arXivSub Start free trial

ICLR 2026 Papers with Code β€” Page 6

International Conference on Learning Representations Β· 2207 papers

Do LLMs Forget What They Should? Evaluating In-Context Forgetting in Large Language Models

Yuli Qian (Peking University), Yutao Xie (Microsoft STC Asia)

CodeLarge Language ModelPrompt EngineeringTextSequentialBenchmark

🎯 What it does: Proposed and implemented the ICF-Bench benchmark for systematically evaluating the selective forgetting (in-context forgetting) capabilities of large language models during reasoning.

Do Not Let Low-Probability Tokens Over-Dominate in RL for LLMs

Zhihe Yang (Chinese University of Hong Kong), Yunjian Xu (Chinese University of Hong Kong)

CodeTransformerLarge Language ModelReinforcement LearningText

🎯 What it does: Identify and mitigate the dominance of low-probability tokens over gradients in RL training, thereby improving the performance of large language models on reasoning tasks.

Do Vision-Language Models Respect Contextual Integrity in Location Disclosure?

Ruixin Yang (Georgia Institute of Technology), Alan Ritter (Georgia Institute of Technology)

CodeSafty and PrivacyLarge Language ModelPrompt EngineeringVision Language ModelImageMultimodalityBenchmarkChain-of-Thought

🎯 What it does: This paper proposes the VLM-GEOPRIVACY benchmark to evaluate whether visual language models maintain contextual integrity when disclosing image geographical locations.

Do We Need All the Synthetic Data? Targeted Image Augmentation via Diffusion Models

Dang Nguyen (University of California Los Angeles), Baharan Mirzasoleiman (University of California Los Angeles)

CodeObject DetectionGenerationData SynthesisConvolutional Neural NetworkTransformerDiffusion modelImage

🎯 What it does: Propose a directional diffusion model data augmentation framework called TADA, which synthesizes augmented samples only for those that are not quickly learned during training;

Doctor-R1: Mastering Clinical Inquiry with Experiential Agentic Reinforcement Learning

Yunghwei Lai (Tsinghua University), Yang Liu (Tsinghua University)

CodeReinforcement Learning from Human FeedbackReinforcement LearningAgentic AITextBiomedical DataBenchmark

🎯 What it does: Proposed and implemented DOCTOR-R1β€”a doctor agent based on experience-driven agent reinforcement learning, capable of performing strategic questioning in multi-round interactions and making accurate diagnoses.

Does FLUX Already Know How to Perform Physically Plausible Image Composition?

Shilin Lu (Nanyang Technological University), Adams Wai-Kin Kong (Nanyang Technological University)

CodeImage HarmonizationGenerationVision Language ModelDiffusion modelImageBenchmark

🎯 What it does: Proposed a training-free image synthesis framework called SHINE, which can seamlessly insert target objects into complex lighting and high-resolution scenes while maintaining background integrity;

Does Higher Interpretability Imply Better Utility? A Pairwise Analysis on Sparse Autoencoders

Xu Wang (University of Hong Kong), Difan Zou (University of Hong Kong)

CodeExplainability and InterpretabilityLarge Language ModelAuto EncoderText

🎯 What it does: Investigate the relationship between interpretability and control performance of sparse autoencoders (SAE) in large language models, systematically evaluating the performance of 90 SAEs across different models, architectures, and sparsity levels; propose selecting efficient control features via βˆ†Token Confidence and compare the control effectiveness of different feature selection methods; further analyze the association between interpretability and control performance after feature selection.

DoFlow: Flow-based Generative Models for Interventional and Counterfactual Forecasting on Time Series

Dongze Wu (Georgia Institute of Technology), Yao Xie (Georgia Institute of Technology)

CodeGenerationAnomaly DetectionRecurrent Neural NetworkFlow-based ModelTime SeriesBiomedical DataOrdinary Differential Equation

🎯 What it does: Propose a causal temporal prediction framework DoFlow based on Continuous Normalizing Flows (CNF), which can perform observation, intervention, and counterfactual prediction on known causal DAGs, and provides explicit trajectory likelihood for anomaly detection.

Doloris: Dual Conditional Diffusion Implicit Bridges with Sparsity Masking Strategy for Unpaired Single-Cell Perturbation Estimation

Changxi Chi (Zhejiang University Westlake University), Stan Z. Li (Hong Kong University of Science and Technology)

CodeDrug DiscoveryDiffusion modelBiomedical Data

🎯 What it does: Propose a new framework called Doloris, which uses a dual-conditional diffusion model and a sparse mask strategy to predict cellular responses on unpaired single-cell perturbation data.

Don’t Pass@k: A Bayesian Framework for Large Language Model Evaluation

Mohsen Hariri (Case Western Reserve University), Vipin Chaudhary (Case Western Reserve University)

CodeLarge Language ModelTextBenchmark

🎯 What it does: Propose a Bayesian framework based on Dirichlet priors to replace traditional evaluation metrics such as Pass@k and avg@N, enabling posterior estimation and confidence interval inference for LLM performance.

Don't Settle Too Early: Self-Reflective Remasking for Diffusion Language Models

Zemin Huang (Zhejiang University), Guo-Jun Qi (MAPLE Lab, Westlake University)

CodeGenerationTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningDiffusion modelText

🎯 What it does: Proposes a self-reflective Remask mechanism that enables diffusion language models to identify and re-mask erroneous words during the generation process, achieving multi-round text correction.

Don't Throw Away Your Pretrained Model

Shangbin Feng (University of Washington), Dong Yu (Tencent AI Seattle Lab)

CodeGenerationComputational EfficiencyTransformerLarge Language ModelMixture of ExpertsText

🎯 What it does: Developed a reasoning strategy called SWITCH GENERATION based on model checkpoint collaboration, using a switcher LM to dynamically select pre-trained, fine-tuned, and aligned models to generate text fragments.

DOPPLER: Dual-Policy Learning for Device Assignment in Asynchronous Dataflow Graphs

Xinyu Yao (Rice University), Chris Jermaine

CodeOptimizationGraph Neural NetworkReinforcement LearningGraph

🎯 What it does: In asynchronous work-conserving (WC) systems, the DOPPLER framework is proposed to address the device allocation problem for dataflow graph (Dataflow Graph) computations, using dual strategies (selection strategy SEL and placement strategy PLC) to learn optimal operation sequences and device allocation schemes through reinforcement learning;

Doubly-Regressing Approach for Subgroup Fairness

Kunwoong Kim (KAIST), Yongdai Kim (Seoul National University)

CodeGenerative Adversarial NetworkTextTabularBenchmark

🎯 What it does: Proposes the concept of subgroup subset fairness and develops the DRAF (Doubly Regressing Adversarial learning for Fairness) algorithm, achieving distributed subgroup fairness and marginal fairness in scenarios with sparse and high-dimensional sensitive attributes using a single discriminator.

Downgrade to Upgrade: Optimizer Simplification Enhances Robustness in LLM Unlearning

Yicheng Lang (Michigan State University), Sijia Liu (Michigan State University)

CodeOptimizationSafty and PrivacyLarge Language ModelTextBenchmark

🎯 What it does: Investigate the impact of optimizer level on the robustness of unlearning in large language models, and propose a hybrid FO-ZO optimization strategy to enhance robustness

Doxing via the Lens: Revealing Location-related Privacy Leakage on Multi-modal Large Reasoning Models

Weidi Luo (University of Georgia), Chaowei Xiao (John Hopkins University)

CodeSafty and PrivacyTransformerLarge Language ModelVision Language ModelImageMultimodalityBenchmarkChain-of-Thought

🎯 What it does: This paper studies the risk of geographical location information leakage in images through multimodal large inference models.

DPad: Efficient Diffusion Language Models with Suffix Dropout

Xinhua Chen (Duke University), Yiran Chen (Duke University)

CodeComputational EfficiencyAI Code AssistantTransformerLarge Language ModelDiffusion modelText

🎯 What it does: Designed and implemented a training-free inference strategy called DPad, which significantly reduces the computational cost of dLLM's suffix attention by pre-trimming suffix tokens using a sliding window and distance-decay dropout, thereby improving inference efficiency.

dParallel: Learnable Parallel Decoding for dLLMs

Zigeng Chen (National University of Singapore), Xinchao Wang (National University of Singapore)

CodeComputational EfficiencyKnowledge DistillationTransformerSupervised Fine-TuningText

🎯 What it does: Propose the dParallel method, achieving highly parallel decoding for dLLM through certainty-forcing distillation, significantly reducing decoding steps;

DR-SAC: Distributionally Robust Soft Actor-Critic for Reinforcement Learning under Uncertainty

Mingxuan Cui (New York University), Zhengyuan Zhou (New York University)

CodeReinforcement LearningAuto Encoder

🎯 What it does: Developed DR-SAC, an offline distributed robust soft actor-critic algorithm for continuous action spaces;

Dr.LLM: Dynamic Layer Routing in LLMs

Ahmed Heakl (Parameter Lab), Seong Joon Oh (Parameter Lab)

CodeComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningText

🎯 What it does: Designed and implemented Dr.LLM, a dynamic layer routing framework that can be attached to frozen LLMs, enabling the model to decide to skip, execute, or repeat individual layers based on input, thereby reducing computational cost while maintaining or improving accuracy.

Draft-based Approximate Inference for LLMs

Kevin Galim (FuriosaAI), Kangwook Lee (University of Wisconsin-Madison)

CodeComputational EfficiencyKnowledge DistillationTransformerLarge Language ModelText

🎯 What it does: Proposed an approximate inference framework based on a draft model, utilizing lookahead to predict future outputs, thereby enhancing the accuracy of KV cache trimming and prompt compression, and further introduced SpecKV, SpecPC, and their cascaded solution SpecKV-PC.

DragFlow: Unleashing DiT Priors with Region-Based Supervision for Drag Editing

Zihan Zhou (Nanyang Technological University), Adams Wai-Kin Kong (Nanyang Technological University)

CodeImage HarmonizationGenerationTransformerLarge Language ModelDiffusion modelFlow-based ModelImageBenchmark

🎯 What it does: Proposes the DragFlow framework, enabling drag-based image editing using Diffusion Transformer (e.g., FLUX), combined with region-level supervision, hard constraint background preservation, and adapter-enhanced inversion;

Dragging with Geometry: From Pixels to Geometry-Guided Image Editing

Xinyu Pu (Southeast University), Pan Zhou (Singapore Management University)

CodeDepth EstimationDiffusion modelImageBenchmark

🎯 What it does: Proposes GeoDrag, a one-step drag-and-drop image editing framework that integrates 3D geometric information with 2D pixel plane information.

Draw-In-Mind: Rebalancing Designer-Painter Roles in Unified Multimodal Models Benefits Image Editing

Ziyun Zeng (Show Lab, National University of Singapore), Mike Zheng Shou (Show Lab, National University of Singapore)

CodeGenerationTransformerVision Language ModelFlow-based ModelImageTextMultimodalityChain-of-Thought

🎯 What it does: Proposed and trained the Draw-In-Mind (DIM) dataset and model, emphasizing shifting design responsibility to the understanding module within a unified multimodal model to enhance image editing performance.

DRBench: A Realistic Benchmark for Enterprise Deep Research

Amirhossein Abaskohi (ServiceNow Research), Issam H. Laradji (ServiceNow Research)

CodeTransformerLarge Language ModelAgentic AIMultimodalityBenchmarkRetrieval-Augmented GenerationChain-of-Thought

🎯 What it does: Proposes DRBench, a benchmark for enterprise deep research tasks, evaluating AI agents using real-world enterprise environments and multimodal data;

DreamOn: Diffusion Language Models For Code Infilling Beyond Fixed-size Canvas

Zirui Wu (University of Hong Kong), Lingpeng Kong (University of Hong Kong)

CodeGenerationAI Code AssistantSupervised Fine-TuningDiffusion modelText

🎯 What it does: Propose the DREAMON framework, which introduces [expand] and [delete] special states into diffusion language models to achieve dynamic length generation, solving the fixed-length limitation.

DRIFT: Decompose, Retrieve, Illustrate, then Formalize Theorems

Meiru Zhang (University of Cambridge), Gerasimos Lampouras (Huawei)

CodeAI Code AssistantTransformerLarge Language ModelTextBenchmarkRetrieval-Augmented Generation

🎯 What it does: Proposed the DRIFT framework, addressing the dual challenges of knowledge and syntax in automatic formalization by decomposing natural language mathematical statements into sub-queries, retrieving dependencies, instantiating theorems, and ultimately achieving automatic formalization.

DRIFT: Learning from Abundant User Dissatisfaction in Real-World Preference Learning

Yifan Wang (Purdue University), Qingkai Zeng (Nankai University)

CodeData-Centric LearningReinforcement Learning from Human FeedbackContrastive LearningText

🎯 What it does: Proposed an iterative preference learning framework called DRIFT, based on real user dissatisfaction feedback, using dissatisfied (DSAT) examples as negative samples and dynamically sampling positive samples for training.

DriftLite: Lightweight Drift Control for Inference-Time Scaling of Diffusion Models

Yinuo Ren (Stanford University), Jiequn Han (Flatiron Institute)

CodeGenerationComputational EfficiencyDiffusion modelBiomedical DataPhysics RelatedStochastic Differential Equation

🎯 What it does: Propose DriftLite, a lightweight, training-free inference-time adaptation method that adjusts pre-trained diffusion models to adapt to new target distributions by dynamically controlling drift during inference.

DRPO: Efficient Reasoning via Decoupled Reward Policy Optimization

Gang Li (Texas A&M University), Tianbao Yang (Texas A&M University)

CodeOptimizationComputational EfficiencyLarge Language ModelReinforcement LearningText

🎯 What it does: Proposes Decoupled Reward Policy Optimization (DRPO), a reinforcement learning framework that decouples positive and negative sample rewards and normalizes the length reward of positive samples, to train large reasoning models (LRMs) for efficient inference, significantly reducing redundant inference paths.

DrVoice: Parallel Speech-Text Voice Conversation Model via Dual-Resolution Speech Representations

Chao-Hong Tan (Alibaba Group), Jieping Ye (Alibaba Group)

CodeGenerationTransformerLarge Language ModelFlow-based ModelTextMultimodalityAudio

🎯 What it does: Built an end-to-end parallel speech-text generation model called DRVOICE, achieving joint autoregressive generation of dual-resolution speech representations and a speech refinement head;

Dual Randomized Smoothing: Beyond Global Noise Variance

Chenhao Sun (ETH ZΓΌrich), Martin Vechev (ETH ZΓΌrich)

CodeClassificationAdversarial AttackConvolutional Neural NetworkTransformerMixture of ExpertsDiffusion modelImage

🎯 What it does: Propose a Dual Random Smoothing (Dual RS) framework that allows using input-dependent noise variance during the random smoothing process, and prove robustness guarantees under local constant variance conditions.

Dual-IPO: Dual-Iterative Preference Optimization for Text-to-Video Generation

Xiaomeng Yang (Shanghai Academy of AI for Science), Hao Li (Fudan University)

CodeGenerationReinforcement LearningVision Language ModelDiffusion modelVideoTextMultimodalityChain-of-Thought

🎯 What it does: Developed the Dual-IPO framework, which significantly improves the quality of text-to-video generation and alignment with human preferences through alternating optimization of the reward model and video generation model.

Dual-Kernel Adapter: Expanding Spatial Horizons for Data-Constrained Medical Image Analysis

Ziquan Zhu (University of Leicester), Tianjin Huang (Mohamed bin Zayed University of Artificial Intelligence)

CodeClassificationSegmentationConvolutional Neural NetworkBiomedical Data

🎯 What it does: Systematic evaluation of adapter fine-tuning in low-data medical imaging tasks, and proposing Dual-Kernel Adapter (DKA) to expand the effective receptive field

Dual-Path Condition Alignment for Diffusion Transformers

Changhao Peng (Peking University), Wei Gao (Peking University)

CodeGenerationTransformerDiffusion modelImage

🎯 What it does: Propose an unsupervised alignment method (DUPA) that does not require an external visual encoder. By performing multiple random noise samplings on the same image and processing them with a Decoupled Diffusion Transformer (DDT), the method enhances the performance of generative models by aligning conditional features obtained through different noise paths.

Dual-Robust Cross-Domain Offline Reinforcement Learning Against Dynamics Shifts

Zhongjian Qiao (City University of Hong Kong), Shuang Qiu (City University of Hong Kong)

CodeDomain AdaptationReinforcement LearningBenchmark

🎯 What it does: Proposed a dual robust method for cross-domain offline reinforcement learning (training and testing) called DROCO, verifying its superior performance in dynamic shift scenarios.

Dual-Scale World Memory for LLM Agents towards Hard-Exploration Problems

Minsoo Kim (Seoul National University), seung-won hwang

CodeTransformerLarge Language ModelReinforcement LearningAgentic AIWorld ModelText

🎯 What it does: Proposed and implemented the GLoW framework, leveraging dual-scale world memory (global trajectory frontier and local advantage reflection) to address the hard exploration challenge faced by LLM agents in text-based games;

DualMap: Enabling Both Cache Affinity and Load Balancing for Distributed LLM Serving

Ying Yuan (Huazhong University of Science and Technology), Zhou Yu (Huawei)

CodeOptimizationComputational EfficiencyTransformerLarge Language ModelText

🎯 What it does: Propose DualMap, a dual-mapping scheduling strategy, to simultaneously achieve KV cache affinity and load balancing in distributed LLM services;

DuPO: Enabling Reliable Self-Verification via Dual Preference Optimization

Shuaijie She (Nanjing University), Yuxuan Wang (ByteDance Seed)

CodeOptimizationReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningContrastive LearningTextBenchmark

🎯 What it does: Proposes the DuPO framework, leveraging bidirectional learning and self-supervised rewards of LLMs for label-free optimization of model performance;

Dyna-Mind: Learning to Simulate from Experience for Better AI Agents

Xiao Yu (Columbia University), Jianfeng Gao (Microsoft Research)

CodeRobotic IntelligenceTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningVision Language ModelImageTextMultimodality

🎯 What it does: Propose the Dyna-Mind two-stage framework, first using RESIM to generate simulated reasoning trajectories based on real-world experiences, then using DYNA-GRPO in online reinforcement learning to integrate intermediate state feedback and enhance the (V)LM agent's simulation and decision-making capabilities.

Dynamic Early Exit in Reasoning Models

Chenxu Yang (Institute of Information Engineering Chinese Academy of Sciences), Weiping Wang (Institute of Information Engineering Chinese Academy of Sciences)

CodeComputational EfficiencyPrompt EngineeringTextBenchmarkChain-of-Thought

🎯 What it does: Proposed a training-free dynamic early exit mechanism called DEER, allowing large-scale inference models to adaptively terminate thinking early and directly provide answers during chain-of-thought (CoT) generation based on their own confidence.

Dynamic Multi-sample Mixup with Gradient Exploration for Open-set Graph Anomaly Detection

Caiyang Yu (Sichuan University of China), Ziyue Qiao (Great Bay University)

CodeAnomaly DetectionGraph Neural NetworkGraph

🎯 What it does: Propose a framework called DEMO for open-set graph anomaly detection, combining multi-sample Mixup, energy gradient adaptive weighting, and pseudo-label generation assisted by historical memory, enabling the detection of unknown anomalies under extremely limited label scenarios;

Dynamic Novel View Synthesis in High Dynamic Range

Kaixuan Zhang (Nanjing University of Science and Technology), Xiatian Zhu (University of Surrey)

CodeGenerationData SynthesisRecurrent Neural NetworkTransformerGaussian SplattingVideo

🎯 What it does: Proposed and implemented the HDR Dynamic Novel View Synthesis (HDR DNVS) task and the HDR-4DGS model, achieving temporally and spatially consistent HDR rendering through Gaussian Splatting combined with dynamic tone mapping.

Dynamic Weight Grafting: Localizing Finetuned Factual Knowledge in Transformers

Todd Nief (University of Chicago), Ari Holtzman (University of Chicago)

CodeExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningText

🎯 What it does: Developed a dynamic weight grafting method to locate how large language models retrieve new relational knowledge after fine-tuning.

Dynamic-dLLM: Dynamic Cache-Budget and Adaptive Parallel Decoding for Training-Free Acceleration of Diffusion LLM

Tianyi Wu (Harbin Institute of Technology), Zhuotao Tian (Harbin Institute of Technology)

CodeComputational EfficiencyTransformerLarge Language ModelDiffusion modelText

🎯 What it does: Proposes a training-free framework called Dynamic-dLLM to accelerate inference for diffusion-based large language models (dLLMs).

DynamicInfer: Runtime-Aware Sparse Offloading for LLMs Inference on a Consumer-Grade GPU

Zhui Zhu (Tsinghua University), Fan Dang (Beijing Jiaotong University)

CodeOptimizationComputational EfficiencyTransformerText

🎯 What it does: Proposed a real-time neuron scheduling and sparse offloading framework called DynamicInfer, enabling large language models to perform efficient inference on consumer-grade GPUs, addressing the issues of low GPU utilization and high latency caused by static hot-cold neuron partitioning.

Dynamics-Predictive Sampling for Active RL Finetuning of Large Reasoning Models

Yixiu Mao (Tsinghua University), Xiangyang Ji (Tsinghua University)

CodeTransformerLarge Language ModelReinforcement LearningPrompt EngineeringText

🎯 What it does: Propose an online prediction method based on dynamic systems (DPS), which uses a hidden Markov model to perform Bayesian inference on the problem-solving state of prompts, enabling efficient selection of informative prompts without expensive rollouts during reinforcement learning fine-tuning of large language models.

Dyslexify: A Mechanistic Defense Against Typographic Attacks in CLIP

Lorenz Hufe (Fraunhofer Heinrich Hertz Institute), Wojciech Samek (Fraunhofer Heinrich Hertz Institute)

CodeExplainability and InterpretabilityComputational EfficiencyAdversarial AttackTransformerVision Language ModelImageBiomedical Data

🎯 What it does: Propose a gradient-agnostic defense method called Dyslexify, which significantly enhances the model's robustness against layout attacks by identifying and ablating attention heads in the CLIP vision encoder that are specialized for processing text.

Early Signs of Steganographic Capabilities in Frontier LLMs

Artur Zolkowski (ML Alignment & Theory Scholars), David Lindner (Google DeepMind)

CodeSafty and PrivacyTransformerPrompt EngineeringTextChain-of-Thought

🎯 What it does: This paper systematically evaluates the capabilities of state-of-the-art large language models in steganography, including the transmission of hidden information and hidden reasoning, and demonstrates their potential risks through case studies.

Earth-Agent: Unlocking the Full Landscape of Earth Observation with Agents

Peilin Feng (Shanghai Artificial Intelligence Laboratory), Weijia Li (Tsinghua University)

CodeTransformerLarge Language ModelAgentic AIImageMultimodalityBenchmark

🎯 What it does: Proposed the Earth-Agent framework, integrating ReAct and MCP protocols with LLM agents, combining 104 specialized tools to support multi-step quantitative reasoning in three EO modes: RGB, spectral, and product;

EAST: Early Action Prediction Sampling Strategy with Token Masking

Iva Sović (University of Zagreb), Marin Orőić (University of Zagreb)

CodeRecognitionTransformerAuto EncoderVideo

🎯 What it does: Propose the EAST framework, which can predict video actions at any observation ratio by training a single model.

Echo: Towards Advanced Audio Comprehension via Audio-Interleaved Reasoning

Daiqing Wu (Chinese Academy of Sciences), Yu ZHOU

CodeTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningMultimodalityRetrieval-Augmented GenerationChain-of-ThoughtAudio

🎯 What it does: Proposed and implemented an audio-interleaved reasoning framework, constructing the Echo large audio language model. It employs a two-stage training approach (SFT + RL) combined with a structured data generation pipeline, aiming to enhance the model's fine-grained understanding and reasoning capabilities for complex audio.

Echoes as Anchors: Probabilistic Costs and Attention Refocusing in LLM Reasoning

Zhuoyuan Hao (Harbin Institute of Technology), Jing Li (Harbin Institute of Technology)

CodeExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextBenchmark

🎯 What it does: Analyzes the phenomenon of large inference models spontaneously repeating prompts at the beginning of generation (Echo of Prompt), and proposes a probability framework based on rejection sampling, an attention reorientation mechanism, and two methods to improve inference performance;

EchoMotion: Unified Human Video and Motion Generation via Dual-Modality Diffusion Transformer

Yuxiao Yang (Tsinghua University), Jieping Ye (Alibaba Group)

CodeGenerationData SynthesisTransformerDiffusion modelVideoMultimodality

🎯 What it does: Propose the EchoMotion framework, which jointly models the joint distribution of videos and human motion to generate more realistic complex human action videos.

EditBench: Evaluating LLM Abilities to Perform Real-World Instructed Code Edits

Wayne Chi (Carnegie Mellon University), Chris Donahue (Carnegie Mellon University)

CodeAI Code AssistantTransformerLarge Language ModelTextBenchmark

🎯 What it does: Developed a benchmark called EditBench, collecting real user code editing instructions and contexts from VS Code, constructing 540 multilingual, multi-programming language, real-world editing tasks, and using it to evaluate the editing capabilities of 40 LLMs.

EditScore: Unlocking Online RL for Image Editing via High-Fidelity Reward Modeling

Xin Luo (University of Science and Technology of China), Zheng Liu (Beijing Academy of Artificial Intelligence)

CodeImage TranslationSupervised Fine-TuningReinforcement LearningVision Language ModelImageBenchmarkChain-of-Thought

🎯 What it does: Proposed an image editing reward model called EditScore and the corresponding benchmark EditReward-Bench, applying them to Best-of-N selection and online reinforcement learning, significantly enhancing the performance of editing models.

EdiVal-Agent: An Object-Centric Framework for Automated, Fine-Grained Evaluation of Multi-Turn Editing

Tianyu Chen (University of Texas at Austin), Mingyuan Zhou (University of Texas at Austin)

CodeObject DetectionAgentic AIPrompt EngineeringVision Language ModelImageTextBenchmark

🎯 What it does: Proposed EdiVal-Agent, an object-oriented automated evaluation framework for fine-grained assessment of multi-round instructive image editing;

EEPO: Exploration-Enhanced Policy Optimization via Sample-Then-Forget

Liang Chen (Chinese University of Hong Kong), Kam-Fai Wong (Chinese University of Hong Kong)

CodeOptimizationLarge Language ModelReinforcement LearningTextBenchmark

🎯 What it does: Under the RLVR framework, EEPO is proposed by introducing a temporary 'forgetting' step between two-stage sampling, forcing the model to skip already-occurred high-probability trajectories in subsequent sampling, thereby enhancing exploration.

Efficient Agent Training for Computer Use

Yanheng He (Shanghai Jiao Tong University), Pengfei Liu (Shanghai Jiao Tong University)

CodeData SynthesisReinforcement Learning from Human FeedbackTransformerLarge Language ModelAgentic AITextSequentialChain-of-Thought

🎯 What it does: Training a computer to act as an agent can achieve performance surpassing Claude 3.7 Sonnet using only 312 human trajectories expanded via AI synthesis.

Efficient algorithms for Incremental Metric Bipartite Matching

Ritesh Seth (IIIT Delhi), Syamantak Das (IIIT Delhi)

CodeOptimizationComputational EfficiencyImageGraphTabular

🎯 What it does: Propose a deterministic incremental algorithm that maintains an approximate minimum cost bipartite matching in any metric space, supporting online point insertions with an amortized time per update of O~(n(1+Ξ΄ logΒ²(1/Ξ΄) log(nΞ”))).

Efficient and Sharp Off-Policy Learning under Unobserved Confounding

Konstantin Hess (LMU Munich), Stefan Feuerriegel (LMU Munich)

CodeReinforcement LearningBiomedical Data

🎯 What it does: Proposed an efficient and sharp personalized off-policy learning method for unobserved confounding scenarios.

Efficient Credal Prediction through Decalibration

Paul Hofman (LMU Munich), Eyke HΓΌllermeier (LMU Munich)

CodeClassificationAnomaly DetectionOptimizationExplainability and InterpretabilityComputational EfficiencyTransformerVision Language ModelImageTextBiomedical Data

🎯 What it does: Propose a post-processing, model-agnostic method that generates feasible probability intervals under a relative likelihood budget by applying controlled offset (decalibration) to the logits of a trained model, thereby constructing a credal set to express the model's epistemic uncertainty.

Efficient Degradation-agnostic Image Restoration via Channel-Wise Functional Decomposition and Manifold Regularization

Bin Ren (Mohamed bin Zayed University of Artificial Intelligence), Nicu Sebe (University of Trento)

CodeRestorationConvolutional Neural NetworkContrastive LearningImage

🎯 What it does: Proposed MIRAGE, an efficient degradation-agnostic image restoration framework;

Efficient Estimation of Kernel Surrogate Models for Task Attribution

Zhenshuo Zhang (Northeastern University), Hongyang R. Zhang (Northeastern University)

CodeExplainability and InterpretabilityComputational EfficiencyImageText

🎯 What it does: Address the task attribution problem by proposing Kernel Surrogate Models (KERNELSM) for efficiently estimating the impact of training tasks on the performance of the target task, and achieving efficient learning without retraining through first-order gradient approximation.

Efficient Message-Passing Transformer for Error Correcting Codes

Seong-Joon Park (POSTECH), Jong-Seon No (Seoul National University)

CodeOptimizationComputational EfficiencyTransformerGraph

🎯 What it does: Designed an efficient message-passing Transformer (EfficientMPT) for error correction in linear block codes.

Efficient Orthogonal Fine-Tuning with Principal Subspace Adaptation

Fei Wu (University of Exeter), Shiqiang Wang (University of Exeter)

CodeComputational EfficiencyRepresentation LearningTransformerSupervised Fine-TuningImageText

🎯 What it does: Proposes the PSOFT method, performing orthogonal fine-tuning within the principal subspace of pre-trained models, balancing semantic preservation, expressiveness, and multidimensional efficiency.

Efficient Prediction of Large Protein Complexes via Subunit-Guided Hierarchical Refinement

Chixiang Lu (University of Hong Kong), Haibo Jiang (University of Hong Kong)

CodeProtein Structure PredictionTransformerBiomedical Data

🎯 What it does: Propose HIERAFOLD, a hierarchical coarse-to-fine pipeline for predicting the structure of large protein complexes, which utilizes PAE-guided subunit segmentation, interface-aware refinement, and confidence-weighted assembly to significantly reduce memory usage and enable prediction of complexes with over 5k tokens.

Efficient Quantization of Mixture-of-Experts with Theoretical Generalization Guarantees

Mohammed Nowaz Rabbani Chowdhury (Rensselaer Polytechnic Institute), Meng Wang (Rensselaer Polytechnic Institute)

CodeComputational EfficiencyTransformerMixture of ExpertsText

🎯 What it does: Proposed an expert-level mixed-precision quantization method based on router norm variations, which can assign different bit-widths to different experts in MoE models.

Efficient Reasoning with Balanced Thinking

Yulin Li (Harbin Institute of Technology), Zhuotao Tian (Harbin Institute of Technology)

CodeComputational EfficiencyTransformerLarge Language ModelTextBenchmark

🎯 What it does: Propose a training-free framework named REBALANCE that dynamically controls overthinking and underthinking in large reasoning models, achieving efficient balanced reasoning.

Efficient Regression-based Training of Normalizing Flows for Boltzmann Generators

Danyal Rehman, Joey Bose

CodeGenerationData SynthesisDrug DiscoveryProtein Structure PredictionFlow-based ModelBiomedical Data

🎯 What it does: Propose REGFLOW, a regression-based training method that achieves high-quality sampling in single-step regularized flows while providing accurate likelihoods;

Efficient Reinforcement Learning by Guiding World Models with Non-Curated Data

Yi Zhao (Aalto University), Joni Pajarinen (Aalto University)

CodeRobotic IntelligenceReinforcement LearningWorld ModelSequentialRetrieval-Augmented Generation

🎯 What it does: Pre-train a task-agnostic world model using unsupervised, mixed-quality, multi-body non-episodic offline data, and during online fine-tuning, reduce distribution shift through experience replay and execution guidance, significantly improving sample efficiency.

Efficient Resource-Constrained Training of Transformers via Subspace Optimization

Le-Trung Nguyen (TΓ©lΓ©com Paris), Van-Tam Nguyen (TΓ©lΓ©com Paris)

CodeComputational EfficiencyTransformerSupervised Fine-TuningImageText

🎯 What it does: Proposed the Weight-Activation Subspace Iteration (WASI) method for efficiently training Transformer models on edge devices, compressing both weights and activations.

Efficient Submodular Maximization for Sums of Concave over Modular Functions

Yang Lv (Beijing University of Technology), Ruiqi Yang (Beijing University of Technology)

CodeOptimizationGraph

🎯 What it does: This paper proposes a method utilizing accelerated approximate projected gradient ascent (AAPGA), combined with continuous relaxation, convex/concave extension, randomization, and pipelined rounding, to solve submodular maximization problems under cardinality, knapsack, and partition matroid constraints for 'additive concave modular' (SCM) functions;

Efficient Zero-shot Inpainting with Decoupled Diffusion Guidance

Badr MOUFAD, Jimmy Olsson (KTH Royal Institute of Technology)

CodeRestorationDiffusion modelImage

🎯 What it does: This paper proposes a zero-shot image inpainting method called DING, which utilizes pre-trained diffusion models to perform fast and high-quality image restoration in the latent space;

Efficient-SAM2: Accelerating SAM2 with Object-Aware Visual Encoding and Memory Retrieval

Jing Zhang (Institute of Automation Chinese Academy of Sciences), Qingyi Gu (Institute of Automation Chinese Academy of Sciences)

CodeSegmentationComputational EfficiencyTransformerVideo

🎯 What it does: Accelerate the post-training of SAM2 by introducing two modules: object-aware sparse window routing (SWR) and sparse memory retrieval (SMR), reducing redundant computations in the image encoder and memory attention.

Egalitarian Gradient Descent: A Simple Approach to Accelerated Grokking

Ali Saheb Pasand (McGill University), Elvis Dohmatob (Concordia University)

CodeOptimizationImage

🎯 What it does: Proposed Egalitarian Gradient Descent (EGD), a simple hyperparameter-free method that balances optimization speed along the principal gradient directions to accelerate the model's grokking process;

EgoDex: Learning Dexterous Manipulation from Large-Scale Egocentric Video

Ryan Hoque (Apple), Jian Zhang (Apple)

CodePose EstimationRobotic IntelligenceTransformerDiffusion modelFlow-based ModelVideoBenchmark

🎯 What it does: Constructed the largest egocentric viewpoint grasping dataset EgoDex, and evaluated imitation learning methods for hand trajectory prediction based on it.

EgoHandICL: Egocentric 3D Hand Reconstruction with In-Context Learning

Binzhu Xie (Chinese University of Hong Kong), Pheng-Ann Heng (Khalifa University)

CodePose EstimationRetrievalTransformerVision Language ModelAuto EncoderImageVideoMultimodality

🎯 What it does: Propose the EgoHandICL framework to achieve adaptive 3D hand reconstruction, leveraging VLM for retrieving example templates, ICL tokenizer for multimodal context tokenization, and MAE-based architecture;

EgoNight: Towards Egocentric Vision Understanding at Night with a Challenging Benchmark

Deheng Zhang (INSAIT, Sofia University St Kliment Ohridski), Danda Pani Paudel (INSAIT, Sofia University St Kliment Ohridski)

CodeRecognitionData SynthesisDepth EstimationRetrievalTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelVideoTextMultimodalityBenchmark

🎯 What it does: This paper proposes the EgoNight benchmark kit, covering nighttime egocentric visual tasks, including VQA, day-night correspondence retrieval, and depth estimation, and generates 3,658 high-quality QA pairs through human-machine collaborative automatic annotation.

EigenBench: A Comparative Behavioral Measure of Value Alignment

Jonathn Chang (Cornell University), Lionel Levine (Cornell University)

CodeExplainability and InterpretabilityLarge Language ModelTextBenchmark

🎯 What it does: Propose the EigenBench method, which quantifies the average alignment of language models with a given constitution (value system) using model peer review and the EigenTrust algorithm.

Einstein Fields: A Neural Perspective To Computational General Relativity

Sandeep Suresh Cranganore (JKU Linz), Johannes Brandstetter (JKU Linz)

CodeMeshPhysics Related

🎯 What it does: Neurally compress the metric tensor in 4D general relativity simulations, constructing the EinFields model, which can reconstruct the complete spacetime geometry and perform differentiation with extremely low storage requirements.

EIP: Weighted Ranking of LLMs by Quantifying Question Difficulty

Xingjian Hu (Lehigh University), Lichao Sun (Lehigh University)

CodeComputational EfficiencyLarge Language ModelTextBenchmark

🎯 What it does: Proposed and validated the Empirical Interaction Propagation (EIP) framework for simultaneously estimating problem difficulty and model capability in large language model assessments;

Eliciting Numerical Predictive Distributions of LLMs Without Auto-Regression

Julianna Piskorz (University of Cambridge), Mihaela van der Schaar (University of Cambridge)

CodeExplainability and InterpretabilityComputational EfficiencyRepresentation LearningTransformerLarge Language ModelTime Series

🎯 What it does: Investigate how to recover numerical prediction point estimates (mean, median, greedy value) and their uncertainty distributions directly from hidden states by probing the internal representations of large language models (LLMs), without requiring autoregressive sampling.

Eliminating Inductive Bias in Reward Models with Information-Theoretic Guidance

Zhuo Li (Alibaba), guanjunjiang

CodeReinforcement Learning from Human FeedbackTransformerReinforcement LearningText

🎯 What it does: Studied a reward model debiasing method called DIR, aiming to eliminate inductive bias caused by low-quality human preference data in RLHF training (e.g., response length, flattery tone, formatting, etc.).

ELViS: Efficient Visual Similarity from Local Descriptors that Generalizes Across Domains

Pavel Suma (Czech Technical University in Prague), Giorgos Tolias (Czech Technical University in Prague)

CodeRetrievalDomain AdaptationTransformerImage

🎯 What it does: Propose ELViS, a lightweight image-to-image similarity model based on a local feature similarity matrix, for re-ranking in retrieval tasks.

Embedding-Based Context-Aware Reranker

Ye Yuan (McGill University), Siqi Liu (RBC Borealis)

CodeRetrievalComputational EfficiencyRepresentation LearningTransformerContrastive LearningTextRetrieval-Augmented Generation

🎯 What it does: Proposed a lightweight, embedding-based, context-aware re-ranker called EBCAR, specifically designed for re-ranking tasks in retrieval-augmented generation (RAG);

Emergence of Spatial Representation in an Actor-Critic Agent with Hippocampus-Inspired Sequence Generator

Xiao-Xiong Lin (University of Freiburg), Christian Leibold (University of Freiburg)

CodeComputational EfficiencyRepresentation LearningConvolutional Neural NetworkRecurrent Neural NetworkReinforcement LearningImage

🎯 What it does: Developed a shift-register RNN inspired by the hippocampus CA3, combined with sparse DG input and actor-critic learning, achieving self-centered perception-driven navigation in the visual-based DeepMind Lab continuous maze;

Emergent Coordination in Multi-Agent Language Models

Christoph Riedl (Northeastern University)

CodeTransformerLarge Language ModelPrompt EngineeringText

🎯 What it does: This paper designs a non-communicative group guessing game, quantitatively evaluates the emergence, collaboration, and identity differences in multi-agent LLM systems using information theory (PID, TDMI) and mixed-effects models, and investigates how prompts (Plain, Persona, ToM) can control systems to transition from disordered collections to higher-order collaborative collectives.

Emergent Misalignment is Easy, Narrow Misalignment is Hard

Anna Soligo (Imperial College London), Neel Nanda (Imperial College London)

CodeExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningText

🎯 What it does: Investigated the 'emergent misalignment' phenomenon in large language models after fine-tuning on fine-grained harmful data, and proposed linear direction representations along with their monitoring and mitigation methods.

EMFuse: Energy-based Model Fusion for Decision Making

Kejie He (Nanjing University), Yang Yu (Nanjing University)

CodeTransformerLarge Language ModelReinforcement LearningTextSequential

🎯 What it does: Proposed a unified energy model fusion framework (EMFuse) that merges models through energy addition in two decision tasks: direct strategy fusion and dynamics model fusion, and designed EMSelect strategy selection and ADETM energy dynamics model to achieve efficient uncertainty estimation and fusion.

EmoPrefer: Can Large Language Models Understand Human Emotion Preferences?

Zheng Lian (Tongji University), Jianhua Tao (Tsinghua University)

CodeRecognitionTransformerLarge Language ModelPrompt EngineeringMixture of ExpertsVision Language ModelVideoTextMultimodalityBenchmark

🎯 What it does: Constructed the first emotional preference dataset EmoPrefer-Data and proposed EmoPrefer-Bench, establishing a benchmark for evaluating the emotional preference judgment of multimodal large language models (MLLM) in descriptive emotional recognition (DMER).

EmotionHallucer: Evaluating Emotion Hallucinations in Multimodal Large Language Models

Bohao Xing (Lappeenranta-Lahti University of Technology), Heikki KΓ€lviΓ€inen

CodeExplainability and InterpretabilityLarge Language ModelTextMultimodalityBenchmarkAudio

🎯 What it does: Proposed the EmotionHallucer benchmark to assess hallucinations in multimodal large language models for emotion understanding.

EmotionThinker: Prosody-Aware Reinforcement Learning for Explainable Speech Emotion Reasoning

Dingdong WANG, Helen M. Meng (The Chinese University of Hong Kong)

CodeClassificationRecognitionExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextMultimodalityChain-of-ThoughtAudio

🎯 What it does: Propose EmotionThinker, a reinforcement learning-based speech emotion recognition model capable of generating interpretable emotional judgments and reasoning processes.

Empowering Efficiency and Efficacy in WebAgent via Enabling Info-Rich Seeking

Zhengwei Tao (Peking University), Zhiqiang Gao (Southeast University)

CodeRetrievalComputational EfficiencyLarge Language ModelAgentic AITabular

🎯 What it does: Developed a WebLeaper framework that trains LLMs for efficient information seeking through entity-dense tree-structured task synthesis and information-oriented trajectory filtering.

Empowering LLM Tool Invocation with Tool-call Reward Model

Da Ma (Shanghai Jiao Tong University), Lu Chen (Shanghai Jiao Tong University)

CodeAI Code AssistantReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningText

🎯 What it does: Proposed the Tool Call Reward Model (TRM), providing fine-grained rewards for large language models when using external tools, and combined PPO and GRPO reinforcement learning to enhance performance in search and code generation tasks.

Empowering Small VLMs to Think with Dynamic Memorization and Exploration

Jiazhen Liu (Hong Kong University of Science and Technology), Long Chen (Hong Kong University of Science and Technology)

CodeSupervised Fine-TuningReinforcement LearningPrompt EngineeringVision Language ModelImageTextMultimodalityChain-of-Thought

🎯 What it does: Proposes a dynamic memory-exploration training paradigm called DyME, enabling small-scale vision-language models (SVLM) to acquire reasoning capabilities.

Enabling arbitrary inference in spatio-temporal dynamic systems: A physics-inspired perspective

Yan Ge, Yang Wang (University Of Science And Technology Of China)

CodeComputational EfficiencyConvolutional Neural NetworkGraph Neural NetworkGraphTime SeriesPhysics Related

🎯 What it does: Proposed the PhySTA framework, combining Graph-Time Fourier Neural Operator with a multi-scale adaptive interaction module to model continuous spatiotemporal dynamics in graph structures and enable arbitrary region inference.

Enabling Fine-Tuning of Direct Feedback Alignment via Feedback-Weight Matching

Yunseok Lee (UNIST), Seulki Lee (KAIST)

CodeClassificationConvolutional Neural NetworkTransformerSupervised Fine-TuningImageText

🎯 What it does: Proposed the feedback-weight matching method, enabling direct feedback alignment (DFA) to reliably fine-tune pre-trained networks.

Enabling True Global Perception in State Space Models for Visual Tasks

Jie Hui (Xi'an Jiaotong University), Jianji Wang (Xi'an Jiaotong University)

CodeObject DetectionSegmentationImage

🎯 What it does: Define global modeling theoretically, design GSSM and implement the GMamba module, using 2D-DFT pre-modulation to enhance the global perception of SSM, thereby achieving efficient and plug-and-play visual global modeling.

End-to-End Probabilistic Framework for Learning with Hard Constraints

Utkarsh Utkarsh (MIT CSAIL), Bernie Wang

CodeOptimizationTime SeriesPhysics Related

🎯 What it does: Propose the ProbHardE2E framework, achieving end-to-end probabilistic prediction while strictly satisfying hard constraints and providing uncertainty quantification.