id
stringlengths
9
16
title
stringlengths
4
278
categories
stringlengths
5
104
abstract
stringlengths
6
4.09k
2502.04356
Open Foundation Models in Healthcare: Challenges, Paradoxes, and Opportunities with GenAI Driven Personalized Prescription
cs.CL cs.AI cs.LG
In response to the success of proprietary Large Language Models (LLMs) such as OpenAI's GPT-4, there is a growing interest in developing open, non-proprietary LLMs and AI foundation models (AIFMs) for transparent use in academic, scientific, and non-commercial applications. Despite their inability to match the refined functionalities of their proprietary counterparts, open models hold immense potential to revolutionize healthcare applications. In this paper, we examine the prospects of open-source LLMs and AIFMs for developing healthcare applications and make two key contributions. Firstly, we present a comprehensive survey of the current state-of-the-art open-source healthcare LLMs and AIFMs and introduce a taxonomy of these open AIFMs, categorizing their utility across various healthcare tasks. Secondly, to evaluate the general-purpose applications of open LLMs in healthcare, we present a case study on personalized prescriptions. This task is particularly significant due to its critical role in delivering tailored, patient-specific medications that can greatly improve treatment outcomes. In addition, we compare the performance of open-source models with proprietary models in settings with and without Retrieval-Augmented Generation (RAG). Our findings suggest that, although less refined, open LLMs can achieve performance comparable to proprietary models when paired with grounding techniques such as RAG. Furthermore, to highlight the clinical significance of LLMs-empowered personalized prescriptions, we perform subjective assessment through an expert clinician. We also elaborate on ethical considerations and potential risks associated with the misuse of powerful LLMs and AIFMs, highlighting the need for a cautious and responsible implementation in healthcare.
2502.04357
Reusing Embeddings: Reproducible Reward Model Research in Large Language Model Alignment without GPUs
cs.CL cs.AI cs.LG
Large Language Models (LLMs) have made substantial strides in structured tasks through Reinforcement Learning (RL), demonstrating proficiency in mathematical reasoning and code generation. However, applying RL in broader domains like chatbots and content generation -- through the process known as Reinforcement Learning from Human Feedback (RLHF) -- presents unique challenges. Reward models in RLHF are critical, acting as proxies that evaluate the alignment of LLM outputs with human intent. Despite advancements, the development of reward models is hindered by challenges such as computational heavy training, costly evaluation, and therefore poor reproducibility. We advocate for using embedding-based input in reward model research as an accelerated solution to those challenges. By leveraging embeddings for reward modeling, we can enhance reproducibility, reduce computational demands on hardware, improve training stability, and significantly reduce training and evaluation costs, hence facilitating fair and efficient comparisons in this active research area. We then show a case study of reproducing existing reward model ensemble research using embedding-based reward models. We discussed future avenues for research, aiming to contribute to safer and more effective LLM deployments.
2502.04358
Position: Scaling LLM Agents Requires Asymptotic Analysis with LLM Primitives
cs.CL cs.AI cs.CC cs.LG cs.NE
Decomposing hard problems into subproblems often makes them easier and more efficient to solve. With large language models (LLMs) crossing critical reliability thresholds for a growing slate of capabilities, there is an increasing effort to decompose systems into sets of LLM-based agents, each of whom can be delegated sub-tasks. However, this decomposition (even when automated) is often intuitive, e.g., based on how a human might assign roles to members of a human team. How close are these role decompositions to optimal? This position paper argues that asymptotic analysis with LLM primitives is needed to reason about the efficiency of such decomposed systems, and that insights from such analysis will unlock opportunities for scaling them. By treating the LLM forward pass as the atomic unit of computational cost, one can separate out the (often opaque) inner workings of a particular LLM from the inherent efficiency of how a set of LLMs are orchestrated to solve hard problems. In other words, if we want to scale the deployment of LLMs to the limit, instead of anthropomorphizing LLMs, asymptotic analysis with LLM primitives should be used to reason about and develop more powerful decompositions of large problems into LLM agents.
2502.04359
Exploring Spatial Language Grounding Through Referring Expressions
cs.CL cs.AI cs.CV
Spatial Reasoning is an important component of human cognition and is an area in which the latest Vision-language models (VLMs) show signs of difficulty. The current analysis works use image captioning tasks and visual question answering. In this work, we propose using the Referring Expression Comprehension task instead as a platform for the evaluation of spatial reasoning by VLMs. This platform provides the opportunity for a deeper analysis of spatial comprehension and grounding abilities when there is 1) ambiguity in object detection, 2) complex spatial expressions with a longer sentence structure and multiple spatial relations, and 3) expressions with negation ('not'). In our analysis, we use task-specific architectures as well as large VLMs and highlight their strengths and weaknesses in dealing with these specific situations. While all these models face challenges with the task at hand, the relative behaviors depend on the underlying models and the specific categories of spatial semantics (topological, directional, proximal, etc.). Our results highlight these challenges and behaviors and provide insight into research gaps and future directions.
2502.04360
MARAGE: Transferable Multi-Model Adversarial Attack for Retrieval-Augmented Generation Data Extraction
cs.CL cs.CR cs.LG
Retrieval-Augmented Generation (RAG) offers a solution to mitigate hallucinations in Large Language Models (LLMs) by grounding their outputs to knowledge retrieved from external sources. The use of private resources and data in constructing these external data stores can expose them to risks of extraction attacks, in which attackers attempt to steal data from these private databases. Existing RAG extraction attacks often rely on manually crafted prompts, which limit their effectiveness. In this paper, we introduce a framework called MARAGE for optimizing an adversarial string that, when appended to user queries submitted to a target RAG system, causes outputs containing the retrieved RAG data verbatim. MARAGE leverages a continuous optimization scheme that integrates gradients from multiple models with different architectures simultaneously to enhance the transferability of the optimized string to unseen models. Additionally, we propose a strategy that emphasizes the initial tokens in the target RAG data, further improving the attack's generalizability. Evaluations show that MARAGE consistently outperforms both manual and optimization-based baselines across multiple LLMs and RAG datasets, while maintaining robust transferability to previously unseen models. Moreover, we conduct probing tasks to shed light on the reasons why MARAGE is more effective compared to the baselines and to analyze the impact of our approach on the model's internal state.
2502.04361
Predicting 3D Motion from 2D Video for Behavior-Based VR Biometrics
cs.CV cs.AI cs.HC
Critical VR applications in domains such as healthcare, education, and finance that use traditional credentials, such as PIN, password, or multi-factor authentication, stand the chance of being compromised if a malicious person acquires the user credentials or if the user hands over their credentials to an ally. Recently, a number of approaches on user authentication have emerged that use motions of VR head-mounted displays (HMDs) and hand controllers during user interactions in VR to represent the user's behavior as a VR biometric signature. One of the fundamental limitations of behavior-based approaches is that current on-device tracking for HMDs and controllers lacks capability to perform tracking of full-body joint articulation, losing key signature data encapsulated by the user articulation. In this paper, we propose an approach that uses 2D body joints, namely shoulder, elbow, wrist, hip, knee, and ankle, acquired from the right side of the participants using an external 2D camera. Using a Transformer-based deep neural network, our method uses the 2D data of body joints that are not tracked by the VR device to predict past and future 3D tracks of the right controller, providing the benefit of augmenting 3D knowledge in authentication. Our approach provides a minimum equal error rate (EER) of 0.025, and a maximum EER drop of 0.040 over prior work that uses single-unit 3D trajectory as the input.
2502.04362
LLMs can be easily Confused by Instructional Distractions
cs.CL cs.AI
Despite the fact that large language models (LLMs) show exceptional skill in instruction following tasks, this strength can turn into a vulnerability when the models are required to disregard certain instructions. Instruction-following tasks typically involve a clear task description and input text containing the target data to be processed. However, when the input itself resembles an instruction, confusion may arise, even if there is explicit prompting to distinguish between the task instruction and the input. We refer to this phenomenon as instructional distraction. In this paper, we introduce a novel benchmark, named DIM-Bench, specifically designed to assess LLMs' performance under instructional distraction. The benchmark categorizes real-world instances of instructional distraction and evaluates LLMs across four instruction tasks: rewriting, proofreading, translation, and style transfer -- alongside five input tasks: reasoning, code generation, mathematical reasoning, bias detection, and question answering. Our experimental results reveal that even the most advanced LLMs are susceptible to instructional distraction, often failing to accurately follow user intent in such cases.
2502.04363
On-device Sora: Enabling Diffusion-Based Text-to-Video Generation for Mobile Devices
cs.CV
We present On-device Sora, a first pioneering solution for diffusion-based on-device text-to-video generation that operates efficiently on smartphone-grade devices. Building on Open-Sora, On-device Sora applies three novel techniques to address the challenges of diffusion-based text-to-video generation on computation- and memory-limited mobile devices. First, Linear Proportional Leap (LPL) reduces the excessive denoising steps required in video diffusion through an efficient leap-based approach. Second, Temporal Dimension Token Merging (TDTM) minimizes intensive token-processing computation in attention layers by merging consecutive tokens along the temporal dimension. Third, Concurrent Inference with Dynamic Loading (CI-DL) dynamically partitions large models into smaller blocks and loads them into memory for concurrent model inference, effectively addressing the challenges of limited device memory. We implement On-device Sora on the iPhone 15 Pro, and the experimental evaluations demonstrate that it is capable of generating high-quality videos on the device, comparable to those produced by Open-Sora running on high-end GPUs. These results show that On-device Sora enables efficient and high-quality video generation on resource-constrained mobile devices, expanding accessibility, ensuring user privacy, reducing dependence on cloud infrastructure, and lowering associated costs. We envision the proposed On-device Sora as a significant first step toward democratizing state-of-the-art generative technologies, enabling video generation capabilities on commodity mobile and embedded devices. The code implementation is publicly available at an GitHub repository: https://github.com/eai-lab/On-device-Sora.
2502.04364
Lost in Edits? A $\lambda$-Compass for AIGC Provenance
cs.CV cs.AI cs.HC cs.LG
Recent advancements in diffusion models have driven the growth of text-guided image editing tools, enabling precise and iterative modifications of synthesized content. However, as these tools become increasingly accessible, they also introduce significant risks of misuse, emphasizing the critical need for robust attribution methods to ensure content authenticity and traceability. Despite the creative potential of such tools, they pose significant challenges for attribution, particularly in adversarial settings where edits can be layered to obscure an image's origins. We propose LambdaTracer, a novel latent-space attribution method that robustly identifies and differentiates authentic outputs from manipulated ones without requiring any modifications to generative or editing pipelines. By adaptively calibrating reconstruction losses, LambdaTracer remains effective across diverse iterative editing processes, whether automated through text-guided editing tools such as InstructPix2Pix and ControlNet or performed manually with editing software such as Adobe Photoshop. Extensive experiments reveal that our method consistently outperforms baseline approaches in distinguishing maliciously edited images, providing a practical solution to safeguard ownership, creativity, and credibility in the open, fast-evolving AI ecosystems.
2502.04365
AI-Based Thermal Video Analysis in Privacy-Preserving Healthcare: A Case Study on Detecting Time of Birth
cs.CV cs.AI
Approximately 10% of newborns need some assistance to start breathing and 5\% proper ventilation. It is crucial that interventions are initiated as soon as possible after birth. Accurate documentation of Time of Birth (ToB) is thereby essential for documenting and improving newborn resuscitation performance. However, current clinical practices rely on manual recording of ToB, typically with minute precision. In this study, we present an AI-driven, video-based system for automated ToB detection using thermal imaging, designed to preserve the privacy of healthcare providers and mothers by avoiding the use of identifiable visual data. Our approach achieves 91.4% precision and 97.4% recall in detecting ToB within thermal video clips during performance evaluation. Additionally, our system successfully identifies ToB in 96% of test cases with an absolute median deviation of 1 second compared to manual annotations. This method offers a reliable solution for improving ToB documentation and enhancing newborn resuscitation outcomes.
2502.04366
Contrastive Token-level Explanations for Graph-based Rumour Detection
cs.CL cs.AI cs.LG
The widespread use of social media has accelerated the dissemination of information, but it has also facilitated the spread of harmful rumours, which can disrupt economies, influence political outcomes, and exacerbate public health crises, such as the COVID-19 pandemic. While Graph Neural Network (GNN)-based approaches have shown significant promise in automated rumour detection, they often lack transparency, making their predictions difficult to interpret. Existing graph explainability techniques fall short in addressing the unique challenges posed by the dependencies among feature dimensions in high-dimensional text embeddings used in GNN-based models. In this paper, we introduce Contrastive Token Layerwise Relevance Propagation (CT-LRP), a novel framework designed to enhance the explainability of GNN-based rumour detection. CT-LRP extends current graph explainability methods by providing token-level explanations that offer greater granularity and interpretability. We evaluate the effectiveness of CT-LRP across multiple GNN models trained on three publicly available rumour detection datasets, demonstrating that it consistently produces high-fidelity, meaningful explanations, paving the way for more robust and trustworthy rumour detection systems.
2502.04367
Hybrid Deep Learning Framework for Classification of Kidney CT Images: Diagnosis of Stones, Cysts, and Tumors
eess.IV cs.CV cs.LG
Medical image classification is a vital research area that utilizes advanced computational techniques to improve disease diagnosis and treatment planning. Deep learning models, especially Convolutional Neural Networks (CNNs), have transformed this field by providing automated and precise analysis of complex medical images. This study introduces a hybrid deep learning model that integrates a pre-trained ResNet101 with a custom CNN to classify kidney CT images into four categories: normal, stone, cyst, and tumor. The proposed model leverages feature fusion to enhance classification accuracy, achieving 99.73% training accuracy and 100% testing accuracy. Using a dataset of 12,446 CT images and advanced feature mapping techniques, the hybrid CNN model outperforms standalone ResNet101. This architecture delivers a robust and efficient solution for automated kidney disease diagnosis, providing improved precision, recall, and reduced testing time, making it highly suitable for clinical applications.
2502.04369
HSI: A Holistic Style Injector for Arbitrary Style Transfer
cs.CV cs.LG eess.IV
Attention-based arbitrary style transfer methods have gained significant attention recently due to their impressive ability to synthesize style details. However, the point-wise matching within the attention mechanism may overly focus on local patterns such that neglect the remarkable global features of style images. Additionally, when processing large images, the quadratic complexity of the attention mechanism will bring high computational load. To alleviate above problems, we propose Holistic Style Injector (HSI), a novel attention-style transformation module to deliver artistic expression of target style. Specifically, HSI performs stylization only based on global style representation that is more in line with the characteristics of style transfer, to avoid generating local disharmonious patterns in stylized images. Moreover, we propose a dual relation learning mechanism inside the HSI to dynamically render images by leveraging semantic similarity in content and style, ensuring the stylized images preserve the original content and improve style fidelity. Note that the proposed HSI achieves linear computational complexity because it establishes feature mapping through element-wise multiplication rather than matrix multiplication. Qualitative and quantitative results demonstrate that our method outperforms state-of-the-art approaches in both effectiveness and efficiency.
2502.04370
DreamDPO: Aligning Text-to-3D Generation with Human Preferences via Direct Preference Optimization
cs.CL cs.GR cs.LG
Text-to-3D generation automates 3D content creation from textual descriptions, which offers transformative potential across various fields. However, existing methods often struggle to align generated content with human preferences, limiting their applicability and flexibility. To address these limitations, in this paper, we propose DreamDPO, an optimization-based framework that integrates human preferences into the 3D generation process, through direct preference optimization. Practically, DreamDPO first constructs pairwise examples, then compare their alignment with human preferences using reward or large multimodal models, and lastly optimizes the 3D representation with a preference-driven loss function. By leveraging pairwise comparison to reflect preferences, DreamDPO reduces reliance on precise pointwise quality evaluations while enabling fine-grained controllability through preference-guided optimization. Experiments demonstrate that DreamDPO achieves competitive results, and provides higher-quality and more controllable 3D content compared to existing methods. The code and models will be open-sourced.
2502.04371
PerPO: Perceptual Preference Optimization via Discriminative Rewarding
cs.AI cs.CL cs.LG
This paper presents Perceptual Preference Optimization (PerPO), a perception alignment method aimed at addressing the visual discrimination challenges in generative pre-trained multimodal large language models (MLLMs). To align MLLMs with human visual perception process, PerPO employs discriminative rewarding to gather diverse negative samples, followed by listwise preference optimization to rank them.By utilizing the reward as a quantitative margin for ranking, our method effectively bridges generative preference optimization and discriminative empirical risk minimization. PerPO significantly enhances MLLMs' visual discrimination capabilities while maintaining their generative strengths, mitigates image-unconditional reward hacking, and ensures consistent performance across visual tasks. This work marks a crucial step towards more perceptually aligned and versatile MLLMs. We also hope that PerPO will encourage the community to rethink MLLM alignment strategies.
2502.04372
Mining Unstructured Medical Texts With Conformal Active Learning
cs.CL cs.LG stat.ML
The extraction of relevant data from Electronic Health Records (EHRs) is crucial to identifying symptoms and automating epidemiological surveillance processes. By harnessing the vast amount of unstructured text in EHRs, we can detect patterns that indicate the onset of disease outbreaks, enabling faster, more targeted public health responses. Our proposed framework provides a flexible and efficient solution for mining data from unstructured texts, significantly reducing the need for extensive manual labeling by specialists. Experiments show that our framework achieving strong performance with as few as 200 manually labeled texts, even for complex classification problems. Additionally, our approach can function with simple lightweight models, achieving competitive and occasionally even better results compared to more resource-intensive deep learning models. This capability not only accelerates processing times but also preserves patient privacy, as the data can be processed on weaker on-site hardware rather than being transferred to external systems. Our methodology, therefore, offers a practical, scalable, and privacy-conscious approach to real-time epidemiological monitoring, equipping health institutions to respond rapidly and effectively to emerging health threats.
2502.04375
An Analysis for Reasoning Bias of Language Models with Small Initialization
cs.CL cs.LG
Transformer-based Large Language Models (LLMs) have revolutionized Natural Language Processing by demonstrating exceptional performance across diverse tasks. This study investigates the impact of the parameter initialization scale on the training behavior and task preferences of LLMs. We discover that smaller initialization scales encourage models to favor reasoning tasks, whereas larger initialization scales lead to a preference for memorization tasks. We validate this reasoning bias via real datasets and meticulously designed anchor functions. Further analysis of initial training dynamics suggests that specific model components, particularly the embedding space and self-attention mechanisms, play pivotal roles in shaping these learning biases. We provide a theoretical framework from the perspective of model training dynamics to explain these phenomena. Additionally, experiments on real-world language tasks corroborate our theoretical insights. This work enhances our understanding of how initialization strategies influence LLM performance on reasoning tasks and offers valuable guidelines for training models.
2502.04376
MEETING DELEGATE: Benchmarking LLMs on Attending Meetings on Our Behalf
cs.CL cs.AI
In contemporary workplaces, meetings are essential for exchanging ideas and ensuring team alignment but often face challenges such as time consumption, scheduling conflicts, and inefficient participation. Recent advancements in Large Language Models (LLMs) have demonstrated their strong capabilities in natural language generation and reasoning, prompting the question: can LLMs effectively delegate participants in meetings? To explore this, we develop a prototype LLM-powered meeting delegate system and create a comprehensive benchmark using real meeting transcripts. Our evaluation reveals that GPT-4/4o maintain balanced performance between active and cautious engagement strategies. In contrast, Gemini 1.5 Pro tends to be more cautious, while Gemini 1.5 Flash and Llama3-8B/70B display more active tendencies. Overall, about 60\% of responses address at least one key point from the ground-truth. However, improvements are needed to reduce irrelevant or repetitive content and enhance tolerance for transcription errors commonly found in real-world settings. Additionally, we implement the system in practical settings and collect real-world feedback from demos. Our findings underscore the potential and challenges of utilizing LLMs as meeting delegates, offering valuable insights into their practical application for alleviating the burden of meetings.
2502.04377
MapFusion: A Novel BEV Feature Fusion Network for Multi-modal Map Construction
cs.CV cs.AI
Map construction task plays a vital role in providing precise and comprehensive static environmental information essential for autonomous driving systems. Primary sensors include cameras and LiDAR, with configurations varying between camera-only, LiDAR-only, or camera-LiDAR fusion, based on cost-performance considerations. While fusion-based methods typically perform best, existing approaches often neglect modality interaction and rely on simple fusion strategies, which suffer from the problems of misalignment and information loss. To address these issues, we propose MapFusion, a novel multi-modal Bird's-Eye View (BEV) feature fusion method for map construction. Specifically, to solve the semantic misalignment problem between camera and LiDAR BEV features, we introduce the Cross-modal Interaction Transform (CIT) module, enabling interaction between two BEV feature spaces and enhancing feature representation through a self-attention mechanism. Additionally, we propose an effective Dual Dynamic Fusion (DDF) module to adaptively select valuable information from different modalities, which can take full advantage of the inherent information between different modalities. Moreover, MapFusion is designed to be simple and plug-and-play, easily integrated into existing pipelines. We evaluate MapFusion on two map construction tasks, including High-definition (HD) map and BEV map segmentation, to show its versatility and effectiveness. Compared with the state-of-the-art methods, MapFusion achieves 3.6% and 6.2% absolute improvements on the HD map construction and BEV map segmentation tasks on the nuScenes dataset, respectively, demonstrating the superiority of our approach.
2502.04378
DILLEMA: Diffusion and Large Language Models for Multi-Modal Augmentation
cs.CV cs.GR cs.LG cs.SE
Ensuring the robustness of deep learning models requires comprehensive and diverse testing. Existing approaches, often based on simple data augmentation techniques or generative adversarial networks, are limited in producing realistic and varied test cases. To address these limitations, we present a novel framework for testing vision neural networks that leverages Large Language Models and control-conditioned Diffusion Models to generate synthetic, high-fidelity test cases. Our approach begins by translating images into detailed textual descriptions using a captioning model, allowing the language model to identify modifiable aspects of the image and generate counterfactual descriptions. These descriptions are then used to produce new test images through a text-to-image diffusion process that preserves spatial consistency and maintains the critical elements of the scene. We demonstrate the effectiveness of our method using two datasets: ImageNet1K for image classification and SHIFT for semantic segmentation in autonomous driving. The results show that our approach can generate significant test cases that reveal weaknesses and improve the robustness of the model through targeted retraining. We conducted a human assessment using Mechanical Turk to validate the generated images. The responses from the participants confirmed, with high agreement among the voters, that our approach produces valid and realistic images.
2502.04379
Can Large Language Models Capture Video Game Engagement?
cs.CV cs.AI cs.CL cs.HC
Can out-of-the-box pretrained Large Language Models (LLMs) detect human affect successfully when observing a video? To address this question, for the first time, we evaluate comprehensively the capacity of popular LLMs to annotate and successfully predict continuous affect annotations of videos when prompted by a sequence of text and video frames in a multimodal fashion. Particularly in this paper, we test LLMs' ability to correctly label changes of in-game engagement in 80 minutes of annotated videogame footage from 20 first-person shooter games of the GameVibe corpus. We run over 2,400 experiments to investigate the impact of LLM architecture, model size, input modality, prompting strategy, and ground truth processing method on engagement prediction. Our findings suggest that while LLMs rightfully claim human-like performance across multiple domains, they generally fall behind capturing continuous experience annotations provided by humans. We examine some of the underlying causes for the relatively poor overall performance, highlight the cases where LLMs exceed expectations, and draw a roadmap for the further exploration of automated emotion labelling via LLMs.
2502.04380
Diversity as a Reward: Fine-Tuning LLMs on a Mixture of Domain-Undetermined Data
cs.CL cs.AI cs.LG
Fine-tuning large language models (LLMs) using diverse datasets is crucial for enhancing their overall performance across various domains. In practical scenarios, existing methods based on modeling the mixture proportions of data composition often struggle with data whose domain labels are missing, imprecise or non-normalized, while methods based on data selection usually encounter difficulties in balancing multi-domain performance. To address these challenges, in this paper, we study the role of data diversity in enhancing the overall abilities of LLMs by empirically constructing contrastive data pools and theoretically deriving explanations for both inter- and intra-diversity. Building upon the insights gained, we propose a new method that gives the LLM a dual identity: an output model to cognitively probe and select data based on diversity reward, as well as an input model to be tuned with the selected data. Extensive experiments show that the proposed method notably boosts performance across domain-undetermined data and a series of foundational downstream tasks when applied to various advanced LLMs. We release our code and hope this study can shed light on the understanding of data diversity and advance feedback-driven data-model co-development for LLMs.
2502.04381
Limitations of Large Language Models in Clinical Problem-Solving Arising from Inflexible Reasoning
cs.CL cs.AI
Large Language Models (LLMs) have attained human-level accuracy on medical question-answer (QA) benchmarks. However, their limitations in navigating open-ended clinical scenarios have recently been shown, raising concerns about the robustness and generalizability of LLM reasoning across diverse, real-world medical tasks. To probe potential LLM failure modes in clinical problem-solving, we present the medical abstraction and reasoning corpus (M-ARC). M-ARC assesses clinical reasoning through scenarios designed to exploit the Einstellung effect -- the fixation of thought arising from prior experience, targeting LLM inductive biases toward inflexible pattern matching from their training data rather than engaging in flexible reasoning. We find that LLMs, including current state-of-the-art o1 and Gemini models, perform poorly compared to physicians on M-ARC, often demonstrating lack of commonsense medical reasoning and a propensity to hallucinate. In addition, uncertainty estimation analyses indicate that LLMs exhibit overconfidence in their answers, despite their limited accuracy. The failure modes revealed by M-ARC in LLM medical reasoning underscore the need to exercise caution when deploying these models in clinical settings.
2502.04382
Sparse Autoencoders for Hypothesis Generation
cs.CL cs.AI cs.CY
We describe HypotheSAEs, a general method to hypothesize interpretable relationships between text data (e.g., headlines) and a target variable (e.g., clicks). HypotheSAEs has three steps: (1) train a sparse autoencoder on text embeddings to produce interpretable features describing the data distribution, (2) select features that predict the target variable, and (3) generate a natural language interpretation of each feature (e.g., "mentions being surprised or shocked") using an LLM. Each interpretation serves as a hypothesis about what predicts the target variable. Compared to baselines, our method better identifies reference hypotheses on synthetic datasets (at least +0.06 in F1) and produces more predictive hypotheses on real datasets (~twice as many significant findings), despite requiring 1-2 orders of magnitude less compute than recent LLM-based methods. HypotheSAEs also produces novel discoveries on two well-studied tasks: explaining partisan differences in Congressional speeches and identifying drivers of engagement with online headlines.
2502.04384
Enhancing Reasoning to Adapt Large Language Models for Domain-Specific Applications
cs.CL cs.AI cs.LG cs.SY eess.SY
This paper presents SOLOMON, a novel Neuro-inspired Large Language Model (LLM) Reasoning Network architecture that enhances the adaptability of foundation models for domain-specific applications. Through a case study in semiconductor layout design, we demonstrate how SOLOMON enables swift adaptation of general-purpose LLMs to specialized tasks by leveraging Prompt Engineering and In-Context Learning techniques. Our experiments reveal the challenges LLMs face in spatial reasoning and applying domain knowledge to practical problems. Results show that SOLOMON instances significantly outperform their baseline LLM counterparts and achieve performance comparable to state-of-the-art reasoning model, o1-preview. We discuss future research directions for developing more adaptive AI systems that can continually learn, adapt, and evolve in response to new information and changing requirements.
2502.04385
TexLiDAR: Automated Text Understanding for Panoramic LiDAR Data
cs.CV cs.AI
Efforts to connect LiDAR data with text, such as LidarCLIP, have primarily focused on embedding 3D point clouds into CLIP text-image space. However, these approaches rely on 3D point clouds, which present challenges in encoding efficiency and neural network processing. With the advent of advanced LiDAR sensors like Ouster OS1, which, in addition to 3D point clouds, produce fixed resolution depth, signal, and ambient panoramic 2D images, new opportunities emerge for LiDAR based tasks. In this work, we propose an alternative approach to connect LiDAR data with text by leveraging 2D imagery generated by the OS1 sensor instead of 3D point clouds. Using the Florence 2 large model in a zero-shot setting, we perform image captioning and object detection. Our experiments demonstrate that Florence 2 generates more informative captions and achieves superior performance in object detection tasks compared to existing methods like CLIP. By combining advanced LiDAR sensor data with a large pre-trained model, our approach provides a robust and accurate solution for challenging detection scenarios, including real-time applications requiring high accuracy and robustness.
2502.04386
Towards Fair Medical AI: Adversarial Debiasing of 3D CT Foundation Embeddings
cs.CV cs.AI cs.LG
Self-supervised learning has revolutionized medical imaging by enabling efficient and generalizable feature extraction from large-scale unlabeled datasets. Recently, self-supervised foundation models have been extended to three-dimensional (3D) computed tomography (CT) data, generating compact, information-rich embeddings with 1408 features that achieve state-of-the-art performance on downstream tasks such as intracranial hemorrhage detection and lung cancer risk forecasting. However, these embeddings have been shown to encode demographic information, such as age, sex, and race, which poses a significant risk to the fairness of clinical applications. In this work, we propose a Variation Autoencoder (VAE) based adversarial debiasing framework to transform these embeddings into a new latent space where demographic information is no longer encoded, while maintaining the performance of critical downstream tasks. We validated our approach on the NLST lung cancer screening dataset, demonstrating that the debiased embeddings effectively eliminate multiple encoded demographic information and improve fairness without compromising predictive accuracy for lung cancer risk at 1-year and 2-year intervals. Additionally, our approach ensures the embeddings are robust against adversarial bias attacks. These results highlight the potential of adversarial debiasing techniques to ensure fairness and equity in clinical applications of self-supervised 3D CT embeddings, paving the way for their broader adoption in unbiased medical decision-making.
2502.04387
FedP$^2$EFT: Federated Learning to Personalize Parameter Efficient Fine-Tuning for Multilingual LLMs
cs.CL cs.AI
Federated learning (FL) has enabled the training of multilingual large language models (LLMs) on diverse and decentralized multilingual data, especially on low-resource languages. To improve client-specific performance, personalization via the use of parameter-efficient fine-tuning (PEFT) modules such as LoRA is common. This involves a personalization strategy (PS), such as the design of the PEFT adapter structures (e.g., in which layers to add LoRAs and what ranks) and choice of hyperparameters (e.g., learning rates) for fine-tuning. Instead of manual PS configuration, we propose FedP$^2$EFT, a federated learning-to-personalize method for multilingual LLMs in cross-device FL settings. Unlike most existing PEFT structure selection methods, which are prone to overfitting low-data regimes, FedP$^2$EFT collaboratively learns the optimal personalized PEFT structure for each client via Bayesian sparse rank selection. Evaluations on both simulated and real-world multilingual FL benchmarks demonstrate that FedP$^2$EFT largely outperforms existing personalized fine-tuning methods, while complementing a range of existing FL methods.
2502.04388
Position: Emergent Machina Sapiens Urge Rethinking Multi-Agent Paradigms
cs.MA cs.AI
Artificially intelligent (AI) agents that are capable of autonomous learning and independent decision-making hold great promise for addressing complex challenges across domains like transportation, energy systems, and manufacturing. However, the surge in AI systems' design and deployment driven by various stakeholders with distinct and unaligned objectives introduces a crucial challenge: how can uncoordinated AI systems coexist and evolve harmoniously in shared environments without creating chaos? To address this, we advocate for a fundamental rethinking of existing multi-agent frameworks, such as multi-agent systems and game theory, which are largely limited to predefined rules and static objective structures. We posit that AI agents should be empowered to dynamically adjust their objectives, make compromises, form coalitions, and safely compete or cooperate through evolving relationships and social feedback. Through this paper, we call for a shift toward the emergent, self-organizing, and context-aware nature of these systems.
2502.04389
Overcoming Vision Language Model Challenges in Diagram Understanding: A Proof-of-Concept with XML-Driven Large Language Models Solutions
cs.SE cs.AI
Diagrams play a crucial role in visually conveying complex relationships and processes within business documentation. Despite recent advances in Vision-Language Models (VLMs) for various image understanding tasks, accurately identifying and extracting the structures and relationships depicted in diagrams continues to pose significant challenges. This study addresses these challenges by proposing a text-driven approach that bypasses reliance on VLMs' visual recognition capabilities. Instead, it utilizes the editable source files--such as xlsx, pptx or docx--where diagram elements (e.g., shapes, lines, annotations) are preserved as textual metadata. In our proof-of-concept, we extracted diagram information from xlsx-based system design documents and transformed the extracted shape data into textual input for Large Language Models (LLMs). This approach allowed the LLM to analyze relationships and generate responses to business-oriented questions without the bottleneck of image-based processing. Experimental comparisons with a VLM-based method demonstrated that the proposed text-driven framework yielded more accurate answers for questions requiring detailed comprehension of diagram structures.The results obtained in this study are not limited to the tested .xlsx files but can also be extended to diagrams in other documents with source files, such as Office pptx and docx formats. These findings highlight the feasibility of circumventing VLM constraints through direct textual extraction from original source files. By enabling robust diagram understanding through LLMs, our method offers a promising path toward enhanced workflow efficiency and information analysis in real-world business scenarios.
2502.04390
In Praise of Stubbornness: The Case for Cognitive-Dissonance-Aware Knowledge Updates in LLMs
cs.CL cs.AI cs.LG q-bio.NC
Despite remarkable capabilities, large language models (LLMs) struggle to continually update their knowledge without catastrophic forgetting. In contrast, humans effortlessly integrate new information, detect conflicts with existing beliefs, and selectively update their mental models. This paper introduces a cognitive-inspired investigation paradigm to study continual knowledge updating in LLMs. We implement two key components inspired by human cognition: (1) Dissonance and Familiarity Awareness, analyzing model behavior to classify information as novel, familiar, or dissonant; and (2) Targeted Network Updates, which track neural activity to identify frequently used (stubborn) and rarely used (plastic) neurons. Through carefully designed experiments in controlled settings, we uncover a number of empirical findings demonstrating the potential of this approach. First, dissonance detection is feasible using simple activation and gradient features, suggesting potential for cognitive-inspired training. Second, we find that non-dissonant updates largely preserve prior knowledge regardless of targeting strategy, revealing inherent robustness in LLM knowledge integration. Most critically, we discover that dissonant updates prove catastrophically destructive to the model's knowledge base, indiscriminately affecting even information unrelated to the current updates. This suggests fundamental limitations in how neural networks handle contradictions and motivates the need for new approaches to knowledge updating that better mirror human cognitive mechanisms.
2502.04391
Towards Fair and Robust Face Parsing for Generative AI: A Multi-Objective Approach
cs.CV cs.AI
Face parsing is a fundamental task in computer vision, enabling applications such as identity verification, facial editing, and controllable image synthesis. However, existing face parsing models often lack fairness and robustness, leading to biased segmentation across demographic groups and errors under occlusions, noise, and domain shifts. These limitations affect downstream face synthesis, where segmentation biases can degrade generative model outputs. We propose a multi-objective learning framework that optimizes accuracy, fairness, and robustness in face parsing. Our approach introduces a homotopy-based loss function that dynamically adjusts the importance of these objectives during training. To evaluate its impact, we compare multi-objective and single-objective U-Net models in a GAN-based face synthesis pipeline (Pix2PixHD). Our results show that fairness-aware and robust segmentation improves photorealism and consistency in face generation. Additionally, we conduct preliminary experiments using ControlNet, a structured conditioning model for diffusion-based synthesis, to explore how segmentation quality influences guided image generation. Our findings demonstrate that multi-objective face parsing improves demographic consistency and robustness, leading to higher-quality GAN-based synthesis.
2502.04392
Division-of-Thoughts: Harnessing Hybrid Language Model Synergy for Efficient On-Device Agents
cs.CL cs.AI
The rapid expansion of web content has made on-device AI assistants indispensable for helping users manage the increasing complexity of online tasks. The emergent reasoning ability in large language models offer a promising path for next-generation on-device AI agents. However, deploying full-scale Large Language Models (LLMs) on resource-limited local devices is challenging. In this paper, we propose Division-of-Thoughts (DoT), a collaborative reasoning framework leveraging the synergy between locally deployed Smaller-scale Language Models (SLMs) and cloud-based LLMs. DoT leverages a Task Decomposer to elicit the inherent planning abilities in language models to decompose user queries into smaller sub-tasks, which allows hybrid language models to fully exploit their respective strengths. Besides, DoT employs a Task Scheduler to analyze the pair-wise dependency of sub-tasks and create a dependency graph, facilitating parallel reasoning of sub-tasks and the identification of key steps. To allocate the appropriate model based on the difficulty of sub-tasks, DoT leverages a Plug-and-Play Adapter, which is an additional task head attached to the SLM that does not alter the SLM's parameters. To boost adapter's task allocation capability, we propose a self-reinforced training method that relies solely on task execution feedback. Extensive experiments on various benchmarks demonstrate that our DoT significantly reduces LLM costs while maintaining competitive reasoning accuracy. Specifically, DoT reduces the average reasoning time and API costs by 66.12% and 83.57%, while achieving comparable reasoning accuracy with the best baseline methods.
2502.04393
UniCP: A Unified Caching and Pruning Framework for Efficient Video Generation
cs.CV
Diffusion Transformers (DiT) excel in video generation but encounter significant computational challenges due to the quadratic complexity of attention. Notably, attention differences between adjacent diffusion steps follow a U-shaped pattern. Current methods leverage this property by caching attention blocks, however, they still struggle with sudden error spikes and large discrepancies. To address these issues, we propose UniCP a unified caching and pruning framework for efficient video generation. UniCP optimizes both temporal and spatial dimensions through. Error Aware Dynamic Cache Window (EDCW): Dynamically adjusts cache window sizes for different blocks at various timesteps, adapting to abrupt error changes. PCA based Slicing (PCAS) and Dynamic Weight Shift (DWS): PCAS prunes redundant attention components, and DWS integrates caching and pruning by enabling dynamic switching between pruned and cached outputs. By adjusting cache windows and pruning redundant components, UniCP enhances computational efficiency and maintains video detail fidelity. Experimental results show that UniCP outperforms existing methods in both performance and efficiency.
2502.04394
DECT: Harnessing LLM-assisted Fine-Grained Linguistic Knowledge and Label-Switched and Label-Preserved Data Generation for Diagnosis of Alzheimer's Disease
cs.CL cs.AI
Alzheimer's Disease (AD) is an irreversible neurodegenerative disease affecting 50 million people worldwide. Low-cost, accurate identification of key markers of AD is crucial for timely diagnosis and intervention. Language impairment is one of the earliest signs of cognitive decline, which can be used to discriminate AD patients from normal control individuals. Patient-interviewer dialogues may be used to detect such impairments, but they are often mixed with ambiguous, noisy, and irrelevant information, making the AD detection task difficult. Moreover, the limited availability of AD speech samples and variability in their speech styles pose significant challenges in developing robust speech-based AD detection models. To address these challenges, we propose DECT, a novel speech-based domain-specific approach leveraging large language models (LLMs) for fine-grained linguistic analysis and label-switched label-preserved data generation. Our study presents four novelties: We harness the summarizing capabilities of LLMs to identify and distill key Cognitive-Linguistic information from noisy speech transcripts, effectively filtering irrelevant information. We leverage the inherent linguistic knowledge of LLMs to extract linguistic markers from unstructured and heterogeneous audio transcripts. We exploit the compositional ability of LLMs to generate AD speech transcripts consisting of diverse linguistic patterns to overcome the speech data scarcity challenge and enhance the robustness of AD detection models. We use the augmented AD textual speech transcript dataset and a more fine-grained representation of AD textual speech transcript data to fine-tune the AD detection model. The results have shown that DECT demonstrates superior model performance with an 11% improvement in AD detection accuracy on the datasets from DementiaBank compared to the baselines.
2502.04395
Time-VLM: Exploring Multimodal Vision-Language Models for Augmented Time Series Forecasting
cs.CV cs.LG
Recent advancements in time series forecasting have explored augmenting models with text or vision modalities to improve accuracy. While text provides contextual understanding, it often lacks fine-grained temporal details. Conversely, vision captures intricate temporal patterns but lacks semantic context, limiting the complementary potential of these modalities. To address this, we propose Time-VLM, a novel multimodal framework that leverages pre-trained Vision-Language Models (VLMs) to bridge temporal, visual, and textual modalities for enhanced forecasting. Our framework comprises three key components: (1) a Retrieval-Augmented Learner, which extracts enriched temporal features through memory bank interactions; (2) a Vision-Augmented Learner, which encodes time series as informative images; and (3) a Text-Augmented Learner, which generates contextual textual descriptions. These components collaborate with frozen pre-trained VLMs to produce multimodal embeddings, which are then fused with temporal features for final prediction. Extensive experiments across diverse datasets demonstrate that Time-VLM achieves superior performance, particularly in few-shot and zero-shot scenarios, thereby establishing a new direction for multimodal time series forecasting.
2502.04397
Multimodal Medical Code Tokenizer
cs.CL cs.AI cs.LG
Foundation models trained on patient electronic health records (EHRs) require tokenizing medical data into sequences of discrete vocabulary items. Existing tokenizers treat medical codes from EHRs as isolated textual tokens. However, each medical code is defined by its textual description, its position in ontological hierarchies, and its relationships to other codes, such as disease co-occurrences and drug-treatment associations. Medical vocabularies contain more than 600,000 codes with critical information for clinical reasoning. We introduce MedTok, a multimodal medical code tokenizer that uses the text descriptions and relational context of codes. MedTok processes text using a language model encoder and encodes the relational structure with a graph encoder. It then quantizes both modalities into a unified token space, preserving modality-specific and cross-modality information. We integrate MedTok into five EHR models and evaluate it on operational and clinical tasks across in-patient and out-patient datasets, including outcome prediction, diagnosis classification, drug recommendation, and risk stratification. Swapping standard EHR tokenizers with MedTok improves AUPRC across all EHR models, by 4.10% on MIMIC-III, 4.78% on MIMIC-IV, and 11.30% on EHRShot, with the largest gains in drug recommendation. Beyond EHR modeling, we demonstrate using MedTok tokenizer with medical QA systems. Our results demonstrate the potential of MedTok as a unified tokenizer for medical codes, improving tokenization for medical foundation models.
2502.04398
XMTC: Explainable Early Classification of Multivariate Time Series in Reach-to-Grasp Hand Kinematics
cs.LG cs.GR cs.HC
Hand kinematics can be measured in Human-Computer Interaction (HCI) with the intention to predict the user's intention in a reach-to-grasp action. Using multiple hand sensors, multivariate time series data are being captured. Given a number of possible actions on a number of objects, the goal is to classify the multivariate time series data, where the class shall be predicted as early as possible. Many machine-learning methods have been developed for such classification tasks, where different approaches produce favorable solutions on different data sets. We, therefore, employ an ensemble approach that includes and weights different approaches. To provide a trustworthy classification production, we present the XMTC tool that incorporates coordinated multiple-view visualizations to analyze the predictions. Temporal accuracy plots, confusion matrix heatmaps, temporal confidence heatmaps, and partial dependence plots allow for the identification of the best trade-off between early prediction and prediction quality, the detection and analysis of challenging classification conditions, and the investigation of the prediction evolution in an overview and detail manner. We employ XMTC to real-world HCI data in multiple scenarios and show that good classification predictions can be achieved early on with our classifier as well as which conditions are easy to distinguish, which multivariate time series measurements impose challenges, and which features have most impact.
2502.04399
Online Location Planning for AI-Defined Vehicles: Optimizing Joint Tasks of Order Serving and Spatio-Temporal Heterogeneous Model Fine-Tuning
cs.LG cs.AI cs.SY eess.SY
Advances in artificial intelligence (AI) including foundation models (FMs), are increasingly transforming human society, with smart city driving the evolution of urban living.Meanwhile, vehicle crowdsensing (VCS) has emerged as a key enabler, leveraging vehicles' mobility and sensor-equipped capabilities. In particular, ride-hailing vehicles can effectively facilitate flexible data collection and contribute towards urban intelligence, despite resource limitations. Therefore, this work explores a promising scenario, where edge-assisted vehicles perform joint tasks of order serving and the emerging foundation model fine-tuning using various urban data. However, integrating the VCS AI task with the conventional order serving task is challenging, due to their inconsistent spatio-temporal characteristics: (i) The distributions of ride orders and data point-of-interests (PoIs) may not coincide in geography, both following a priori unknown patterns; (ii) they have distinct forms of temporal effects, i.e., prolonged waiting makes orders become instantly invalid while data with increased staleness gradually reduces its utility for model fine-tuning.To overcome these obstacles, we propose an online framework based on multi-agent reinforcement learning (MARL) with careful augmentation. A new quality-of-service (QoS) metric is designed to characterize and balance the utility of the two joint tasks, under the effects of varying data volumes and staleness. We also integrate graph neural networks (GNNs) with MARL to enhance state representations, capturing graph-structured, time-varying dependencies among vehicles and across locations. Extensive experiments on our testbed simulator, utilizing various real-world foundation model fine-tuning tasks and the New York City Taxi ride order dataset, demonstrate the advantage of our proposed method.
2502.04400
Adaptive Prototype Knowledge Transfer for Federated Learning with Mixed Modalities and Heterogeneous Tasks
cs.LG cs.AI cs.CR cs.MM
Multimodal Federated Learning (MFL) enables multiple clients to collaboratively train models on multimodal data while ensuring clients' privacy. However, modality and task heterogeneity hinder clients from learning a unified representation, weakening local model generalization, especially in MFL with mixed modalities where only some clients have multimodal data. In this work, we propose an Adaptive prototype-based Multimodal Federated Learning (AproMFL) framework for mixed modalities and heterogeneous tasks to address the aforementioned issues. Our AproMFL transfers knowledge through adaptively-constructed prototypes without a prior public dataset. Clients adaptively select prototype construction methods in line with tasks; server converts client prototypes into unified multimodal prototypes and aggregates them to form global prototypes, avoid clients keeping unified labels. We divide the model into various modules and only aggregate mapping modules to reduce communication and computation overhead. To address aggregation issues in heterogeneity, we develop a client relationship graph-based scheme to dynamically adjust aggregation weights. Extensive experiments on representative datasets evidence effectiveness of AproMFL.
2502.04402
Beyond Interpolation: Extrapolative Reasoning with Reinforcement Learning and Graph Neural Networks
cs.LG cs.AI
Despite incredible progress, many neural architectures fail to properly generalize beyond their training distribution. As such, learning to reason in a correct and generalizable way is one of the current fundamental challenges in machine learning. In this respect, logic puzzles provide a great testbed, as we can fully understand and control the learning environment. Thus, they allow to evaluate performance on previously unseen, larger and more difficult puzzles that follow the same underlying rules. Since traditional approaches often struggle to represent such scalable logical structures, we propose to model these puzzles using a graph-based approach. Then, we investigate the key factors enabling the proposed models to learn generalizable solutions in a reinforcement learning setting. Our study focuses on the impact of the inductive bias of the architecture, different reward systems and the role of recurrent modeling in enabling sequential reasoning. Through extensive experiments, we demonstrate how these elements contribute to successful extrapolation on increasingly complex puzzles.These insights and frameworks offer a systematic way to design learning-based systems capable of generalizable reasoning beyond interpolation.
2502.04403
Agency Is Frame-Dependent
cs.AI
Agency is a system's capacity to steer outcomes toward a goal, and is a central topic of study across biology, philosophy, cognitive science, and artificial intelligence. Determining if a system exhibits agency is a notoriously difficult question: Dennett (1989), for instance, highlights the puzzle of determining which principles can decide whether a rock, a thermostat, or a robot each possess agency. We here address this puzzle from the viewpoint of reinforcement learning by arguing that agency is fundamentally frame-dependent: Any measurement of a system's agency must be made relative to a reference frame. We support this claim by presenting a philosophical argument that each of the essential properties of agency proposed by Barandiaran et al. (2009) and Moreno (2018) are themselves frame-dependent. We conclude that any basic science of agency requires frame-dependence, and discuss the implications of this claim for reinforcement learning.
2502.04404
Step Back to Leap Forward: Self-Backtracking for Boosting Reasoning of Language Models
cs.CL cs.AI
The integration of slow-thinking mechanisms into large language models (LLMs) offers a promising way toward achieving Level 2 AGI Reasoners, as exemplified by systems like OpenAI's o1. However, several significant challenges remain, including inefficient overthinking and an overreliance on auxiliary reward models. We point out that these limitations stem from LLMs' inability to internalize the search process, a key component of effective reasoning. A critical step toward addressing this issue is enabling LLMs to autonomously determine when and where to backtrack, a fundamental operation in traditional search algorithms. To this end, we propose a self-backtracking mechanism that equips LLMs with the ability to backtrack during both training and inference. This mechanism not only enhances reasoning ability but also efficiency by transforming slow-thinking processes into fast-thinking through self-improvement. Empirical evaluations demonstrate that our proposal significantly enhances the reasoning capabilities of LLMs, achieving a performance gain of over 40 percent compared to the optimal-path supervised fine-tuning method. We believe this study introduces a novel and promising pathway for developing more advanced and robust Reasoners.
2502.04405
FAS: Fast ANN-SNN Conversion for Spiking Large Language Models
cs.LG cs.AI cs.CL
Spiking Large Language Models have been shown as a good alternative to LLMs in various scenarios. Existing methods for creating Spiking LLMs, i.e., direct training and ANN-SNN conversion, often suffer from performance degradation and relatively high computational costs. To address these issues, we propose a novel Fast ANN-SNN conversion strategy (FAS) that transforms LLMs into spiking LLMs in two stages. The first stage employs a full-parameter fine-tuning of pre-trained models, so it does not need any direct training from scratch. The second stage introduces a coarse-to-fine calibration method to reduce conversion errors and improve accuracy. Our experiments on both language and vision-language tasks across four different scales of LLMs demonstrate that FAS can achieve state-of-the-art performance yet with significantly reduced inference latency and computational costs. For example, FAS only takes 8 timesteps to achieve an accuracy of 3% higher than that of the OPT-7B model, while reducing energy consumption by 96.63%.
2502.04406
Calibrated Physics-Informed Uncertainty Quantification
cs.LG cs.AI physics.comp-ph
Neural PDEs offer efficient alternatives to computationally expensive numerical PDE solvers for simulating complex physical systems. However, their lack of robust uncertainty quantification (UQ) limits deployment in critical applications. We introduce a model-agnostic, physics-informed conformal prediction (CP) framework that provides guaranteed uncertainty estimates without requiring labelled data. By utilising a physics-based approach, we are able to quantify and calibrate the model's inconsistencies with the PDE rather than the uncertainty arising from the data. Our approach uses convolutional layers as finite-difference stencils and leverages physics residual errors as nonconformity scores, enabling data-free UQ with marginal and joint coverage guarantees across prediction domains for a range of complex PDEs. We further validate the efficacy of our method on neural PDE models for plasma modelling and shot design in fusion reactors.
2502.04407
Illuminating Spaces: Deep Reinforcement Learning and Laser-Wall Partitioning for Architectural Layout Generation
cs.LG cs.AI
Space layout design (SLD), occurring in the early stages of the design process, nonetheless influences both the functionality and aesthetics of the ultimate architectural outcome. The complexity of SLD necessitates innovative approaches to efficiently explore vast solution spaces. While image-based generative AI has emerged as a potential solution, they often rely on pixel-based space composition methods that lack intuitive representation of architectural processes. This paper leverages deep Reinforcement Learning (RL), as it offers a procedural approach that intuitively mimics the process of human designers. Effectively using RL for SLD requires an explorative space composing method to generate desirable design solutions. We introduce "laser-wall", a novel space partitioning method that conceptualizes walls as emitters of imaginary light beams to partition spaces. This approach bridges vector-based and pixel-based partitioning methods, offering both flexibility and exploratory power in generating diverse layouts. We present two planning strategies: one-shot planning, which generates entire layouts in a single pass, and dynamic planning, which allows for adaptive refinement by continuously transforming laser-walls. Additionally, we introduce on-light and off-light wall transformations for smooth and fast layout refinement, as well as identity-less and identity-full walls for versatile room assignment. We developed SpaceLayoutGym, an open-source OpenAI Gym compatible simulator for generating and evaluating space layouts. The RL agent processes the input design scenarios and generates solutions following a reward function that balances geometrical and topological requirements. Our results demonstrate that the RL-based laser-wall approach can generate diverse and functional space layouts that satisfy both geometric constraints and topological requirements and is architecturally intuitive.
2502.04408
Transforming Multimodal Models into Action Models for Radiotherapy
cs.LG cs.AI
Radiotherapy is a crucial cancer treatment that demands precise planning to balance tumor eradication and preservation of healthy tissue. Traditional treatment planning (TP) is iterative, time-consuming, and reliant on human expertise, which can potentially introduce variability and inefficiency. We propose a novel framework to transform a large multimodal foundation model (MLM) into an action model for TP using a few-shot reinforcement learning (RL) approach. Our method leverages the MLM's extensive pre-existing knowledge of physics, radiation, and anatomy, enhancing it through a few-shot learning process. This allows the model to iteratively improve treatment plans using a Monte Carlo simulator. Our results demonstrate that this method outperforms conventional RL-based approaches in both quality and efficiency, achieving higher reward scores and more optimal dose distributions in simulations on prostate cancer data. This proof-of-concept suggests a promising direction for integrating advanced AI models into clinical workflows, potentially enhancing the speed, quality, and standardization of radiotherapy treatment planning.
2502.04409
Learning low-dimensional representations of ensemble forecast fields using autoencoder-based methods
cs.LG physics.ao-ph
Large-scale numerical simulations often produce high-dimensional gridded data that is challenging to process for downstream applications. A prime example is numerical weather prediction, where atmospheric processes are modeled using discrete gridded representations of the physical variables and dynamics. Uncertainties are assessed by running the simulations multiple times, yielding ensembles of simulated fields as a high-dimensional stochastic representation of the forecast distribution. The high-dimensionality and large volume of ensemble datasets poses major computing challenges for subsequent forecasting stages. Data-driven dimensionality reduction techniques could help to reduce the data volume before further processing by learning meaningful and compact representations. However, existing dimensionality reduction methods are typically designed for deterministic and single-valued inputs, and thus cannot handle ensemble data from multiple randomized simulations. In this study, we propose novel dimensionality reduction approaches specifically tailored to the format of ensemble forecast fields. We present two alternative frameworks, which yield low-dimensional representations of ensemble forecasts while respecting their probabilistic character. The first approach derives a distribution-based representation of an input ensemble by applying standard dimensionality reduction techniques in a member-by-member fashion and merging the member representations into a joint parametric distribution model. The second approach achieves a similar representation by encoding all members jointly using a tailored variational autoencoder. We evaluate and compare both approaches in a case study using 10 years of temperature and wind speed forecasts over Europe. The approaches preserve key spatial and statistical characteristics of the ensemble and enable probabilistic reconstructions of the forecast fields.
2502.04411
Mediator: Memory-efficient LLM Merging with Less Parameter Conflicts and Uncertainty Based Routing
cs.LG cs.AI cs.CL
Model merging aggregates Large Language Models (LLMs) finetuned on different tasks into a stronger one. However, parameter conflicts between models leads to performance degradation in averaging. While model routing addresses this issue by selecting individual models during inference, it imposes excessive storage and compute costs, and fails to leverage the common knowledge from different models. In this work, we observe that different layers exhibit varying levels of parameter conflicts. Building on this insight, we average layers with minimal parameter conflicts and use a novel task-level expert routing for layers with significant conflicts. To further reduce storage costs, inspired by task arithmetic sparsity, we decouple multiple fine-tuned experts into a dense expert and several sparse experts. Considering the out-of-distribution samples, we select and merge appropriate experts based on the task uncertainty of the input data. We conduct extensive experiments on both LLaMA and Qwen with varying parameter scales, and evaluate on real-world reasoning tasks. Results demonstrate that our method consistently achieves significant performance improvements while requiring less system cost compared to existing methods.
2502.04412
Decoder-Only LLMs are Better Controllers for Diffusion Models
cs.CV cs.AI cs.CL
Groundbreaking advancements in text-to-image generation have recently been achieved with the emergence of diffusion models. These models exhibit a remarkable ability to generate highly artistic and intricately detailed images based on textual prompts. However, obtaining desired generation outcomes often necessitates repetitive trials of manipulating text prompts just like casting spells on a magic mirror, and the reason behind that is the limited capability of semantic understanding inherent in current image generation models. Specifically, existing diffusion models encode the text prompt input with a pre-trained encoder structure, which is usually trained on a limited number of image-caption pairs. The state-of-the-art large language models (LLMs) based on the decoder-only structure have shown a powerful semantic understanding capability as their architectures are more suitable for training on very large-scale unlabeled data. In this work, we propose to enhance text-to-image diffusion models by borrowing the strength of semantic understanding from large language models, and devise a simple yet effective adapter to allow the diffusion models to be compatible with the decoder-only structure. Meanwhile, we also provide a supporting theoretical analysis with various architectures (e.g., encoder-only, encoder-decoder, and decoder-only), and conduct extensive empirical evaluations to verify its effectiveness. The experimental results show that the enhanced models with our adapter module are superior to the stat-of-the-art models in terms of text-to-image generation quality and reliability.
2502.04413
MedRAG: Enhancing Retrieval-augmented Generation with Knowledge Graph-Elicited Reasoning for Healthcare Copilot
cs.CL cs.AI cs.IR
Retrieval-augmented generation (RAG) is a well-suited technique for retrieving privacy-sensitive Electronic Health Records (EHR). It can serve as a key module of the healthcare copilot, helping reduce misdiagnosis for healthcare practitioners and patients. However, the diagnostic accuracy and specificity of existing heuristic-based RAG models used in the medical domain are inadequate, particularly for diseases with similar manifestations. This paper proposes MedRAG, a RAG model enhanced by knowledge graph (KG)-elicited reasoning for the medical domain that retrieves diagnosis and treatment recommendations based on manifestations. MedRAG systematically constructs a comprehensive four-tier hierarchical diagnostic KG encompassing critical diagnostic differences of various diseases. These differences are dynamically integrated with similar EHRs retrieved from an EHR database, and reasoned within a large language model. This process enables more accurate and specific decision support, while also proactively providing follow-up questions to enhance personalized medical decision-making. MedRAG is evaluated on both a public dataset DDXPlus and a private chronic pain diagnostic dataset (CPDD) collected from Tan Tock Seng Hospital, and its performance is compared against various existing RAG methods. Experimental results show that, leveraging the information integration and relational abilities of the KG, our MedRAG provides more specific diagnostic insights and outperforms state-of-the-art models in reducing misdiagnosis rates. Our code will be available at https://github.com/SNOWTEAM2023/MedRAG
2502.04415
TerraQ: Spatiotemporal Question-Answering on Satellite Image Archives
cs.CV cs.AI
TerraQ is a spatiotemporal question-answering engine for satellite image archives. It is a natural language processing system that is built to process requests for satellite images satisfying certain criteria. The requests can refer to image metadata and entities from a specialized knowledge base (e.g., the Emilia-Romagna region). With it, users can make requests like "Give me a hundred images of rivers near ports in France, with less than 20% snow coverage and more than 10% cloud coverage", thus making Earth Observation data more easily accessible, in-line with the current landscape of digital assistants.
2502.04416
CMoE: Fast Carving of Mixture-of-Experts for Efficient LLM Inference
cs.LG cs.AI
Large language models (LLMs) achieve impressive performance by scaling model parameters, but this comes with significant inference overhead. Feed-forward networks (FFNs), which dominate LLM parameters, exhibit high activation sparsity in hidden neurons. To exploit this, researchers have proposed using a mixture-of-experts (MoE) architecture, where only a subset of parameters is activated. However, existing approaches often require extensive training data and resources, limiting their practicality. We propose CMoE (Carved MoE), a novel framework to efficiently carve MoE models from dense models. CMoE achieves remarkable performance through efficient expert grouping and lightweight adaptation. First, neurons are grouped into shared and routed experts based on activation rates. Next, we construct a routing mechanism without training from scratch, incorporating a differentiable routing process and load balancing. Using modest data, CMoE produces a well-designed, usable MoE from a 7B dense model within five minutes. With lightweight fine-tuning, it achieves high-performance recovery in under an hour. We make our code publicly available at https://github.com/JarvisPei/CMoE.
2502.04417
NeuralMOVES: A lightweight and microscopic vehicle emission estimation model based on reverse engineering and surrogate learning
cs.LG cs.AI
The transportation sector significantly contributes to greenhouse gas emissions, necessitating accurate emission models to guide mitigation strategies. Despite its field validation and certification, the industry-standard Motor Vehicle Emission Simulator (MOVES) faces challenges related to complexity in usage, high computational demands, and its unsuitability for microscopic real-time applications. To address these limitations, we present NeuralMOVES, a comprehensive suite of high-performance, lightweight surrogate models for vehicle CO2 emissions. Developed based on reverse engineering and Neural Networks, NeuralMOVES achieves a remarkable 6.013% Mean Average Percentage Error relative to MOVES across extensive tests spanning over two million scenarios with diverse trajectories and the factors regarding environments and vehicles. NeuralMOVES is only 2.4 MB, largely condensing the original MOVES and the reverse engineered MOVES into a compact representation, while maintaining high accuracy. Therefore, NeuralMOVES significantly enhances accessibility while maintaining the accuracy of MOVES, simplifying CO2 evaluation for transportation analyses and enabling real-time, microscopic applications across diverse scenarios without reliance on complex software or extensive computational resources. Moreover, this paper provides, for the first time, a framework for reverse engineering industrial-grade software tailored specifically to transportation scenarios, going beyond MOVES. The surrogate models are available at https://github.com/edgar-rs/neuralMOVES.
2502.04418
Autotelic Reinforcement Learning: Exploring Intrinsic Motivations for Skill Acquisition in Open-Ended Environments
cs.LG cs.AI
This paper presents a comprehensive overview of autotelic Reinforcement Learning (RL), emphasizing the role of intrinsic motivations in the open-ended formation of skill repertoires. We delineate the distinctions between knowledge-based and competence-based intrinsic motivations, illustrating how these concepts inform the development of autonomous agents capable of generating and pursuing self-defined goals. The typology of Intrinsically Motivated Goal Exploration Processes (IMGEPs) is explored, with a focus on the implications for multi-goal RL and developmental robotics. The autotelic learning problem is framed within a reward-free Markov Decision Process (MDP), WHERE agents must autonomously represent, generate, and master their own goals. We address the unique challenges in evaluating such agents, proposing various metrics for measuring exploration, generalization, and robustness in complex environments. This work aims to advance the understanding of autotelic RL agents and their potential for enhancing skill acquisition in a diverse and dynamic setting.
2502.04419
Understanding and Mitigating the Bias Inheritance in LLM-based Data Augmentation on Downstream Tasks
cs.LG cs.AI cs.CL
Generating synthetic datasets via large language models (LLMs) themselves has emerged as a promising approach to improve LLM performance. However, LLMs inherently reflect biases present in their training data, leading to a critical challenge: when these models generate synthetic data for training, they may propagate and amplify their inherent biases that can significantly impact model fairness and robustness on downstream tasks--a phenomenon we term bias inheritance. This work presents the first systematic investigation in understanding, analyzing, and mitigating bias inheritance. We study this problem by fine-tuning LLMs with a combined dataset consisting of original and LLM-augmented data, where bias ratio represents the proportion of augmented data. Through systematic experiments across 10 classification and generation tasks, we analyze how 6 different types of biases manifest at varying bias ratios. Our results reveal that bias inheritance has nuanced effects on downstream tasks, influencing both classification tasks and generation tasks differently. Then, our analysis identifies three key misalignment factors: misalignment of values, group data, and data distributions. Based on these insights, we propose three mitigation strategies: token-based, mask-based, and loss-based approaches. Experiments demonstrate that these strategies also work differently on various tasks and bias, indicating the substantial challenges to fully mitigate bias inheritance. We hope this work can provide valuable insights to the research of LLM data augmentation.
2502.04420
KVTuner: Sensitivity-Aware Layer-wise Mixed Precision KV Cache Quantization for Efficient and Nearly Lossless LLM Inference
cs.LG cs.AI cs.CL
KV cache quantization can improve Large Language Models (LLMs) inference throughput and latency in long contexts and large batch-size scenarios while preserving LLMs effectiveness. However, current methods have three unsolved issues: overlooking layer-wise sensitivity to KV cache quantization, high overhead of online fine-grained decision-making, and low flexibility to different LLMs and constraints. Therefore, we thoroughly analyze the inherent correlation of layer-wise transformer attention patterns to KV cache quantization errors and study why key cache is more important than value cache for quantization error reduction. We further propose a simple yet effective framework KVTuner to adaptively search for the optimal hardware-friendly layer-wise KV quantization precision pairs for coarse-grained KV cache with multi-objective optimization and directly utilize the offline searched configurations during online inference. To reduce the computational cost of offline calibration, we utilize the intra-layer KV precision pair pruning and inter-layer clustering to reduce the search space. Experimental results show that we can achieve nearly lossless 3.25-bit mixed precision KV cache quantization for LLMs like Llama-3.1-8B-Instruct and 4.0-bit for sensitive models like Qwen2.5-7B-Instruct on mathematical reasoning tasks. The maximum inference throughput can be improved by 38.3% compared with KV8 quantization over various context lengths.
2502.04421
Assessing and Prioritizing Ransomware Risk Based on Historical Victim Data
cs.CR cs.AI cs.LG
We present an approach to identifying which ransomware adversaries are most likely to target specific entities, thereby assisting these entities in formulating better protection strategies. Ransomware poses a formidable cybersecurity threat characterized by profit-driven motives, a complex underlying economy supporting criminal syndicates, and the overt nature of its attacks. This type of malware has consistently ranked among the most prevalent, with a rapid escalation in activity observed. Recent estimates indicate that approximately two-thirds of organizations experienced ransomware attacks in 2023 \cite{Sophos2023Ransomware}. A central tactic in ransomware campaigns is publicizing attacks to coerce victims into paying ransoms. Our study utilizes public disclosures from ransomware victims to predict the likelihood of an entity being targeted by a specific ransomware variant. We employ a Large Language Model (LLM) architecture that uses a unique chain-of-thought, multi-shot prompt methodology to define adversary SKRAM (Skills, Knowledge, Resources, Authorities, and Motivation) profiles from ransomware bulletins, threat reports, and news items. This analysis is enriched with publicly available victim data and is further enhanced by a heuristic for generating synthetic data that reflects victim profiles. Our work culminates in the development of a machine learning model that assists organizations in prioritizing ransomware threats and formulating defenses based on the tactics, techniques, and procedures (TTP) of the most likely attackers.
2502.04423
Primary Care Diagnoses as a Reliable Predictor for Orthopedic Surgical Interventions
cs.LG cs.AI cs.CL
Referral workflow inefficiencies, including misaligned referrals and delays, contribute to suboptimal patient outcomes and higher healthcare costs. In this study, we investigated the possibility of predicting procedural needs based on primary care diagnostic entries, thereby improving referral accuracy, streamlining workflows, and providing better care to patients. A de-identified dataset of 2,086 orthopedic referrals from the University of Texas Health at Tyler was analyzed using machine learning models built on Base General Embeddings (BGE) for semantic extraction. To ensure real-world applicability, noise tolerance experiments were conducted, and oversampling techniques were employed to mitigate class imbalance. The selected optimum and parsimonious embedding model demonstrated high predictive accuracy (ROC-AUC: 0.874, Matthews Correlation Coefficient (MCC): 0.540), effectively distinguishing patients requiring surgical intervention. Dimensionality reduction techniques confirmed the model's ability to capture meaningful clinical relationships. A threshold sensitivity analysis identified an optimal decision threshold (0.30) to balance precision and recall, maximizing referral efficiency. In the predictive modeling analysis, the procedure rate increased from 11.27% to an optimal 60.1%, representing a 433% improvement with significant implications for operational efficiency and healthcare revenue. The results of our study demonstrate that referral optimization can enhance primary and surgical care integration. Through this approach, precise and timely predictions of procedural requirements can be made, thereby minimizing delays, improving surgical planning, and reducing administrative burdens. In addition, the findings highlight the potential of clinical decision support as a scalable solution for improving patient outcomes and the efficiency of the healthcare system.
2502.04424
EmoBench-M: Benchmarking Emotional Intelligence for Multimodal Large Language Models
cs.CL cs.AI
With the integration of Multimodal large language models (MLLMs) into robotic systems and various AI applications, embedding emotional intelligence (EI) capabilities into these models is essential for enabling robots to effectively address human emotional needs and interact seamlessly in real-world scenarios. Existing static, text-based, or text-image benchmarks overlook the multimodal complexities of real-world interactions and fail to capture the dynamic, multimodal nature of emotional expressions, making them inadequate for evaluating MLLMs' EI. Based on established psychological theories of EI, we build EmoBench-M, a novel benchmark designed to evaluate the EI capability of MLLMs across 13 valuation scenarios from three key dimensions: foundational emotion recognition, conversational emotion understanding, and socially complex emotion analysis. Evaluations of both open-source and closed-source MLLMs on EmoBench-M reveal a significant performance gap between them and humans, highlighting the need to further advance their EI capabilities. All benchmark resources, including code and datasets, are publicly available at https://emo-gml.github.io/.
2502.04426
Decoding AI Judgment: How LLMs Assess News Credibility and Bias
cs.CL cs.AI cs.CY
Large Language Models (LLMs) are increasingly used to assess news credibility, yet little is known about how they make these judgments. While prior research has examined political bias in LLM outputs or their potential for automated fact-checking, their internal evaluation processes remain largely unexamined. Understanding how LLMs assess credibility provides insights into AI behavior and how credibility is structured and applied in large-scale language models. This study benchmarks the reliability and political classifications of state-of-the-art LLMs - Gemini 1.5 Flash (Google), GPT-4o mini (OpenAI), and LLaMA 3.1 (Meta) - against structured, expert-driven rating systems such as NewsGuard and Media Bias Fact Check. Beyond assessing classification performance, we analyze the linguistic markers that shape LLM decisions, identifying which words and concepts drive their evaluations. We uncover patterns in how LLMs associate credibility with specific linguistic features by examining keyword frequency, contextual determinants, and rank distributions. Beyond static classification, we introduce a framework in which LLMs refine their credibility assessments by retrieving external information, querying other models, and adapting their responses. This allows us to investigate whether their assessments reflect structured reasoning or rely primarily on prior learned associations.
2502.04428
Confident or Seek Stronger: Exploring Uncertainty-Based On-device LLM Routing From Benchmarking to Generalization
cs.CL cs.AI cs.LG
Large language models (LLMs) are increasingly deployed and democratized on edge devices. To improve the efficiency of on-device deployment, small language models (SLMs) are often adopted due to their efficient decoding latency and reduced energy consumption. However, these SLMs often generate inaccurate responses when handling complex queries. One promising solution is uncertainty-based SLM routing, offloading high-stakes queries to stronger LLMs when resulting in low-confidence responses on SLM. This follows the principle of "If you lack confidence, seek stronger support" to enhance reliability. Relying on more powerful LLMs is yet effective but increases invocation costs. Therefore, striking a routing balance between efficiency and efficacy remains a critical challenge. Additionally, efficiently generalizing the routing strategy to new datasets remains under-explored. In this paper, we conduct a comprehensive investigation into benchmarking and generalization of uncertainty-driven routing strategies from SLMs to LLMs over 1500+ settings. Our findings highlight: First, uncertainty-correctness alignment in different uncertainty quantification (UQ) methods significantly impacts routing performance. Second, uncertainty distributions depend more on both the specific SLM and the chosen UQ method, rather than downstream data. Building on the insight, we propose a calibration data construction instruction pipeline and open-source a constructed hold-out set to enhance routing generalization on new downstream scenarios. The experimental results indicate calibration data effectively bootstraps routing performance without any new data.
2502.04457
"In order that" -- a data driven study of symptoms and causes of obsolescence
cs.CL cs.CY
The paper is an empirical case study of grammatical obsolescence in progress. The main studied variable is the purpose subordinator in order that, which is shown to be steadily decreasing in the frequency of use starting from the beginning of the twentieth century. This work applies a data-driven approach for the investigation and description of obsolescence, recently developed by the Rudnicka (2019). The methodology combines philological analysis with statistical methods used on data acquired from mega-corpora. Moving from the description of possible symptoms of obsolescence to different causes for it, the paper aims at presenting a comprehensive account of the studied phenomenon. Interestingly, a very significant role in the decline of in order that can be ascribed to the so-called higher-order processes, understood as processes influencing the constructional level from above. Two kinds of higher-order processes are shown to play an important role, namely i) an externally-motivated higher-order process exemplified by the drastic socio-cultural changes of the 19th and 20th centuries; ii) an internally-motivated higher-order processes instantiated by the rise of the to-infinitive (rise of infinite clauses).
2502.04463
Training Language Models to Reason Efficiently
cs.LG cs.CL
Scaling model size and training data has led to great advances in the performance of Large Language Models (LLMs). However, the diminishing returns of this approach necessitate alternative methods to improve model capabilities, particularly in tasks requiring advanced reasoning. Large reasoning models, which leverage long chain-of-thoughts, bring unprecedented breakthroughs in problem-solving capabilities but at a substantial deployment cost associated to longer generations. Reducing inference costs is crucial for the economic feasibility, user experience, and environmental sustainability of these models. In this work, we propose to train large reasoning models to reason efficiently. More precisely, we use reinforcement learning (RL) to train reasoning models to dynamically allocate inference-time compute based on task complexity. Our method incentivizes models to minimize unnecessary computational overhead while maintaining accuracy, thereby achieving substantial efficiency gains. It enables the derivation of a family of reasoning models with varying efficiency levels, controlled via a single hyperparameter. Experiments on two open-weight large reasoning models demonstrate significant reductions in inference cost while preserving most of the accuracy.
2502.04465
FocalCodec: Low-Bitrate Speech Coding via Focal Modulation Networks
cs.LG cs.AI cs.SD eess.AS
Large language models have revolutionized natural language processing through self-supervised pretraining on massive datasets. Inspired by this success, researchers have explored adapting these methods to speech by discretizing continuous audio into tokens using neural audio codecs. However, existing approaches face limitations, including high bitrates, the loss of either semantic or acoustic information, and the reliance on multi-codebook designs when trying to capture both, which increases architectural complexity for downstream tasks. To address these challenges, we introduce FocalCodec, an efficient low-bitrate codec based on focal modulation that utilizes a single binary codebook to compress speech between 0.16 and 0.65 kbps. FocalCodec delivers competitive performance in speech resynthesis and voice conversion at lower bitrates than the current state-of-the-art, while effectively handling multilingual speech and noisy environments. Evaluation on downstream tasks shows that FocalCodec successfully preserves sufficient semantic and acoustic information, while also being well-suited for generative modeling. Demo samples, code and checkpoints are available at https://lucadellalib.github.io/focalcodec-web/.
2502.04467
Efficient variable-length hanging tether parameterization for marsupial robot planning in 3D environments
cs.RO
This paper presents a novel approach to efficiently parameterize and estimate the state of a hanging tether for path and trajectory planning of a UGV tied to a UAV in a marsupial configuration. Most implementations in the state of the art assume a taut tether or make use of the catenary curve to model the shape of the hanging tether. The catenary model is complex to compute and must be instantiated thousands of times during the planning process, becoming a time-consuming task, while the taut tether assumption simplifies the problem, but might overly restrict the movement of the platforms. In order to accelerate the planning process, this paper proposes defining an analytical model to efficiently compute the hanging tether state, and a method to get a tether state parameterization free of collisions. We exploit the existing similarity between the catenary and parabola curves to derive analytical expressions of the tether state.
2502.04468
Iterative Importance Fine-tuning of Diffusion Models
cs.LG eess.IV math.PR
Diffusion models are an important tool for generative modelling, serving as effective priors in applications such as imaging and protein design. A key challenge in applying diffusion models for downstream tasks is efficiently sampling from resulting posterior distributions, which can be addressed using the $h$-transform. This work introduces a self-supervised algorithm for fine-tuning diffusion models by estimating the $h$-transform, enabling amortised conditional sampling. Our method iteratively refines the $h$-transform using a synthetic dataset resampled with path-based importance weights. We demonstrate the effectiveness of this framework on class-conditional sampling and reward fine-tuning for text-to-image diffusion models.
2502.04469
No Images, No Problem: Retaining Knowledge in Continual VQA with Questions-Only Memory
cs.CV cs.AI
Continual Learning in Visual Question Answering (VQACL) requires models to learn new visual-linguistic tasks (plasticity) while retaining knowledge from previous tasks (stability). The multimodal nature of VQACL presents unique challenges, requiring models to balance stability across visual and textual domains while maintaining plasticity to adapt to novel objects and reasoning tasks. Existing methods, predominantly designed for unimodal tasks, often struggle to balance these demands effectively. In this work, we introduce QUestion-only replay with Attention Distillation (QUAD), a novel approach for VQACL that leverages only past task questions for regularisation, eliminating the need to store visual data and addressing both memory and privacy concerns. QUAD achieves stability by introducing a question-only replay mechanism that selectively uses questions from previous tasks to prevent overfitting to the current task's answer space, thereby mitigating the out-of-answer-set problem. Complementing this, we propose attention consistency distillation, which uniquely enforces both intra-modal and inter-modal attention consistency across tasks, preserving essential visual-linguistic associations. Extensive experiments on VQAv2 and NExT-QA demonstrate that QUAD significantly outperforms state-of-the-art methods, achieving robust performance in continual VQA.
2502.04470
Color in Visual-Language Models: CLIP deficiencies
cs.CV cs.AI
This work explores how color is encoded in CLIP (Contrastive Language-Image Pre-training) which is currently the most influential VML (Visual Language model) in Artificial Intelligence. After performing different experiments on synthetic datasets created for this task, we conclude that CLIP is able to attribute correct color labels to colored visual stimulus, but, we come across two main deficiencies: (a) a clear bias on achromatic stimuli that are poorly related to the color concept, thus white, gray and black are rarely assigned as color labels; and (b) the tendency to prioritize text over other visual information. Here we prove it is highly significant in color labelling through an exhaustive Stroop-effect test. With the aim to find the causes of these color deficiencies, we analyse the internal representation at the neuron level. We conclude that CLIP presents an important amount of neurons selective to text, specially in deepest layers of the network, and a smaller amount of multi-modal color neurons which could be the key of understanding the concept of color properly. Our investigation underscores the necessity of refining color representation mechanisms in neural networks to foster a more comprehensive comprehension of colors as humans understand them, thereby advancing the efficacy and versatility of multimodal models like CLIP in real-world scenarios.
2502.04471
Identifying Flaky Tests in Quantum Code: A Machine Learning Approach
cs.SE cs.LG
Testing and debugging quantum software pose significant challenges due to the inherent complexities of quantum mechanics, such as superposition and entanglement. One challenge is indeterminacy, a fundamental characteristic of quantum systems, which increases the likelihood of flaky tests in quantum programs. To the best of our knowledge, there is a lack of comprehensive studies on quantum flakiness in the existing literature. In this paper, we present a novel machine learning platform that leverages multiple machine learning models to automatically detect flaky tests in quantum programs. Our evaluation shows that the extreme gradient boosting and decision tree-based models outperform other models (i.e., random forest, k-nearest neighbors, and support vector machine), achieving the highest F1 score and Matthews Correlation Coefficient in a balanced dataset and an imbalanced dataset, respectively. Furthermore, we expand the currently limited dataset for researchers interested in quantum flaky tests. In the future, we plan to explore the development of unsupervised learning techniques to detect and classify quantum flaky tests more effectively. These advancements aim to improve the reliability and robustness of quantum software testing.
2502.04475
Augmented Conditioning Is Enough For Effective Training Image Generation
cs.CV cs.AI cs.LG
Image generation abilities of text-to-image diffusion models have significantly advanced, yielding highly photo-realistic images from descriptive text and increasing the viability of leveraging synthetic images to train computer vision models. To serve as effective training data, generated images must be highly realistic while also sufficiently diverse within the support of the target data distribution. Yet, state-of-the-art conditional image generation models have been primarily optimized for creative applications, prioritizing image realism and prompt adherence over conditional diversity. In this paper, we investigate how to improve the diversity of generated images with the goal of increasing their effectiveness to train downstream image classification models, without fine-tuning the image generation model. We find that conditioning the generation process on an augmented real image and text prompt produces generations that serve as effective synthetic datasets for downstream training. Conditioning on real training images contextualizes the generation process to produce images that are in-domain with the real image distribution, while data augmentations introduce visual diversity that improves the performance of the downstream classifier. We validate augmentation-conditioning on a total of five established long-tail and few-shot image classification benchmarks and show that leveraging augmentations to condition the generation process results in consistent improvements over the state-of-the-art on the long-tailed benchmark and remarkable gains in extreme few-shot regimes of the remaining four benchmarks. These results constitute an important step towards effectively leveraging synthetic data for downstream training.
2502.04476
ADIFF: Explaining audio difference using natural language
cs.SD cs.AI eess.AS
Understanding and explaining differences between audio recordings is crucial for fields like audio forensics, quality assessment, and audio generation. This involves identifying and describing audio events, acoustic scenes, signal characteristics, and their emotional impact on listeners. This paper stands out as the first work to comprehensively study the task of explaining audio differences and then propose benchmark, baselines for the task. First, we present two new datasets for audio difference explanation derived from the AudioCaps and Clotho audio captioning datasets. Using Large Language Models (LLMs), we generate three levels of difference explanations: (1) concise descriptions of audio events and objects, (2) brief sentences about audio events, acoustic scenes, and signal properties, and (3) comprehensive explanations that include semantics and listener emotions. For the baseline, we use prefix tuning where audio embeddings from two audio files are used to prompt a frozen language model. Our empirical analysis and ablation studies reveal that the naive baseline struggles to distinguish perceptually similar sounds and generate detailed tier 3 explanations. To address these limitations, we propose ADIFF, which introduces a cross-projection module, position captioning, and a three-step training process to enhance the model's ability to produce detailed explanations. We evaluate our model using objective metrics and human evaluation and show our model enhancements lead to significant improvements in performance over naive baseline and SoTA Audio-Language Model (ALM) Qwen Audio. Lastly, we conduct multiple ablation studies to study the effects of cross-projection, language model parameters, position captioning, third stage fine-tuning, and present our findings. Our benchmarks, findings, and strong baseline pave the way for nuanced and human-like explanations of audio differences.
2502.04478
OneTrack-M: A multitask approach to transformer-based MOT models
cs.CV cs.LG
Multi-Object Tracking (MOT) is a critical problem in computer vision, essential for understanding how objects move and interact in videos. This field faces significant challenges such as occlusions and complex environmental dynamics, impacting model accuracy and efficiency. While traditional approaches have relied on Convolutional Neural Networks (CNNs), introducing transformers has brought substantial advancements. This work introduces OneTrack-M, a transformer-based MOT model designed to enhance tracking computational efficiency and accuracy. Our approach simplifies the typical transformer-based architecture by eliminating the need for a decoder model for object detection and tracking. Instead, the encoder alone serves as the backbone for temporal data interpretation, significantly reducing processing time and increasing inference speed. Additionally, we employ innovative data pre-processing and multitask training techniques to address occlusion and diverse objective challenges within a single set of weights. Experimental results demonstrate that OneTrack-M achieves at least 25% faster inference times compared to state-of-the-art models in the literature while maintaining or improving tracking accuracy metrics. These improvements highlight the potential of the proposed solution for real-time applications such as autonomous vehicles, surveillance systems, and robotics, where rapid responses are crucial for system effectiveness.
2502.04480
On Techniques for Barely Coupled Multiphysics
cs.CE
A technique to combine codes to solve barely coupled multiphysics problems has been developed. Each field is advanced separately until a stop is triggered. This could be due to a preset time increment, a preset number of timesteps, a preset decrease of residuals, a preset change in unknowns, a preset change in geometry, or any other physically meaningful quantity. The technique allows for a simple implementation in coupled codes using the loose coupling approach. Examples from evaporative cooling of electric motors, a problem that has come to the forefront with the rise of electric propulsion in the aerospace sector (drones and air taxis in particular) shows the viability and accuracy of the proposed procedure.
2502.04483
Measuring Physical Plausibility of 3D Human Poses Using Physics Simulation
cs.CV
Modeling humans in physical scenes is vital for understanding human-environment interactions for applications involving augmented reality or assessment of human actions from video (e.g. sports or physical rehabilitation). State-of-the-art literature begins with a 3D human pose, from monocular or multiple views, and uses this representation to ground the person within a 3D world space. While standard metrics for accuracy capture joint position errors, they do not consider physical plausibility of the 3D pose. This limitation has motivated researchers to propose other metrics evaluating jitter, floor penetration, and unbalanced postures. Yet, these approaches measure independent instances of errors and are not representative of balance or stability during motion. In this work, we propose measuring physical plausibility from within physics simulation. We introduce two metrics to capture the physical plausibility and stability of predicted 3D poses from any 3D Human Pose Estimation model. Using physics simulation, we discover correlations with existing plausibility metrics and measuring stability during motion. We evaluate and compare the performances of two state-of-the-art methods, a multi-view triangulated baseline, and ground truth 3D markers from the Human3.6m dataset.
2502.04484
The ML Supply Chain in the Era of Software 2.0: Lessons Learned from Hugging Face
cs.SE cs.LG
The last decade has seen widespread adoption of Machine Learning (ML) components in software systems. This has occurred in nearly every domain, from natural language processing to computer vision. These ML components range from relatively simple neural networks to complex and resource-intensive large language models. However, despite this widespread adoption, little is known about the supply chain relationships that produce these models, which can have implications for compliance and security. In this work, we conduct an extensive analysis of 760,460 models and 175,000 datasets mined from the popular model-sharing site Hugging Face. First, we evaluate the current state of documentation in the Hugging Face supply chain, report real-world examples of shortcomings, and offer actionable suggestions for improvement. Next, we analyze the underlying structure of the extant supply chain. Finally, we explore the current licensing landscape against what was reported in prior work and discuss the unique challenges posed in this domain. Our results motivate multiple research avenues, including the need for better license management for ML models/datasets, better support for model documentation, and automated inconsistency checking and validation. We make our research infrastructure and dataset available to facilitate future research.
2502.04485
Active Task Disambiguation with LLMs
cs.CL cs.AI cs.LG
Despite the impressive performance of large language models (LLMs) across various benchmarks, their ability to address ambiguously specified problems--frequent in real-world interactions--remains underexplored. To address this gap, we introduce a formal definition of task ambiguity and frame the problem of task disambiguation through the lens of Bayesian Experimental Design. By posing clarifying questions, LLM agents can acquire additional task specifications, progressively narrowing the space of viable solutions and reducing the risk of generating unsatisfactory outputs. Yet, generating effective clarifying questions requires LLM agents to engage in a form of meta-cognitive reasoning, an ability LLMs may presently lack. Our proposed approach of active task disambiguation enables LLM agents to generate targeted questions maximizing the information gain. Effectively, this approach shifts the load from implicit to explicit reasoning about the space of viable solutions. Empirical results demonstrate that this form of question selection leads to more effective task disambiguation in comparison to approaches relying on reasoning solely within the space of questions.
2502.04488
Building A Unified AI-centric Language System: analysis, framework and future work
cs.CL cs.AI
Recent advancements in large language models have demonstrated that extended inference through techniques can markedly improve performance, yet these gains come with increased computational costs and the propagation of inherent biases found in natural languages. This paper explores the design of a unified AI-centric language system that addresses these challenges by offering a more concise, unambiguous, and computationally efficient alternative to traditional human languages. We analyze the limitations of natural language such as gender bias, morphological irregularities, and contextual ambiguities and examine how these issues are exacerbated within current Transformer architectures, where redundant attention heads and token inefficiencies prevail. Drawing on insights from emergent artificial communication systems and constructed languages like Esperanto and Lojban, we propose a framework that translates diverse natural language inputs into a streamlined AI-friendly language, enabling more efficient model training and inference while reducing memory footprints. Finally, we outline a pathway for empirical validation through controlled experiments, paving the way for a universal interchange format that could revolutionize AI-to-AI and human-to-AI interactions by enhancing clarity, fairness, and overall performance.
2502.04489
CNN Autoencoders for Hierarchical Feature Extraction and Fusion in Multi-sensor Human Activity Recognition
cs.LG cs.AI
Deep learning methods have been widely used for Human Activity Recognition (HAR) using recorded signals from Iner-tial Measurement Units (IMUs) sensors that are installed on various parts of the human body. For this type of HAR, sev-eral challenges exist, the most significant of which is the analysis of multivarious IMU sensors data. Here, we introduce a Hierarchically Unsupervised Fusion (HUF) model designed to extract, and fuse features from IMU sensors data via a hybrid structure of Convolutional Neural Networks (CNN)s and Autoencoders (AE)s. First, we design a stack CNN-AE to embed short-time signals into sets of high dimensional features. Second, we develop another CNN-AE network to locally fuse the extracted features from each sensor unit. Finally, we unify all the sensor features through a third CNN-AE architecture as globally feature fusion to create a unique feature set. Additionally, we analyze the effects of varying the model hyperparameters. The best results are achieved with eight convolutional layers in each AE. Furthermore, it is determined that an overcomplete AE with 256 kernels in the code layer is suitable for feature extraction in the first block of the proposed HUF model; this number reduces to 64 in the last block of the model to customize the size of the applied features to the classifier. The tuned model is applied to the UCI-HAR, DaLiAc, and Parkinson's disease gait da-tasets, achieving the classification accuracies of 97%, 97%, and 88%, respectively, which are nearly 3% better com-pared to the state-of-the-art supervised methods.
2502.04491
Provable Sample-Efficient Transfer Learning Conditional Diffusion Models via Representation Learning
cs.LG math.ST stat.ML stat.TH
While conditional diffusion models have achieved remarkable success in various applications, they require abundant data to train from scratch, which is often infeasible in practice. To address this issue, transfer learning has emerged as an essential paradigm in small data regimes. Despite its empirical success, the theoretical underpinnings of transfer learning conditional diffusion models remain unexplored. In this paper, we take the first step towards understanding the sample efficiency of transfer learning conditional diffusion models through the lens of representation learning. Inspired by practical training procedures, we assume that there exists a low-dimensional representation of conditions shared across all tasks. Our analysis shows that with a well-learned representation from source tasks, the samplecomplexity of target tasks can be reduced substantially. In addition, we investigate the practical implications of our theoretical results in several real-world applications of conditional diffusion models. Numerical experiments are also conducted to verify our results.
2502.04492
Multi-Agent Reinforcement Learning with Focal Diversity Optimization
cs.CL
The advancement of Large Language Models (LLMs) and their finetuning strategies has triggered the renewed interests in multi-agent reinforcement learning. In this paper, we introduce a focal diversity-optimized multi-agent reinforcement learning approach, coined as MARL-Focal, with three unique characteristics. First, we develop an agent-fusion framework for encouraging multiple LLM based agents to collaborate in producing the final inference output for each LLM query. Second, we develop a focal-diversity optimized agent selection algorithm that can choose a small subset of the available agents based on how well they can complement one another to generate the query output. Finally, we design a conflict-resolution method to detect output inconsistency among multiple agents and produce our MARL-Focal output through reward-aware and policy-adaptive inference fusion. Extensive evaluations on five benchmarks show that MARL-Focal is cost-efficient and adversarial-robust. Our multi-agent fusion model achieves performance improvement of 5.51\% compared to the best individual LLM-agent and offers stronger robustness over the TruthfulQA benchmark. Code is available at https://github.com/sftekin/rl-focal
2502.04493
LUND-PROBE -- LUND Prostate Radiotherapy Open Benchmarking and Evaluation dataset
physics.med-ph cs.CV eess.IV
Radiotherapy treatment for prostate cancer relies on computed tomography (CT) and/or magnetic resonance imaging (MRI) for segmentation of target volumes and organs at risk (OARs). Manual segmentation of these volumes is regarded as the gold standard for ground truth in machine learning applications but to acquire such data is tedious and time-consuming. A publicly available clinical dataset is presented, comprising MRI- and synthetic CT (sCT) images, target and OARs segmentations, and radiotherapy dose distributions for 432 prostate cancer patients treated with MRI-guided radiotherapy. An extended dataset with 35 patients is also included, with the addition of deep learning (DL)-generated segmentations, DL segmentation uncertainty maps, and DL segmentations manually adjusted by four radiation oncologists. The publication of these resources aims to aid research within the fields of automated radiotherapy treatment planning, segmentation, inter-observer analyses, and DL model uncertainty investigation. The dataset is hosted on the AIDA Data Hub and offers a free-to-use resource for the scientific community, valuable for the advancement of medical imaging and prostate cancer radiotherapy research.
2502.04495
Discovering Physics Laws of Dynamical Systems via Invariant Function Learning
cs.LG
We consider learning underlying laws of dynamical systems governed by ordinary differential equations (ODE). A key challenge is how to discover intrinsic dynamics across multiple environments while circumventing environment-specific mechanisms. Unlike prior work, we tackle more complex environments where changes extend beyond function coefficients to entirely different function forms. For example, we demonstrate the discovery of ideal pendulum's natural motion $\alpha^2 \sin{\theta_t}$ by observing pendulum dynamics in different environments, such as the damped environment $\alpha^2 \sin(\theta_t) - \rho \omega_t$ and powered environment $\alpha^2 \sin(\theta_t) + \rho \frac{\omega_t}{\left|\omega_t\right|}$. Here, we formulate this problem as an \emph{invariant function learning} task and propose a new method, known as \textbf{D}isentanglement of \textbf{I}nvariant \textbf{F}unctions (DIF), that is grounded in causal analysis. We propose a causal graph and design an encoder-decoder hypernetwork that explicitly disentangles invariant functions from environment-specific dynamics. The discovery of invariant functions is guaranteed by our information-based principle that enforces the independence between extracted invariant functions and environments. Quantitative comparisons with meta-learning and invariant learning baselines on three ODE systems demonstrate the effectiveness and efficiency of our method. Furthermore, symbolic regression explanation results highlight the ability of our framework to uncover intrinsic laws.
2502.04497
Distributed Resilient Asymmetric Bipartite Consensus: A Data-Driven Event-Triggered Mechanism
eess.SY cs.SY
The problem of asymmetric bipartite consensus control is investigated within the context of nonlinear, discrete-time, networked multi-agent systems (MAS) subject to aperiodic denial-of-service (DoS) attacks. To address the challenges posed by these aperiodic DoS attacks, a data-driven event-triggered (DDET) mechanism has been developed. This mechanism is specifically designed to synchronize the states of the follower agents with the leader's state, even in the face of aperiodic communication disruptions and data losses. Given the constraints of unavailable agents' states and data packet loss during these attacks, the DDET control framework resiliently achieves leader-follower consensus. The effectiveness of the proposed framework is validated through two numerical examples, which showcase its ability to adeptly handle the complexities arising from aperiodic DoS attacks in nonlinear MAS settings.
2502.04498
Verifiable Format Control for Large Language Model Generations
cs.CL
Recent Large Language Models (LLMs) have demonstrated satisfying general instruction following ability. However, small LLMs with about 7B parameters still struggle fine-grained format following (e.g., JSON format), which seriously hinder the advancements of their applications. Most existing methods focus on benchmarking general instruction following while overlook how to improve the specific format following ability for small LLMs. Besides, these methods often rely on evaluations based on advanced LLMs (e.g., GPT-4), which can introduce the intrinsic bias of LLMs and be costly due to the API calls. In this paper, we first curate a fully verifiable format following dataset VFF. In contrast to existing works often adopting external LLMs for instruction-following validations, every sample of VFF can be easily validated with a Python function. Further, we propose to leverage this verifiable feature to synthesize massive data for progressively training small LLMs, in order to improve their format following abilities. Experimental results highlight the prevalent limitations in the format following capabilities of 7B level open-source LLMs and demonstrate the effectiveness of our method in enhancing this essential ability.
2502.04499
Revisiting Intermediate-Layer Matching in Knowledge Distillation: Layer-Selection Strategy Doesn't Matter (Much)
cs.LG cs.AI cs.CL
Knowledge distillation (KD) is a popular method of transferring knowledge from a large "teacher" model to a small "student" model. KD can be divided into two categories: prediction matching and intermediate-layer matching. We explore an intriguing phenomenon: layer-selection strategy does not matter (much) in intermediate-layer matching. In this paper, we show that seemingly nonsensical matching strategies such as matching the teacher's layers in reverse still result in surprisingly good student performance. We provide an interpretation for this phenomenon by examining the angles between teacher layers viewed from the student's perspective.
2502.04501
ULPT: Prompt Tuning with Ultra-Low-Dimensional Optimization
cs.CL
Large language models achieve state-of-the-art performance but are costly to fine-tune due to their size. Parameter-efficient fine-tuning methods, such as prompt tuning, address this by reducing trainable parameters while maintaining strong performance. However, prior methods tie prompt embeddings to the model's dimensionality, which may not scale well with larger LLMs and more customized LLMs. In this paper, we propose Ultra-Low-dimensional Prompt Tuning (ULPT), which optimizes prompts in a low-dimensional space (e.g., 2D) and use a random but frozen matrix for the up-projection. To enhance alignment, we introduce learnable shift and scale embeddings. ULPT drastically reduces the trainable parameters, e.g., 2D only using 2% parameters compared with vanilla prompt tuning while retaining most of the performance across 21 NLP tasks. Our theoretical analysis shows that random projections can capture high-rank structures effectively, and experimental results demonstrate ULPT's competitive performance over existing parameter-efficient methods.
2502.04506
When One LLM Drools, Multi-LLM Collaboration Rules
cs.CL
This position paper argues that in many realistic (i.e., complex, contextualized, subjective) scenarios, one LLM is not enough to produce a reliable output. We challenge the status quo of relying solely on a single general-purpose LLM and argue for multi-LLM collaboration to better represent the extensive diversity of data, skills, and people. We first posit that a single LLM underrepresents real-world data distributions, heterogeneous skills, and pluralistic populations, and that such representation gaps cannot be trivially patched by further training a single LLM. We then organize existing multi-LLM collaboration methods into a hierarchy, based on the level of access and information exchange, ranging from API-level, text-level, logit-level, to weight-level collaboration. Based on these methods, we highlight how multi-LLM collaboration addresses challenges that a single LLM struggles with, such as reliability, democratization, and pluralism. Finally, we identify the limitations of existing multi-LLM methods and motivate future work. We envision multi-LLM collaboration as an essential path toward compositional intelligence and collaborative AI development.
2502.04507
Fast Video Generation with Sliding Tile Attention
cs.CV
Diffusion Transformers (DiTs) with 3D full attention power state-of-the-art video generation, but suffer from prohibitive compute cost -- when generating just a 5-second 720P video, attention alone takes 800 out of 945 seconds of total inference time. This paper introduces sliding tile attention (STA) to address this challenge. STA leverages the observation that attention scores in pretrained video diffusion models predominantly concentrate within localized 3D windows. By sliding and attending over the local spatial-temporal region, STA eliminates redundancy from full attention. Unlike traditional token-wise sliding window attention (SWA), STA operates tile-by-tile with a novel hardware-aware sliding window design, preserving expressiveness while being hardware-efficient. With careful kernel-level optimizations, STA offers the first efficient 2D/3D sliding-window-like attention implementation, achieving 58.79% MFU. Precisely, STA accelerates attention by 2.8-17x over FlashAttention-2 (FA2) and 1.6-10x over FlashAttention-3 (FA3). On the leading video DiT, HunyuanVideo, STA reduces end-to-end latency from 945s (FA3) to 685s without quality degradation, requiring no training. Enabling finetuning further lowers latency to 268s with only a 0.09% drop on VBench.
2502.04510
Heterogeneous Swarms: Jointly Optimizing Model Roles and Weights for Multi-LLM Systems
cs.CL
We propose Heterogeneous Swarms, an algorithm to design multi-LLM systems by jointly optimizing model roles and weights. We represent multi-LLM systems as directed acyclic graphs (DAGs) of LLMs with topological message passing for collaborative generation. Given a pool of LLM experts and a utility function, Heterogeneous Swarms employs two iterative steps: role-step and weight-step. For role-step, we interpret model roles as learning a DAG that specifies the flow of inputs and outputs between LLMs. Starting from a swarm of random continuous adjacency matrices, we decode them into discrete DAGs, call the LLMs in topological order, evaluate on the utility function (e.g. accuracy on a task), and optimize the adjacency matrices with particle swarm optimization based on the utility score. For weight-step, we assess the contribution of individual LLMs in the multi-LLM systems and optimize model weights with swarm intelligence. We propose JFK-score to quantify the individual contribution of each LLM in the best-found DAG of the role-step, then optimize model weights with particle swarm optimization based on the JFK-score. Experiments demonstrate that Heterogeneous Swarms outperforms 15 role- and/or weight-based baselines by 18.5% on average across 12 tasks. Further analysis reveals that Heterogeneous Swarms discovers multi-LLM systems with heterogeneous model roles and substantial collaborative gains, and benefits from the diversity of language models.
2502.04511
Beyond Sample-Level Feedback: Using Reference-Level Feedback to Guide Data Synthesis
cs.CL
LLMs demonstrate remarkable capabilities in following natural language instructions, largely due to instruction-tuning on high-quality datasets. While synthetic data generation has emerged as a scalable approach for creating such datasets, maintaining consistent quality standards remains challenging. Recent approaches incorporate feedback to improve data quality, but typically operate at the sample level, generating and applying feedback for each response individually. In this work, we propose Reference-Level Feedback, a novel methodology that instead collects feedback based on high-quality reference samples from carefully curated seed data. We use this feedback to capture rich signals of desirable characteristics and propagate it throughout the data synthesis process. We present REFED, a dataset of 10K instruction-response pairs synthesized using such feedback. We demonstrate the effectiveness of our approach by showing that Llama-3.1-8B-Instruct finetuned on REFED achieves state-of-the-art performance among similar-sized SFT-based models on AlpacaEval 2.0 and strong results on Arena-Hard. Through extensive experiments, we show that our approach consistently outperforms traditional sample-level feedback methods with significantly fewer feedback collections and improves performance across different model architectures.
2502.04512
Safety is Essential for Responsible Open-Ended Systems
cs.AI
AI advancements have been significantly driven by a combination of foundation models and curiosity-driven learning aimed at increasing capability and adaptability. A growing area of interest within this field is Open-Endedness - the ability of AI systems to continuously and autonomously generate novel and diverse artifacts or solutions. This has become relevant for accelerating scientific discovery and enabling continual adaptation in AI agents. This position paper argues that the inherently dynamic and self-propagating nature of Open-Ended AI introduces significant, underexplored risks, including challenges in maintaining alignment, predictability, and control. This paper systematically examines these challenges, proposes mitigation strategies, and calls for action for different stakeholders to support the safe, responsible and successful development of Open-Ended AI.
2502.04515
MedGNN: Towards Multi-resolution Spatiotemporal Graph Learning for Medical Time Series Classification
cs.LG cs.AI
Medical time series has been playing a vital role in real-world healthcare systems as valuable information in monitoring health conditions of patients. Accurate classification for medical time series, e.g., Electrocardiography (ECG) signals, can help for early detection and diagnosis. Traditional methods towards medical time series classification rely on handcrafted feature extraction and statistical methods; with the recent advancement of artificial intelligence, the machine learning and deep learning methods have become more popular. However, existing methods often fail to fully model the complex spatial dynamics under different scales, which ignore the dynamic multi-resolution spatial and temporal joint inter-dependencies. Moreover, they are less likely to consider the special baseline wander problem as well as the multi-view characteristics of medical time series, which largely hinders their prediction performance. To address these limitations, we propose a Multi-resolution Spatiotemporal Graph Learning framework, MedGNN, for medical time series classification. Specifically, we first propose to construct multi-resolution adaptive graph structures to learn dynamic multi-scale embeddings. Then, to address the baseline wander problem, we propose Difference Attention Networks to operate self-attention mechanisms on the finite difference for temporal modeling. Moreover, to learn the multi-view characteristics, we utilize the Frequency Convolution Networks to capture complementary information of medical time series from the frequency domain. In addition, we introduce the Multi-resolution Graph Transformer architecture to model the dynamic dependencies and fuse the information from different resolutions. Finally, we have conducted extensive experiments on multiple medical real-world datasets that demonstrate the superior performance of our method. Our Code is available.
2502.04517
Towards Cost-Effective Reward Guided Text Generation
cs.LG cs.CL
Reward-guided text generation (RGTG) has emerged as a viable alternative to offline reinforcement learning from human feedback (RLHF). RGTG methods can align baseline language models to human preferences without further training like in standard RLHF methods. However, they rely on a reward model to score each candidate token generated by the language model at inference, incurring significant test-time overhead. Additionally, the reward model is usually only trained to score full sequences, which can lead to sub-optimal choices for partial sequences. In this work, we present a novel reward model architecture that is trained, using a Bradley-Terry loss, to prefer the optimal expansion of a sequence with just a \emph{single call} to the reward model at each step of the generation process. That is, a score for all possible candidate tokens is generated simultaneously, leading to efficient inference. We theoretically analyze various RGTG reward models and demonstrate that prior techniques prefer sub-optimal sequences compared to our method during inference. Empirically, our reward model leads to significantly faster inference than other RGTG methods. It requires fewer calls to the reward model and performs competitively compared to previous RGTG and offline RLHF methods.
2502.04519
GenVC: Self-Supervised Zero-Shot Voice Conversion
eess.AS cs.LG
Zero-shot voice conversion has recently made substantial progress, but many models still depend on external supervised systems to disentangle speaker identity and linguistic content. Furthermore, current methods often use parallel conversion, where the converted speech inherits the source utterance's temporal structure, restricting speaker similarity and privacy. To overcome these limitations, we introduce GenVC, a generative zero-shot voice conversion model. GenVC learns to disentangle linguistic content and speaker style in a self-supervised manner, eliminating the need for external models and enabling efficient training on large, unlabeled datasets. Experimental results show that GenVC achieves state-of-the-art speaker similarity while maintaining naturalness competitive with leading approaches. Its autoregressive generation also allows the converted speech to deviate from the source utterance's temporal structure. This feature makes GenVC highly effective for voice anonymization, as it minimizes the preservation of source prosody and speaker characteristics, enhancing privacy protection.
2502.04520
Linear Correlation in LM's Compositional Generalization and Hallucination
cs.CL
The generalization of language models (LMs) is undergoing active debates, contrasting their potential for general intelligence with their struggles with basic knowledge composition (e.g., reverse/transition curse). This paper uncovers the phenomenon of linear correlations in LMs during knowledge composition. For explanation, there exists a linear transformation between certain related knowledge that maps the next token prediction logits from one prompt to another, e.g., "X lives in the city of" $\rightarrow$ "X lives in the country of" for every given X. This mirrors the linearity in human knowledge composition, such as Paris $\rightarrow$ France. Our findings indicate that the linear transformation is resilient to large-scale fine-tuning, generalizing updated knowledge when aligned with real-world relationships, but causing hallucinations when it deviates. Empirical results suggest that linear correlation can serve as a potential identifier of LM's generalization. Finally, we show such linear correlations can be learned with a single feedforward network and pre-trained vocabulary representations, indicating LM generalization heavily relies on the latter.
2502.04521
Generative Autoregressive Transformers for Model-Agnostic Federated MRI Reconstruction
eess.IV cs.CV
Although learning-based models hold great promise for MRI reconstruction, single-site models built on limited local datasets often suffer from poor generalization. This challenge has spurred interest in collaborative model training on multi-site datasets via federated learning (FL) -- a privacy-preserving framework that aggregates model updates instead of sharing imaging data. Conventional FL builds a global model by aggregating locally trained model weights, inherently constraining all sites to a homogeneous model architecture. This rigid homogeneity requirement forces sites to forgo architectures tailored to their compute infrastructure and application-specific demands. Consequently, existing FL methods for MRI reconstruction fail to support model-heterogeneous settings, where individual sites are allowed to use distinct architectures. To overcome this fundamental limitation, here we introduce FedGAT, a novel model-agnostic FL technique based on generative autoregressive transformers. FedGAT decentralizes the training of a global generative prior that captures the distribution of multi-site MR images. For enhanced fidelity, we propose a novel site-prompted GAT prior that controllably synthesizes MR images from desired sites via autoregressive prediction across spatial scales. Each site then trains its site-specific reconstruction model -- using its preferred architecture -- on a hybrid dataset comprising the local MRI dataset and GAT-generated synthetic MRI datasets for other sites. Comprehensive experiments on multi-institutional datasets demonstrate that FedGAT supports flexible collaborations while enjoying superior within-site and across-site reconstruction performance compared to state-of-the-art FL baselines.
2502.04522
ImprovNet: Generating Controllable Musical Improvisations with Iterative Corruption Refinement
cs.SD cs.AI eess.AS
Deep learning has enabled remarkable advances in style transfer across various domains, offering new possibilities for creative content generation. However, in the realm of symbolic music, generating controllable and expressive performance-level style transfers for complete musical works remains challenging due to limited datasets, especially for genres such as jazz, and the lack of unified models that can handle multiple music generation tasks. This paper presents ImprovNet, a transformer-based architecture that generates expressive and controllable musical improvisations through a self-supervised corruption-refinement training strategy. ImprovNet unifies multiple capabilities within a single model: it can perform cross-genre and intra-genre improvisations, harmonize melodies with genre-specific styles, and execute short prompt continuation and infilling tasks. The model's iterative generation framework allows users to control the degree of style transfer and structural similarity to the original composition. Objective and subjective evaluations demonstrate ImprovNet's effectiveness in generating musically coherent improvisations while maintaining structural relationships with the original pieces. The model outperforms Anticipatory Music Transformer in short continuation and infilling tasks and successfully achieves recognizable genre conversion, with 79\% of participants correctly identifying jazz-style improvisations. Our code and demo page can be found at https://github.com/keshavbhandari/improvnet.
2502.04528
Group-Adaptive Threshold Optimization for Robust AI-Generated Text Detection
cs.CL cs.LG
The advancement of large language models (LLMs) has made it difficult to differentiate human-written text from AI-generated text. Several AI-text detectors have been developed in response, which typically utilize a fixed global threshold (e.g., {\theta} = 0.5) to classify machine-generated text. However, we find that one universal threshold can fail to account for subgroup-specific distributional variations. For example, when using a fixed threshold, detectors make more false positive errors on shorter human-written text than longer, and more positive classifications on neurotic writing styles than open among long text. These discrepancies can lead to misclassification that disproportionately affects certain groups. We address this critical limitation by introducing FairOPT, an algorithm for group-specific threshold optimization in AI-generated content classifiers. Our approach partitions data into subgroups based on attributes (e.g., text length and writing style) and learns decision thresholds for each group, which enables careful balancing of performance and fairness metrics within each subgroup. In experiments with four AI text classifiers on three datasets, FairOPT enhances overall F1 score and decreases balanced error rate (BER) discrepancy across subgroups. Our framework paves the way for more robust and fair classification criteria in AI-generated output detection.
2502.04529
Agricultural Field Boundary Detection through Integration of "Simple Non-Iterative Clustering (SNIC) Super Pixels" and "Canny Edge Detection Method"
cs.LG cs.CV
Efficient use of cultivated areas is a necessary factor for sustainable development of agriculture and ensuring food security. Along with the rapid development of satellite technologies in developed countries, new methods are being searched for accurate and operational identification of cultivated areas. In this context, identification of cropland boundaries based on spectral analysis of data obtained from satellite images is considered one of the most optimal and accurate methods in modern agriculture. This article proposes a new approach to determine the suitability and green index of cultivated areas using satellite data obtained through the "Google Earth Engine" (GEE) platform. In this approach, two powerful algorithms, "SNIC (Simple Non-Iterative Clustering) Super Pixels" and "Canny Edge Detection Method", are combined. The SNIC algorithm combines pixels in a satellite image into larger regions (super pixels) with similar characteristics, thereby providing better image analysis. The Canny Edge Detection Method detects sharp changes (edges) in the image to determine the precise boundaries of agricultural fields. This study, carried out using high-resolution multispectral data from the Sentinel-2 satellite and the Google Earth Engine JavaScript API, has shown that the proposed method is effective in accurately and reliably classifying randomly selected agricultural fields. The combined use of these two tools allows for more accurate determination of the boundaries of agricultural fields by minimizing the effects of outliers in satellite images. As a result, more accurate and reliable maps can be created for agricultural monitoring and resource management over large areas based on the obtained data. By expanding the application capabilities of cloud-based platforms and artificial intelligence methods in the agricultural field.