detected
venue	string\|null	Publication venue
quality_score	float	0–1, blended (citations + recency + code/venue)

Quality score methodology

quality_score = max(impact, freshness), clamped to [0, 1], where:

impact = max( log10(citations+1)/4 , log10(influential_citations+1)/2 ) — realized impact (0.5 at 100 citations, ~0.75 at 1,000, 1.0 at 10,000+).
freshness = recency × (0.35 + 0.30·has_code + 0.20·has_venue) — a baseline for recent papers (so a strong paper published this week isn't scored 0 just for lacking citations), where recency is 1.0 for papers ≤60 days old and decays linearly to 0 by ~18 months.

Old highly-cited papers score on impact; brand-new papers score on freshness; old uncited papers score ~0. Useful for filtering training data by quality, not just age.

👉 Want this on YOUR topic, updated daily?

This snapshot is frozen at 2026-06-19. The live FineSet pipeline keeps a dataset like this refreshed every day on whatever topic you describe — new papers in, dedup and quality scoring automatic, export as JSONL/Parquet or push straight to the Hub.

Tell me the topic you'd want and I'll run the pipeline on it — open a discussion on this dataset, it's free and it's how I decide what to build next.

→ fineset.io — describe what you want to train on, get a dataset. Early-access waitlist open (referral skip available).

Downloads last month: 39