ICLR 2025 Papers — Page 6
International Conference on Learning Representations · 3704 papers
Causal Order: The Key to Leveraging Imperfect Experts in Causal Inference
Aniket Vashishtha (University of Illinois Urbana-Champaign), Amit Sharma (Microsoft Research)
TransformerLarge Language ModelPrompt EngineeringTabularBiomedical DataAlzheimer's Disease
🎯 What it does: Proposes using causal order instead of a complete causal graph as expert knowledge output, and designs a query method based on triplet prompts to improve the accuracy of obtaining causal structures from imperfect experts (LLMs and human annotators).
Causal Representation Learning from Multimodal Biomedical Observations
Yuewen Sun (Mohamed bin Zayed University of Artificial Intelligence), Kun Zhang (Mohamed bin Zayed University of Artificial Intelligence)
Representation LearningFlow-based ModelMultimodalityBiomedical Data
🎯 What it does: A causal representation learning framework based on multimodal biomedical observations is proposed, which can identifiably recover latent causal variables from multimodal observations.
Causally Motivated Sycophancy Mitigation for Large Language Models
Haoxi Li (Hong Kong University of Science and Technology), Yue Yu (Hong Kong Polytechnic University)
OptimizationExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper analyzes and eliminates sycophancy caused by user preferences in large language models (LLMs) through a Structural Causal Model (SCM), proposing the CAUSM framework, which suppresses spurious associations by utilizing causal direction calibration and attention head weight reallocation.
CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series
Gideon Stein (Friedrich Schiller University Jena), Joachim Denzler (Friedrich Schiller University Jena)
Time SeriesBenchmark
🎯 What it does: A real-world causal discovery benchmark based on river flow time series, CausalRivers, is proposed, which includes over 1000 monitoring stations, 15-minute resolution data, and provides subgraph sampling tools and three baselines.
CAX: Cellular Automata Accelerated in JAX
Maxence Faldor (Imperial College London), Antoine Cully (Imperial College London)
OptimizationComputational EfficiencyAuto EncoderTime Series
🎯 What it does: Developed and released the CAX library, achieving high-performance simulation and training of discrete, continuous, and neural cellular automata, demonstrating acceleration and new experiments in various tests.
CBGBench: Fill in the Blank of Protein-Molecule Complex Binding Graph
Haitao Lin (Westlake University), Stan Z. Li (Westlake University)
Drug DiscoveryConvolutional Neural NetworkGraph Neural NetworkDiffusion modelContrastive LearningGraphBiomedical DataAlzheimer's DiseaseBenchmark
🎯 What it does: This study proposes the CBGBench benchmark, unifying structure-based drug design (SBDD) and lead optimization tasks into a three-dimensional binding graph completion problem, and integrates 12 mainstream generative models with a unified evaluation framework;
CBMA: Improving Conformal Prediction through Bayesian Model Averaging
Pankaj Bhagwat (University of Alberta), Bei Jiang
TabularBenchmark
🎯 What it does: A CBMA method that combines Bayesian model averaging with conformal prediction is proposed to construct more efficient prediction intervals in the presence of model uncertainty.
CBQ: Cross-Block Quantization for Large Language Models
Xin Ding (University of Science and Technology of China), Yunhe Wang (Huawei Noah's Ark Lab)
CompressionOptimizationTransformerLarge Language ModelText
🎯 What it does: This paper proposes a Cross-Block Quantization method (CBQ) to address the significant quantization error of large language models at extremely low bit precision.
CBraMod: A Criss-Cross Brain Foundation Model for EEG Decoding
Jiquan Wang (Zhejiang University), Gang Pan (Zhejiang University)
ClassificationRepresentation LearningConvolutional Neural NetworkTransformerSupervised Fine-TuningTime SeriesBiomedical Data
🎯 What it does: A basic EEG model called CBraMod based on the criss-cross transformer is proposed, which achieves general representation learning of EEG signals through masked reconstruction pre-training;
CEB: Compositional Evaluation Benchmark for Fairness in Large Language Models
Song Wang (University of Virginia), Jundong Li (University of Virginia)
TransformerLarge Language ModelTextBenchmark
🎯 What it does: A multi-dimensional bias assessment benchmark, CEB, has been constructed to unify the evaluation of social biases in LLMs.
Centrality-guided Pre-training for Graph
Bin Liang (Chinese University of Hong Kong), Kam-Fai Wong (Chinese University of Hong Kong)
ClassificationRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: A graph pre-training framework called CenPre is proposed, which enhances node representations through three modules: node-level and graph-level importance learning and representation alignment, thereby improving downstream tasks such as node classification, link prediction, and graph classification.
Century: A Framework and Dataset for Evaluating Historical Contextualisation of Sensitive Images
Canfer Akbulut (Google DeepMind), Lisa Anne Hendricks (Google DeepMind)
Large Language ModelImage
🎯 What it does: A dataset named Century consisting of 1,500 sensitive historical images has been constructed, and a no-reference historical context description evaluation framework has been proposed.
CertainlyUncertain: A Benchmark and Metric for Multimodal Epistemic and Aleatoric Awareness
Khyathi Chandu, Yejin Choi (University of Washington)
RecognitionData SynthesisTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: A benchmark dataset called CERTAINLYUNCERTAIN was proposed to evaluate the visual-language model (VLM) capabilities in recognizing epistemic and aleatoric uncertainties, along with a confidence-based evaluation metric.
Certified Robustness Under Bounded Levenshtein Distance
Elias Abad Rocamora (Ecole Polytechnique Federale de Lausanne), Volkan Cevher (Ecole Polytechnique Federale de Lausanne)
ClassificationComputational EfficiencyAdversarial AttackConvolutional Neural NetworkText
🎯 What it does: A method for proving the robustness of text classifiers based on Lipschitz constants is proposed, which can provide deterministic certification against attacks under Levenshtein distance in a single forward pass.
Certifying Counterfactual Bias in LLMs
Isha Chaudhary (University of Illinois at Urbana-Champaign), Gagandeep Singh (University of Illinois at Urbana-Champaign)
GenerationTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: Proposed and implemented the LLMCert-B framework, which can quantitatively certify counterfactual biases in large-scale LLM generated texts in a black-box manner and provide high-confidence unbiased probability intervals.
Certifying Language Model Robustness with Fuzzed Randomized Smoothing: An Efficient Defense Against Backdoor Attacks
Bowei He (City University of Hong Kong), Chen Ma (City University of Hong Kong)
TransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper studies a method that combines Fuzzy Random Smoothing (FRS) for the verifiable robustness defense of pre-trained language models (PLMs) with backdoors implanted during the pre-training phase.
CFD: Learning Generalized Molecular Representation via Concept-Enhanced Feedback Disentanglement
Aming WU, Cheng Deng (Xidian University)
Domain AdaptationRepresentation LearningDrug DiscoveryGraph Neural NetworkAuto EncoderContrastive LearningGraph
🎯 What it does: Designed and implemented a Conceptual Feedback Decoupling (CFD) method for learning robust universal molecular representations against distribution shifts.
CFG++: Manifold-constrained Classifier Free Guidance for Diffusion Models
Hyungjin Chung (Korea Advanced Institute of Science and Technology), Jong Chul Ye (Korea Advanced Institute of Science and Technology)
RestorationGenerationData SynthesisDiffusion modelScore-based ModelImage
🎯 What it does: This paper proposes CFG++, an improved Classifier-Free Guidance (CFG) method that utilizes uncertain noise for interpolation rather than extrapolation during the denoising step, resulting in smoother and higher-quality generation and reverse sampling results in text-guided diffusion models.
CG-Bench: Clue-grounded Question Answering Benchmark for Long Video Understanding
Guo Chen (Nanjing University), Limin Wang (Nanjing University)
TransformerLarge Language ModelVision Language ModelVideoMultimodalityBenchmark
🎯 What it does: This paper proposes CG-Bench, a clue-based question answering evaluation benchmark for long videos, covering 1,219 manually annotated videos and 12,129 QAC (triples).
Chain-of-Action: Faithful and Multimodal Question Answering through Large Language Models
Zhenyu Pan (Northwestern University), Han Liu (Northwestern University)
TransformerLarge Language ModelPrompt EngineeringTextMultimodalityTabularRetrieval-Augmented Generation
🎯 What it does: Proposes the Chain-of-Action (CoA) framework, which parallelizes multimodal retrieval and reasoning to address the hallucination and reasoning weaknesses of large language models.
Chain-of-Focus Prompting: Leveraging Sequential Visual Cues to Prompt Large Autoregressive Vision Models
Jiyang Zheng (University of Sydney), Tongliang Liu (University of Sydney)
Object DetectionSegmentationPose EstimationTransformerPrompt EngineeringImage
🎯 What it does: Proposes the Chain-of-Focus (CoF) prompting method for achieving visual step-by-step reasoning and adaptive prompting in large autoregressive visual models (LAVMs);
Chain-of-region: Visual Language Models Need Details for Diagram Analysis
Xue Li (University of California), Haifeng Chen (NEC Laboratories America)
RecognitionSegmentationTransformerPrompt EngineeringVision Language ModelImage
🎯 What it does: Through traditional computer vision methods (OpenCV), the initialization, splitting, and merging of scientific charts are performed to collect visual details and inject them into VLM, achieving fine-grained analysis of scientific charts.
Chain-of-Thought Provably Enables Learning the (Otherwise) Unlearnable
Chenxiao Yang (Toyota Technological Institute at Chicago), David Wipf (Amazon Web Services)
TransformerLarge Language ModelTextChain-of-Thought
🎯 What it does: This study investigates the theory and practice of Chain-of-Thought (CoT) in enhancing learning effectiveness through task decomposition in the context of language models.
CHAMP: Conformalized 3D Human Multi-Hypothesis Pose Estimators
Harry Zhang (Massachusetts Institute of Technology), Luca Carlone (Massachusetts Institute of Technology)
Pose EstimationDiffusion modelVideo
🎯 What it does: The CHAMP method is proposed, which utilizes diffusion models to generate multiple hypotheses of 3D human poses, and filters and aggregates these hypotheses through an end-to-end differentiable conformity scorer to obtain the final prediction.
Charting the Design Space of Neural Graph Representations for Subgraph Matching
Vaibhav Raj (Indian Institute of Technology Bombay), Abir De (Indian Institute of Technology Bombay)
Graph Neural NetworkGraph
🎯 What it does: Systematically explore the design space of neural subgraph matching models and propose an optimal configuration scheme.
ChartMimic: Evaluating LMM's Cross-Modal Reasoning Capability via Chart-to-Code Generation
Cheng Yang (Tsinghua University), Yujiu Yang (Tsinghua University)
GenerationAI Code AssistantTransformerLarge Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Created the ChartMimic benchmark to evaluate the cross-modal reasoning ability of large multimodal models (LMM) from chart visual understanding to code generation;
ChartMoE: Mixture of Diversely Aligned Expert Connector for Chart Understanding
Zhengzhuo Xu (International Digital Economy Academy), Jian Guo (Peking University)
TransformerLarge Language ModelSupervised Fine-TuningMixture of ExpertsVision Language ModelMultimodalityTabular
🎯 What it does: This paper proposes ChartMoE, a multimodal large language model that uses Mixture of Experts (MoE) as a visual-language connector to enhance chart understanding and reasoning capabilities.
CHASE-SQL: Multi-Path Reasoning and Preference Optimized Candidate Selection in Text-to-SQL
Mohammadreza Pourreza (Google Cloud), Sercan O Arik
Large Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: This paper proposes the CHASE-SQL framework, which enhances the execution accuracy of Text-to-SQL through multi-path candidate generation, query repair, and comparative selection agents.
ChatQA 2: Bridging the Gap to Proprietary LLMs in Long Context and RAG Capabilities
Peng Xu (NVIDIA), Bryan Catanzaro (NVIDIA)
GenerationRetrievalTransformerLarge Language ModelSupervised Fine-TuningTextRetrieval-Augmented Generation
🎯 What it does: Developed ChatQA 2 based on Llama 3.0, supporting a 128K context window, while balancing long text understanding and retrieval-augmented generation (RAG) capabilities;
CheapNet: Cross-attention on Hierarchical representations for Efficient protein-ligand binding Affinity Prediction
Hyukjun Lim (Seoul National University), Sangseon Lee (Inha University)
Drug DiscoveryGraph Neural NetworkBiomedical Data
🎯 What it does: We propose CheapNet, a protein-ligand binding affinity prediction model that integrates atomic-level embeddings with hierarchical representations after differentiable clustering aggregation through cross-attention fusion.
Cheating Automatic LLM Benchmarks: Null Models Achieve High Win Rates
Xiaosen Zheng (Sea AI Lab), Min Lin (Sea AI Lab)
TransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: The study demonstrates how to use the 'empty model' to deceive automated LLM evaluation benchmarks through structured fake responses, resulting in a high win rate.
ChemAgent: Self-updating Memories in Large Language Models Improves Chemical Reasoning
Xiangru Tang (Yale University), Mark Gerstein (Yale University)
Drug DiscoveryTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: A self-updating chemical reasoning library (ChemAgent) has been constructed, which breaks down complex problems into subtasks and stores the subtasks and their solutions in multiple types of memory (planning, execution, knowledge), achieving continuous learning and improvement of the LLM.
Chemistry-Inspired Diffusion with Non-Differentiable Guidance
Yuchen Shen (Carnegie Mellon University), Barnabas Poczos
OptimizationDrug DiscoveryDiffusion modelGraph
🎯 What it does: Using a non-differentiable quantum chemistry oracle (such as GFN2-xTB) to guide a 3D molecular diffusion model through zero-order gradient estimation, achieving conditional molecular generation and enhancing molecular stability.
CHiP: Cross-modal Hierarchical Direct Preference Optimization for Multimodal LLMs
Jinlan Fu (National University of Singapore), See-Kiong Ng (National University of Singapore)
OptimizationReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningTextMultimodality
🎯 What it does: Optimizes the hallucination problem of multimodal large language models and proposes a cross-modal hierarchical direct preference optimization method called CHiP.
ChroKnowledge: Unveiling Chronological Knowledge of Language Models in Multiple Domains
Yein Park (Korea University), Jaewoo Kang (Korea University)
TransformerLarge Language ModelPrompt EngineeringTextTime SeriesBiomedical DataBenchmark
🎯 What it does: Proposes the CHROKNOWBENCH and CHROKNOWLEDGE frameworks to evaluate the knowledge retention of large language models across different domains over time, and enhances the accuracy of memory for invariant objects through time series reasoning using CHROKNOWPROMPT.
Chunk-Distilled Language Modeling
Yanhong Li (University of Chicago), Jiawei Zhou (Stony Brook University)
GenerationDomain AdaptationKnowledge DistillationTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: The Chunk-Distilled Language Modeling (CD-LM) method is proposed, which efficiently generates multiple tokens and injects new knowledge without additional training by retrieving multi-word chunks during the generation process and making chunk-level acceptance/rejection decisions.
CipherPrune: Efficient and Scalable Private Transformer Inference
Yancheng Zhang (University of Central Florida), Qian Lou (University of Central Florida)
Safty and PrivacyComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: The CipherPrune framework is proposed, achieving adaptive layer-wise token pruning and multi-order polynomial approximation for encrypted inputs, significantly improving the inference efficiency of private Transformers.
Circuit Representation Learning with Masked Gate Modeling and Verilog-AIG Alignment
Haoyuan WU, Bei Yu (Chinese University of Hong Kong)
Representation LearningGraph Neural NetworkLarge Language ModelTextGraph
🎯 What it does: This paper proposes a constrained mask modeling paradigm MGVGA for circuit representation learning, which can learn fine-grained structural information and abstract functional information without compromising logical equivalence.
Circuit Transformer: A Transformer That Preserves Logical Equivalence
Xihan Li (University College London), Jun Wang (Huawei)
GenerationOptimizationTransformerGraph
🎯 What it does: This study investigates how to ensure the logical equivalence of logic circuit implementations under generative neural networks, proposing the Circuit Transformer to generate strictly equivalent and more compact circuits.
CircuitFusion: Multimodal Circuit Representation Learning for Agile Chip Design
Wenji Fang (Hong Kong University of Science and Technology), Zhiyao Xie (Hong Kong University of Science and Technology)
OptimizationRepresentation LearningGraph Neural NetworkTransformerContrastive LearningMultimodalityGraphRetrieval-Augmented Generation
🎯 What it does: This paper presents CircuitFusion, a multimodal, implementation-aware circuit encoder that first partitions RTL circuits into subcircuits and encodes them using three modalities: HDL, graph structure, and functional summaries. It then learns general representations through a pre-trained self-supervised task and fine-tunes for design quality prediction tasks at the RTL stage; at the same time, it introduces retrieval-augmented reasoning to achieve zero-shot predictions.
CirT: Global Subseasonal-to-Seasonal Forecasting with Geometry-inspired Transformer
Yang Liu (Hong Kong University of Science and Technology), Jia Li (Hong Kong University of Science and Technology)
TransformerTime Series
🎯 What it does: A direct prediction S2S climate forecasting model named CirT has been designed and trained, capable of outputting multivariate averages over intervals of 2–6 weeks.
CityAnchor: City-scale 3D Visual Grounding with Multi-modality LLMs
Jinpeng Li (Wuhan University), Bisheng Yang (Wuhan University)
RecognitionObject DetectionTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodalityPoint Cloud
🎯 What it does: This paper proposes CityAnchor, a two-stage urban-scale 3D visual localization method based on a multimodal large language model. It first performs coarse localization of candidate areas on a 2D projection map, and then conducts fine-grained matching of text and 3D point clouds within the candidate areas to ultimately locate the target object.
CityGaussianV2: Efficient and Geometrically Accurate Reconstruction for Large-Scale Scenes
Yang Liu (Institute of Automation, Chinese Academy of Sciences), Zhaoxiang Zhang (Institute of Automation, Chinese Academy of Sciences)
RestorationCompressionComputational EfficiencyGaussian SplattingPoint Cloud
🎯 What it does: This paper proposes CityGaussianV2, which utilizes 2D Gaussian splatting to achieve high-precision, editable geometric reconstruction in large-scale scenes, and enables efficient deployment through parallel training and compression.
CL-DiffPhyCon: Closed-loop Diffusion Control of Complex Physical Systems
Long Wei (Westlake University), Tailin Wu (Westlake University)
Reinforcement LearningDiffusion modelTime SeriesPhysics Related
🎯 What it does: To address the closed-loop control problem of complex physical systems, the CL-DiffPhyCon method is proposed, which generates control sequences under real-time feedback through an asynchronous diffusion model, significantly reducing sampling costs.
CL-MFAP: A Contrastive Learning-Based Multimodal Foundation Model for Molecular Property Prediction and Antibiotic Screening
Gen Zhou (Western University), Pingzhao Hu (Western University)
Representation LearningDrug DiscoveryGraph Neural NetworkTransformerContrastive LearningMultimodalityGraphBiomedical Data
🎯 What it does: This paper proposes the CL-MFAP model, which utilizes contrastive learning pre-training with three modalities: SMILES, molecular graphs, and Morgan fingerprints, followed by fine-tuning on downstream tasks such as antibacterial activity.
Class Distribution-induced Attention Map for Open-vocabulary Semantic Segmentations
Dong Un Kang (Seoul National University), Se Young Chun (Seoul National University)
SegmentationTransformerVision Language ModelContrastive LearningImage
🎯 What it does: A training-free method called CDAM (Class Distribution-induced Attention Map) is proposed, which constructs attention maps by utilizing the class distribution similarity output by CLIP, thereby improving the localization accuracy of open vocabulary semantic segmentation without the need for additional training or pixel-level annotations.
ClassDiffusion: More Aligned Personalization Tuning with Explicit Class Guidance
Jiannan Huang (Beijing Jiaotong University), Yunchao Wei (Beijing Jiaotong University)
GenerationData SynthesisTransformerDiffusion modelImageVideo
🎯 What it does: This paper proposes ClassDiffusion, which utilizes semantic preservation loss to recover the combinatorial capabilities lost after personalized fine-tuning, thereby enhancing multi-condition text generation performance.
Classic but Everlasting: Traditional Gradient-Based Algorithms Converge Fast Even in Time-Varying Multi-Player Games
Yanzheng Chen (University of Science and Technology of China), Jun Yu (University of Science and Technology of China)
Optimization
🎯 What it does: This study investigates the convergence of the last iteration of the Extra Gradient (EG) and Optimistic Gradient (OG) algorithms in time-varying, monotonic multi-player games, providing a convergence rate of O(1/√T), covering bounded closed sets as well as partially unbounded cases.
ClawMachine: Learning to Fetch Visual Tokens for Referential Comprehension
Tianren Ma (University of Chinese Academy of Sciences), Qixiang Ye (University of Chinese Academy of Sciences)
RecognitionObject DetectionSegmentationTransformerLarge Language ModelVision Language ModelImageTextMultimodality
🎯 What it does: ClawMachine is proposed, a multimodal large language model that directly uses a set of visual tokens for reference understanding and localization without requiring additional syntax.
CLDyB: Towards Dynamic Benchmarking for Continual Learning with Pre-trained Models
Shengzhuang Chen (City University of Hong Kong), Ying Wei (Zhejiang University)
ClassificationTransformerReinforcement LearningImageBenchmark
🎯 What it does: The CLDyB framework is proposed, which dynamically generates task sequences for pre-trained model-based continual learning methods using Markov decision processes and Monte Carlo tree search, providing a unified and repeatable challenging benchmark for evaluating various CL methods.
CLIBD: Bridging Vision and Genomics for Biodiversity Monitoring at Scale
ZeMing Gong (Simon Fraser University), Angel X Chang
ClassificationRetrievalTransformerContrastive LearningImageTextMultimodality
🎯 What it does: A multimodal model CLIBD was constructed, embedding insect images, DNA barcodes, and text classification labels into the same shared space, achieving zero-shot classification and retrieval through alignment.
ClimaQA: An Automated Evaluation Framework for Climate Question Answering Models
Veeramakali Vignesh Manivannan (University of California San Diego), Taylor Berg-Kirkpatrick (University of California San Diego)
Large Language ModelPrompt EngineeringTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: A ClimaQA benchmark has been established in the field of climate science to evaluate the question-answering performance of LLMs.
CLIPDrag: Combining Text-based and Drag-based Instructions for Image Editing
Ziqi Jiang, Long Chen
Diffusion modelImage
🎯 What it does: By combining text descriptions with user-dragged points, the CLIPDrag scheme is proposed to achieve precise and unambiguous image editing.
CLIPure: Purification in Latent Space via CLIP for Adversarially Robust Zero-Shot Classification
Mingkun Zhang (Chinese Academy of Sciences), Xueqi Cheng (Chinese Academy of Sciences)
ClassificationAdversarial AttackDiffusion modelContrastive LearningImageMultimodalityStochastic Differential Equation
🎯 What it does: A zero-shot classification method called CLIPure for adversarial purification in the CLIP multimodal latent space has been developed.
Clique Number Estimation via Differentiable Functions of Adjacency Matrix Permutations
Indradyumna Roy (Indian Institute of Technology Bombay), Abir De (Indian Institute of Technology Bombay)
OptimizationExplainability and InterpretabilityGraph Neural NetworkReinforcement LearningGraph
🎯 What it does: This paper proposes a differentiable maximum clique estimation framework called MXNET, which utilizes soft permutation and a dynamic programming-based information propagation network to directly predict the size of the maximum clique from the graph's adjacency matrix.
CLoSD: Closing the Loop between Simulation and Diffusion for multi-task character control
Guy Tevet (Tel Aviv University), Michiel van de Panne (University of British Columbia)
Robotic IntelligenceTransformerReinforcement LearningDiffusion modelText
🎯 What it does: The CLoSD system is proposed, which implements text-based multi-task role control, combining a real-time diffusion planner (DiP) with a physical simulation reinforcement learning tracking controller to perform interactive actions such as navigation, striking, sitting, and standing in a physical environment.
Closed-Form Merging of Parameter-Efficient Modules for Federated Continual Learning
Riccardo Salami (University of Modena and Reggio Emilia), Simone Calderara (University of Modena and Reggio Emilia)
Federated LearningSupervised Fine-TuningImageText
🎯 What it does: An algorithm called LoRM is proposed for merging LoRA modules using a closed-form solution method in the context of federated incremental learning.
CO-MOT: Boosting End-to-end Transformer-based Multi-Object Tracking via Coopetition Label Assignment and Shadow Sets
Feng yan, Lin Ma (Meituan Inc)
Object DetectionObject TrackingTransformerVideo
🎯 What it does: This paper proposes CO-MOT, which utilizes a combination of CO-opetition Label Assignment and Shadow Set to enhance the synergy between detection and tracking in a Transformer-based end-to-end MOT model, significantly improving issues related to tracking termination and loss.
Co$^{\mathbf{3}}$Gesture: Towards Coherent Concurrent Co-speech 3D Gesture Generation with Interactive Diffusion
Xingqun Qi (Hong Kong University of Science and Technology), Yike Guo (Hong Kong University of Science and Technology)
GenerationData SynthesisPose EstimationTransformerDiffusion modelMultimodalityAudio
🎯 What it does: A Co Gesture 3 framework based on bidirectional collaborative diffusion is proposed to generate temporally coherent and interactive 3D co-speech gestures from two-person dialogue audio.
COAT: Compressing Optimizer states and Activations for Memory-Efficient FP8 Training
Haocheng Xi (University of California), Song Han
CompressionOptimizationComputational EfficiencyTransformerLarge Language ModelVision Language ModelTextMultimodality
🎯 What it does: COAT is proposed, a training framework that quantizes optimizer states and activations to FP8, significantly reducing memory usage and improving speed for large model training.
Cocoon: Robust Multi-Modal Perception with Uncertainty-Aware Sensor Fusion
Minkyoung Cho (University of Michigan), Zhuoqing Mao
Object DetectionAutonomous DrivingTransformerMixture of ExpertsMultimodalityPoint Cloud
🎯 What it does: The Cocoon framework is proposed to enhance accuracy and robustness in 3D object detection through uncertainty-aware cross-modal fusion.
CodeMMLU: A Multi-Task Benchmark for Assessing Code Understanding & Reasoning Capabilities of CodeLLMs
Dung Manh Nguyen, Nghi D. Q. Bui (FPT Software AI Center)
AI Code AssistantTransformerLarge Language ModelPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: A large-scale multiple-choice question benchmark called CodeMMLU is proposed to evaluate the capabilities of CodeLLM in code understanding and reasoning.
CodePlan: Unlocking Reasoning Potential in Large Language Models by Scaling Code-form Planning
Jiaxin Wen (Tsinghua University), Minlie Huang (Tsinghua University)
OptimizationAI Code AssistantTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: The CODEPLAN framework is constructed to improve the multi-step reasoning ability of LLMs by generating and executing plans in the form of code.
CofCA: A STEP-WISE Counterfactual Multi-hop QA benchmark
Jian Wu (Institute of Science Tokyo), Yue Zhang (Westlake University)
Large Language ModelPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: A benchmark named CofCA has been constructed for stepwise causal evaluation of multi-hop question answering, generating adversarial paragraphs and sub-questions through a human-machine loop to assess the true reasoning ability of LLMs.
COFlowNet: Conservative Constraints on Flows Enable High-Quality Candidate Generation
Yudong Zhang (University of Science and Technology of China), Yang Wang (University of Science and Technology of China)
GenerationDrug DiscoveryFlow-based ModelTabular
🎯 What it does: COFlowNet is proposed, a generative flow network for offline environments that can generate diverse and high-scoring candidate substances without relying on online evaluation.
CogCoM: A Visual Language Model with Chain-of-Manipulations Reasoning
Ji Qi (Tsinghua University), Jie Tang (Tsinghua University)
Explainability and InterpretabilityTransformerLarge Language ModelVision Language ModelImageTextMultimodality
🎯 What it does: Proposed and implemented the Chain of Manipulations (CoM) mechanism, enabling visual language models to gradually reason and provide answers through a series of interpretable visual operations such as actively annotating, scaling, counting, and calculating images.
CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer
Zhuoyi Yang (Tsinghua University), Jie Tang (Tsinghua University)
GenerationData SynthesisTransformerDiffusion modelVideoText
🎯 What it does: CogVideoX is a text-to-video generation model based on a diffusion Transformer, capable of generating high-quality videos of up to 10 seconds in length, with coherent actions and semantic consistency at 16fps and a resolution of 768×1360.
CoInD: Enabling Logical Compositions in Diffusion Models
Sachit Gaudi (Michigan State University), Vishnu Boddeti
GenerationData SynthesisDiffusion modelImage
🎯 What it does: This study investigates the generation of arbitrary logical attribute combinations in diffusion models and proposes the COIND objective, which enforces conditional independence to enhance the logical consistency and diversity of samples.
Collab: Controlled Decoding using Mixture of Agents for LLM Alignment
Souradip Chakraborty (JPMorgan AI Research), Sumitra Ganesh (JPMorgan AI Research)
Reinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningAgentic AIText
🎯 What it does: A multi-agent hybrid control decoding method (Collab) is proposed, which achieves alignment with the target task of LLM during the inference phase by dynamically switching between different pre-trained alignment models, without the need for fine-tuning.
CollabEdit: Towards Non-destructive Collaborative Knowledge Editing
Jiamu Zheng (Zhejiang University), Tao Lin (Westlake University)
TransformerLarge Language ModelText
🎯 What it does: The COLLABEDIT framework is proposed, enabling multiple parties to collaboratively edit knowledge in large language models without disclosing editing requests.
Collaborative Discrete-Continuous Black-Box Prompt Learning for Language Models
Hualin Zhang (Mohamed bin Zayed University of Artificial Intelligence), Yi Chang (Jilin University)
OptimizationTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: By jointly optimizing discrete and continuous prompts, a ZO-PoG black-box prompt learning framework is proposed, which can enhance the performance of large language models on downstream tasks without accessing the internal parameters of the model.
Collapsed Language Models Promote Fairness
Jingxuan Xu (Beijing Jiaotong University), Yunchao Wei (Beijing Jiaotong University)
TransformerSupervised Fine-TuningText
🎯 What it does: This paper studies the phenomenon of 'Neural Collapse' in neural networks within language models, finding that debiased models tend to collapse more in the word vectors and token representations of gender-related words. Based on this, it proposes the addition of the (U NC) 3 regularization term during the fine-tuning process, forming a general, low-cost debiasing method; it also makes plug-in improvements to various debiasing strategies (MABEL, ASE, BEC, etc.).
ColPali: Efficient Document Retrieval with Vision Language Models
Manuel Faysse (CentraleSupélec), Pierre Colombo (Equall.ai)
RetrievalTransformerLarge Language ModelVision Language ModelContrastive LearningImageTextBenchmark
🎯 What it does: The paper presents ColPali, an end-to-end retrieval method that directly generates multi-vector embeddings from document page images using a visual language model.
ComaDICE: Offline Cooperative Multi-Agent Reinforcement Learning with Stationary Distribution Shift Regularization
The Viet Bui (Singapore Management University), Tien Anh Mai
Reinforcement LearningTabular
🎯 What it does: An offline multi-agent reinforcement learning algorithm named ComaDICE is proposed, which learns a global policy through static distribution regularization (DICE) in a cooperative environment, and implements local policy training using a centralized training and decentralized execution (CTDE) framework.
Combatting Dimensional Collapse in LLM Pre-Training Data via Submodular File Selection
Ziqing Fan (Shanghai Jiao Tong University), Yanfeng Wang (Shanghai Jiao Tong University)
TransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: A document selection algorithm based on feature decorrelation, DiSF, is proposed to avoid the dimensional collapse problem caused by domain relevance during LLM pre-training.
Combining Induction and Transduction for Abstract Reasoning
Wen-Ding Li (Cornell University), Kevin Ellis (Cornell University)
TransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: The study investigates two few-shot learning methods using neural networks for induction (program synthesis) and recursion (direct prediction) on the ARC task, comparing and integrating them to demonstrate their complementarity in solving different sub-tasks.
COMBO: Compositional World Models for Embodied Multi-Agent Cooperation
Hongxin Zhang (University of Massachusetts), Chuang Gan (Honda Research Institute USA)
Robotic IntelligenceReinforcement Learning from Human FeedbackTransformerVision Language ModelDiffusion modelWorld ModelVideoPoint Cloud
🎯 What it does: The study focuses on multi-agent collaboration and proposes the COMBO framework, which utilizes composable world models and Vision-Language Models for tree search planning.
COME: Test-time Adaption by Conservatively Minimizing Entropy
Qingyang Zhang (Tianjin University), Changqing Zhang (Tianjin University)
Domain AdaptationImage
🎯 What it does: A conservative entropy minimization (COME) method based on subjective logic is proposed to address the model collapse issue caused by overconfidence in traditional entropy minimization during test-time adaptation (TTA).
ComLoRA: A Competitive Learning Approach for Enhancing LoRA
Qiushi Huang (Southern University of Science and Technology), Yu Zhang (Southern University of Science and Technology)
Large Language ModelSupervised Fine-TuningMixture of ExpertsText
🎯 What it does: Combine competitive learning to train multiple LoRA modules, and only use the optimal LoRA during the inference phase to enhance the expressiveness and performance of LoRA.
Commit0: Library Generation from Scratch
Wenting Zhao (Cornell University), Alexander M Rush (Cornell University)
AI Code AssistantLarge Language ModelTextBenchmark
🎯 What it does: Proposes the COMMIT0 benchmark, which requires AI to implement a complete Python library from scratch using natural language specifications and unit tests.
CoMotion: Concurrent Multi-person 3D Motion
Alejandro Newell, Vladlen Koltun
Object TrackingPose EstimationConvolutional Neural NetworkRecurrent Neural NetworkTransformerImageVideo
🎯 What it does: CoMotion is a real-time online multi-person 3D human pose tracking system that utilizes monocular video for end-to-end inference from detection to trajectory updates.
Comparing noisy neural population dynamics using optimal transport distances
Amin Nejatbakhsh (Flatiron Institute), David Lipshutz (Baylor College of Medicine)
Diffusion modelTime Series
🎯 What it does: This paper proposes a new metric method - Causal Optimal Transport (Causal OT) distance, for comparing neural system trajectories with noise and dynamic characteristics.
Comparing Targeting Strategies for Maximizing Social Welfare with Limited Resources
Vibhhu Sharma (Carnegie Mellon University), Bryan Wilder (Carnegie Mellon University)
Tabular
🎯 What it does: Compared the performance of risk-based and treatment effect-based intervention allocation strategies in multi-domain real randomized controlled trials, and explored the impact of confounding levels and welfare functions on their effectiveness.
ComPC: Completing a 3D Point Cloud with 2D Diffusion Priors
Tianxin Huang (National University of Singapore), Gim Hee Lee (National University of Singapore)
RestorationGenerationDiffusion modelGaussian SplattingPoint Cloud
🎯 What it does: A training-free point cloud completion framework based on 3D Gaussian splatting and 2D diffusion models is proposed, capable of completing unseen categories of partial point clouds.
Competing Large Language Models in Multi-Agent Gaming Environments
Jen-tse Huang (Chinese University of Hong Kong), Michael Lyu
TransformerLarge Language ModelPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: This paper presents γ-Bench, a multi-player, multi-round, multi-action game evaluation framework designed to quantify the performance of large language models in complex decision-making scenarios.
Competition Dynamics Shape Algorithmic Phases of In-Context Learning
Core Francisco Park (Harvard University), Hidenori Tanaka (Harvard University)
GenerationData SynthesisTransformerLarge Language ModelSequential
🎯 What it does: This paper proposes a sequence generation task using a finite Markov mixture model to unify the study of the In-Context Learning (ICL) phenomenon in large language models.
Competitive Fair Scheduling with Predictions
Tianming Zhao (University of Sydney), Albert Zomaya
OptimizationReinforcement LearningTabular
🎯 What it does: A learning-enhanced online non-preemptive scheduling algorithm is proposed, aimed at minimizing max-stretch, utilizing predictions of job sizes to optimize scheduling decisions.
Complementary Label Learning with Positive Label Guessing and Negative Label Enhancement
Yuhang Li (Southeast University), Yuheng Jia (Southeast University)
ClassificationImage
🎯 What it does: A novel complementary label learning framework PLNL is proposed, which infers complete labels through two steps: Positive Label Guessing (PLG) and Negative Label Enhancement (NLE), training multi-classifiers from weakly supervised complementary labels.
Complexity Lower Bounds of Adaptive Gradient Algorithms for Non-convex Stochastic Optimization under Relaxed Smoothness
Michael Crawshaw (George Mason University), Mingrui Liu (George Mason University)
Optimization
🎯 What it does: This paper studies the lower bound of the iteration complexity of adaptive gradient algorithms such as AdaGrad under relaxed smoothness conditions ((L, L0, L1)-smooth) in non-convex stochastic optimization, proving that it at least exhibits a quadratic polynomial dependence on the problem parameters.
Composable Interventions for Language Models
Arinbjörn Kolbeinsson (University of Virginia), Thomas Hartvigsen
CompressionKnowledge DistillationTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This study investigates the combinability of various interventions when testing language models, proposing evaluation metrics (Order-free Error, Order Sensitivity) and unified code, and experimentally assesses the combined effects of knowledge editing, compression, and zero-shot learning.
Composing Unbalanced Flows for Flexible Docking and Relaxation
Gabriele Corso (Massachusetts Institute of Technology), Andreas Krause (ETH Zurich)
Protein Structure PredictionFlow-based ModelBiomedical Data
🎯 What it does: A general flow matching framework called Unbalanced Flow Matching (UFM) is proposed to address the issues of protein flexibility and non-physical pose generation in molecular docking, and a flexible docking method FLEXDOCK is constructed based on this framework.
Compositional 4D Dynamic Scenes Understanding with Physics Priors for Video Question Answering
Xingrui Wang (Johns Hopkins University), Alan Yuille (Johns Hopkins University)
Object DetectionObject TrackingData SynthesisConvolutional Neural NetworkRecurrent Neural NetworkReinforcement LearningVision Language ModelVideoPhysics Related
🎯 What it does: Proposed the DynSuperCLEVR dataset and the NS-4DPhysics model for understanding and reasoning about the four-dimensional dynamic properties of 3D objects, such as speed, acceleration, and collisions, in video question answering.
Compositional Entailment Learning for Hyperbolic Vision-Language Models
Avik Pal (University of Amsterdam), Pascal Mettes (University of Amsterdam)
Object DetectionRetrievalRepresentation LearningTransformerVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: A new framework for visual-text representation learning in hyperbolic space, HyCoCLIP, is proposed by introducing the hierarchical relationship between image local boxes and corresponding text segments, using hierarchical contrastive loss and hierarchical entailment loss for joint training.
Compositional simulation-based inference for time series
Manuel Gloeckler (University of Tübingen), Jakob H. Macke (University of Tübingen)
OptimizationComputational EfficiencyScore-based ModelTime SeriesSequentialStochastic Differential Equation
🎯 What it does: This paper proposes an efficient inference method for simulating full-time series parameters by utilizing local state transition information from a Markov simulator, combining local posteriors/likelihoods of single-step transitions.
Computational Explorations of Total Variation Distance
Arnab Bhattacharyya (University of Warwick), N. V. Vinodchandran (University of Nebraska-Lincoln)
🎯 What it does: The study investigates the computation of total variation distance, proposing a simple deterministic polynomial-time algorithm to check the equivalence of product distribution mixtures, and proving that under certain conditions, it is not possible to effectively estimate the total variation distance between arbitrary Ising models.
Computational Limits of Low-Rank Adaptation (LoRA) Fine-Tuning for Transformer Models
Jerry Yao-Chieh Hu (Northwestern University), Han Liu (Northwestern University)
OptimizationComputational EfficiencyTransformerSupervised Fine-Tuning
🎯 What it does: This study investigates the computational limits of LoRA fine-tuning for Transformers, constructs an approximate gradient algorithm, and proves that it can achieve nearly linear time under specific norm thresholds.
Computationally Efficient RL under Linear Bellman Completeness for Deterministic Dynamics
Runzhe Wu (Cornell University), Wen Sun (Cornell University)
Computational EfficiencyReinforcement Learning
🎯 What it does: A reinforcement learning algorithm is proposed under the condition of linear Bellman completeness, with deterministic dynamics, a large action space, and random initial states and rewards, and a theoretical upper bound on regret is provided.
Compute-Constrained Data Selection
Junjie Yin, Alexander M Rush
OptimizationComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper studies data selection in the context of fine-tuning LLMs under computational constraints, proposing a computation-aware data selection objective and cost model. Systematic experiments and performance modeling are conducted across various model sizes, tasks, and data budgets, indicating that traditional high-cost methods are not computationally optimal under most budgets.
Compute-Optimal LLMs Provably Generalize Better with Scale
Marc Anton Finzi, Andrew Gordon Wilson (New York University)
TransformerLarge Language ModelText
🎯 What it does: This study investigates the generalization error of large-scale language models under optimal computation (Chinchilla) conditions and proposes an empirical Freedman martingale confidence inequality, proving that the generalization error decreases with model size.
Computing Circuits Optimization via Model-Based Circuit Genetic Evolution
Zhihai Wang (University of Science and Technology of China), Feng Wu (University of Science and Technology of China)
OptimizationReinforcement Learning
🎯 What it does: A model-based circuit genetic evolution framework (MUTE) is proposed for efficiently optimizing computational circuits such as multipliers and adders;