arXivSub Start free trial

ICLR 2026 Papers with Code β€” Page 21

International Conference on Learning Representations Β· 2207 papers

Towards Self-Robust LLMs: Intrinsic Prompt Noise Resistance via CoIPO

Xin Yang (Zhejiang University), Wenyuan Jiang (ETH Zurich)

CodeReinforcement Learning from Human FeedbackTransformerContrastive LearningTextBenchmark

🎯 What it does: To address the issue of performance fluctuations in large language models when prompts contain minor noise, this paper proposes the CoIPO framework, which enhances the model's robustness to prompt noise through contrastive learning and inverse DPO (Inverse DPO) during the post-training phase.

Towards Understanding Subliminal Learning: When and How Hidden Biases Transfer

Simon Schrodi (University of Freiburg), Thomas Brox (University of Freiburg)

CodeExplainability and InterpretabilityKnowledge DistillationRepresentation LearningTransformerPrompt EngineeringText

🎯 What it does: Studied the unconscious learning phenomenon in language models during distillation, revealing through comparative experiments and mechanism analysis that differential words are the core driving force behind this phenomenon.

Towards Understanding the Nature of Attention with Low-Rank Sparse Decomposition

Zhengfu He (Shanghai Innovation Institute), Xipeng Qiu (Shanghai Innovation Institute)

CodeExplainability and InterpretabilityTransformerAuto EncoderText

🎯 What it does: Designed and implemented a low-rank sparse attention model called Lorsa to replace the original multi-head self-attention (MHSA), decomposing and interpreting attention units through sparsification and single-dimensional OV circuits.

Towards Understanding Valuable Preference Data for Large Language Model Alignment

Zizhuo Zhang (Hong Kong Baptist University), Masashi Sugiyama (University Of Tokyo)

CodeData-Centric LearningReinforcement Learning from Human FeedbackLarge Language ModelSupervised Fine-TuningReinforcement LearningTextBenchmark

🎯 What it does: Propose a truncated influence function (TIF) to evaluate preference data quality, and based on TIF, introduce a lightweight LossDiff-IRM selection method to improve large language model alignment effectiveness.

TPRU: Advancing Temporal and Procedural Understanding in Large Multimodal Models

Zhenkun Gao (East China Normal University), Yuan Xie (East China Normal University)

CodeRobotic IntelligenceSupervised Fine-TuningReinforcement LearningVision Language ModelImageVideoTextMultimodalitySequential

🎯 What it does: Proposed and constructed the TPRU dataset, leveraging three temporal tasks (temporal reordering, next-frame prediction, previous-frame review) and negative samples, combined with reinforcement learning fine-tuning, significantly improving the performance of small-scale multimodal language models in temporal and program reasoning.

TRAC: Tensor-Train based Across-layer Compression for Parameter-Efficient Fine-Tuning

Bangguo Ye, Xiaoqun Zhang (Shanghai Jiao Tong University)

CodeCompressionComputational EfficiencyKnowledge DistillationRepresentation LearningSupervised Fine-TuningImageText

🎯 What it does: Propose a new parameter-efficient fine-tuning framework TRAC, which utilizes Tensor-Train decomposition and shares/freezes cores across layers, while incorporating a lightweight controller to achieve fine-tuning with fewer trainable parameters.

Traceable Black-Box Watermarks For Federated Learning

Jiahao Xu (University of Nevada, Reno), Zikai Zhang (University of Nevada, Reno)

CodeFederated LearningSafty and PrivacyImage

🎯 What it does: Propose a new federated learning framework called TraMark, which injects traceable black-box watermarks into the global model to protect model intellectual property and trace leakage sources.

Traceable Evidence Enhanced Visual Grounded Reasoning: Evaluation and Method

Haochen Wang (New Laboratory of Pattern Recognition, State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences), Zhaoxiang Zhang (New Laboratory of Pattern Recognition, State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences)

CodeExplainability and InterpretabilityTransformerSupervised Fine-TuningReinforcement LearningVision Language ModelMultimodalityBenchmarkChain-of-Thought

🎯 What it does: Propose TreeBench, a benchmark for evaluating vision-based image reasoning, and TreeVGR, a training framework, emphasizing traceable visual evidence and multi-step reasoning.

TRACED: Transition-aware Regret Approximation with Co-learnability for Environment Design

Geonwoo Cho (Gwangju Institute of Science and Technology), Sundong Kim (Gwangju Institute of Science and Technology)

CodeData SynthesisOptimizationReinforcement LearningBenchmark

🎯 What it does: Propose an unsupervised environment design method called TRACED, which enhances the agent's zero-shot generalization capability in RL training through adaptive task generation and a replay mechanism.

TRACEDET: HALLUCINATION DETECTION FROM THE DECODING TRACE OF DIFFUSION LARGE LANGUAGE MODELS

Shenxu Chang (University of Oxford), Jindong Gu (University of Oxford)

CodeAnomaly DetectionExplainability and InterpretabilityTransformerLarge Language ModelDiffusion modelText

🎯 What it does: Propose TraceDet, which detects hallucinatory outputs by leveraging the multi-step denoising process (decoding trajectory) of diffusion-based large language models (D-LLM).

Tracing and Reversing Edits in LLMs

Paul Youssef (Marburg University), JΓΆrg SchlΓΆtterer (Marburg University)

CodeComputational EfficiencyRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningText

🎯 What it does: Proposes a task that uses post-editing weights to track and reverse knowledge edits, along with the corresponding solution.

Tracking Equivalent Mechanistic Interpretations Across Neural Networks

Alan Sun (Carnegie Mellon University), Mariya Toneva (Max Planck Institute for Software Systems)

CodeExplainability and InterpretabilityTransformerTextSequential

🎯 What it does: This paper introduces the concept of 'interpretive equivalence' and designs an algorithm called Congruity, which uses the similarity of linear representations realized by models to determine whether two models implement the same high-level algorithm without requiring explicit explanations.

Train-before-Test Harmonizes Language Model Rankings

Guanhua Zhang (Max Planck Institute for Intelligent Systems), Moritz Hardt (Max Planck Institute for Intelligent Systems)

CodeLarge Language ModelSupervised Fine-TuningTextBenchmark

🎯 What it does: Proposed and implemented the 'train-before-test' method, which uniformly fine-tunes all 61 LLMs using the training set of each benchmark first, then evaluates on the test set to obtain a ranking of model potential.

Training Deep Normalization-Free Spiking Neural Networks with Lateral Inhibition

Peiyu Liu (Peking University), Zhaofei Yu (Peking University)

CodeClassificationSpiking Neural NetworkImage

🎯 What it does: Proposed a completely unnormalized deep spiking neural network (SNN) training framework that leverages the excitatory-inhibitory (E-I) separation and lateral inhibition mechanisms in the cortex to replace traditional normalization techniques.

Training Large Language Models To Reason In Parallel With Global Forking Tokens

Sheng Jia (University of Toronto), Shiva Kasiviswanathan

CodeTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextBenchmark

🎯 What it does: This paper introduces global branch tokens to train large language models for parallel reasoning, thereby improving reasoning accuracy while maintaining diversity.

Training-free Counterfactual Explanation for Temporal Graph Model Inference

Mingjian Lu (Case Western Reserve University), Yinghui Wu (Case Western Reserve University)

CodeExplainability and InterpretabilityGraph Neural NetworkGraphTime Series

🎯 What it does: This paper proposes TemGX, a training-free, instance-oriented temporal graph explanation framework that provides interpretable substructures of TGNN outputs through temporal subgraphs and temporal counterfactual analysis, while supporting time-pattern-based queries.

Training-Free Loosely Speculative Decoding: Accepting Semantically Correct Drafts Beyond Exact Match

Jinze Li (Advanced Micro Devices, Inc.), Emad Barsoum (Advanced Micro Devices, Inc.)

CodeComputational EfficiencyTransformerLarge Language ModelText

🎯 What it does: Proposes a training-agnostic loose speculative decoding (FLy) that enhances LLM inference speed by determining the semantic validity of mismatches using entropy gates and deferred windows.

TRAJECT-Bench:A Trajectory-Aware Benchmark for Evaluating Agentic Tool Use

Pengfei He (Michigan State University), Benoit Dumoulin (Hippocratic AI)

CodeLarge Language ModelAgentic AITextBenchmarkRetrieval-Augmented GenerationChain-of-Thought

🎯 What it does: Created TRAJECT-Bench, a novel evaluation benchmark for tool usage trajectories, assessing LLMs' ability to invoke, parameterize, and arrange tools in real-world tasks.

Trajectory Generation with Conservative Value Guidance for Offline Reinforcement Learning

Tieru Wang (Beijing University of Posts and Telecommunications), Guoshun Nan (Beijing University of Posts and Telecommunications)

CodeData SynthesisTransformerReinforcement LearningWorld ModelSequential

🎯 What it does: Proposed a Transformer-based offline reinforcement learning data augmentation framework called TGCVG, which utilizes conservative value-guided strategies and learned dynamics models to generate high-quality trajectories, and mixes them with original data for offline RL training.

Trajectory-aware Shifted State Space Models for Online Video Super-Resolution

Qiang Zhu (Pengcheng Laboratory), Ronggang Wang (Peking University)

CodeSuper ResolutionOptical FlowVideo

🎯 What it does: Proposed a trajectory-aware Shifted SSM model called TS-Mamba, which selects previous frame tokens using long-term trajectories and aggregates them through Hilbert scanning combined with shifted SSM for online video super-resolution.

TrajFlow: Nation-wide Pseudo GPS Trajectory Generation with Flow Matching Models

Peiran Li (Hitotsubashi University), Renhe Jiang (University of Tokyo)

CodeData SynthesisFlow-based ModelTime SeriesSequentialOrdinary Differential Equation

🎯 What it does: Proposed a flow matching generative framework named TrajFlow for generating multi-scale, multi-transportation mode synthetic GPS trajectories nationwide.

TrajTok: What makes for a good trajectory tokenizer in behavior generation?

Zhiyuan Zhang, Junchi Yan (Shanghai Jiao Tong University)

CodeGenerationAutonomous DrivingRepresentation LearningData-Centric LearningSequential

🎯 What it does: This paper investigates the role of trajectory tokenizers in behavior generation, proposing the TrajTok tokenizer, which combines regularized grids and data-driven filtering/expansion approaches to balance coverage, utilization, symmetry, and robustness, while incorporating spatially aware label smoothing during training;

Transducing Language Models

Vésteinn Snæbjarnarson (ETH Zürich), Tim Vieira

CodeGenerationRepresentation LearningLarge Language ModelTextBiomedical Data

🎯 What it does: Propose a framework based on finite-state transducers (FST) that converts existing language models into different units on demand, directly calculating the probabilities of the transformed units during inference without retraining;

Transformers as Unsupervised Learning Algorithms: A study on Gaussian Mixtures

Zhiheng Chen (Fudan University), Guanhua Fang (Fudan University)

CodeMeta LearningTransformerTabular

🎯 What it does: Propose a meta-learning framework called TGMM based on Transformer, which can simultaneously solve Gaussian Mixture Model (GMM) estimation tasks with different numbers of components during inference;

Transitive RL: Value Learning via Divide and Conquer

Seohong Park (University of California, Berkeley), Sergey Levine (University of California, Berkeley)

CodeReinforcement LearningSequentialBenchmark

🎯 What it does: Propose Transitive RL (TRL), an offline goal-conditioned reinforcement learning (GCRL) value learning algorithm based on the divide-and-conquer paradigm.

TraPO: A Semi-Supervised Reinforcement Learning Framework for Boosting LLM Reasoning

Shenzhi Yang (Zhejiang University), Gang Chen (Zhejiang University)

CodeTransformerLarge Language ModelReinforcement LearningText

🎯 What it does: Propose a semi-supervised reinforcement learning framework TRAPO, which utilizes a small number of labeled samples to guide RLVR training on large-scale unlabeled data, and dynamically selects reliable unlabeled samples through trajectory similarity.

Tree Search for LLM Agent Reinforcement Learning

Yuxiang Ji (Xiamen University), Liaoni Wu (Southern University of Science and Technology)

CodeLarge Language ModelReinforcement LearningAgentic AIText

🎯 What it does: Proposed a multi-step interactive reinforcement learning method called Tree-GRPO based on tree search, which uses a tree structure with complete think-action-observe nodes for rollout;

Tree-sliced Sobolev IPM

Viet-Hoang Tran (National University of Singapore), Tan Minh Nguyen (National University of Singapore)

CodeGenerationRepresentation LearningDiffusion modelScore-based ModelImageText

🎯 What it does: Propose Tree-Sliced Sobolev Integral Probability Metric (TS-Sobolev) and its spherical variant STS-Sobolev as new methods to compute tree-cutting distances for any pβ‰₯1;

TreeGrad-Ranker: Feature Ranking via $O(L)$-Time Gradients for Decision Trees

Weida Li, Bryan Kian Hsiang Low (National University of Singapore)

CodeExplainability and InterpretabilityTabular

🎯 What it does: This paper proposes a gradient ascent-based feature ranking method called TreeGrad-Ranker, which directly addresses the joint optimization problem corresponding to insertion/deletion metrics.

Triangle Multiplication is All You Need for Biomolecular Structure Representations

Jeffrey Ouyang-Zhang (Genesis Research), Maruan Al-Shedivat (Genesis Research)

CodeProtein Structure PredictionTransformerDiffusion modelBiomedical Data

🎯 What it does: Proposed a simplified biomolecular structure prediction backbone network called Pairmixer, removing Triangle Attention and retaining only Triangle Multiplication and FFN to reduce computational costs;

TRIBE: TRImodal Brain Encoder for whole-brain fMRI response prediction

StΓ©phane d'Ascoli (Meta AI), Jean-Remi King

CodeRepresentation LearningTransformerLarge Language ModelMixture of ExpertsVideoTextMultimodalityMagnetic Resonance ImagingAudio

🎯 What it does: TRIBE is a multimodal deep encoding model that leverages pre-trained features from text, audio, and video, combining Transformers and multi-subject conditional layers to predict whole-brain fMRI responses when watching videos.

Tricks or Traps? A Deep Dive into RL for LLM Reasoning

Zihe Liu (Alibaba Group), Bo Zheng (Alibaba Group)

CodeTransformerLarge Language ModelReinforcement LearningText

🎯 What it does: Systematically reproduce and fine-grained evaluate commonly used reinforcement learning (RL) techniques in LLM inference tasks under a unified open-source framework, ultimately proposing LitePPO, which requires only two core techniques (advantage normalization and token-level loss aggregation), significantly improving the inference accuracy of baseline models.

Trion: FFT-based Dynamic Subspace Selection for Low-Rank Adaptive Optimization of LLMs

Ionut-Vlad Modoranu (Institute of Science and Technology Austria), Dan Alistarh (Institute of Science and Technology Austria)

CodeOptimizationComputational EfficiencyTransformerText

🎯 What it does: Propose a dynamic column selection method based on Discrete Cosine Transform (DCT), replacing traditional SVD/QR low-rank projections, and construct two low-rank adaptive optimizers: Trion and DCT-AdamW;

Triple-BERT: Do We Really Need MARL for Order Dispatch on Ride-Sharing Platforms?

Zijian Zhao (Hong Kong University of Science and Technology), Sen Li (Hong Kong University of Science and Technology)

CodeOptimizationTransformerReinforcement LearningTabular

🎯 What it does: Proposed and implemented Triple-BERT, a centralized framework based on single-agent reinforcement learning for large-scale ride-hailing order dispatch.

TripleSumm: Adaptive Triple-Modality Fusion for Video Summarization

Sumin Kim (Seoul National University), Joonseok Lee (Seoul National University)

CodeTransformerVision Language ModelVideoTextMultimodalityAudio

🎯 What it does: Propose the TripleSumm architecture, which can dynamically weight and fuse visual, textual, and audio multimodal information at the frame level for video summarization.

TrojanTO: Action-Level Backdoor Attacks Against Trajectory Optimization Models

Yang Dai (National University of Defense Technology), Li Shen (Sun Yat-sen University)

CodeOptimizationAdversarial AttackTransformerSequential

🎯 What it does: This paper proposes a post-training action-level backdoor attack framework called TrojanTO for trajectory optimization (Trajectory Optimization) models, which can implant a powerful trigger on only 0.3% of trajectory samples while maintaining the original task performance;

Trust-Region Adaptive Policy Optimization

Mingyu Su (Tsinghua University), Hongning Wang (Tsinghua University)

CodeTransformerSupervised Fine-TuningReinforcement LearningTextBenchmark

🎯 What it does: This paper proposes the TRAPO framework, which alternates supervised fine-tuning (SFT) and reinforcement learning (RL) within each training instance, guided by expert prefixes to enable model learning and exploration.

TrustJudge: Inconsistencies of LLM-as-a-Judge and How to Alleviate Them

Yidong Wang (Peking University), Shikun Zhang (Peking University)

CodeLarge Language ModelTextBenchmark

🎯 What it does: Systematically analyze inconsistencies in LLM evaluation frameworks and propose the TrustJudge method to address inconsistencies in score comparison and pairwise transitivity.

Truthful or Fabricated? Using Causal Attribution to Mitigate Reward Hacking in Explanations

Pedro Lobato Ferreira (University of Amsterdam), Ivan Titov (University of Edinburgh)

CodeExplainability and InterpretabilityReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningPrompt EngineeringTextChain-of-Thought

🎯 What it does: Investigate the reward hacking phenomenon in chain-of-thought (CoT) explanations of reward models, proposing to supplement reward model inputs with interpretive information derived from causal attribution to detect and reduce the probability of generating unfaithful CoT.

TS-Attn: Temporal-wise Separable Attention for Multi-Event Video Generation

Hongyu Zhang (Nankai University), Daquan Zhou (Nankai University)

CodeGenerationTransformerDiffusion modelVideoTextMultimodality

🎯 What it does: Proposes a training-agnostic temporal separable attention mechanism, TS-Attn, to enhance the accuracy and coherence of multi-event video generation.

TS-DDAE: A Novel Temporal-Spectral Denoising Diffusion AutoEncoder for Wireless Signal Recognition Model Pre-training

Yaoqi Liu (Beijing University of Posts and Telecommunications), Chuan Shi (Beijing University of Posts and Telecommunications)

CodeRecognitionConvolutional Neural NetworkTransformerDiffusion modelAuto EncoderTime SeriesSequential

🎯 What it does: Propose a wireless signal pre-training framework TS-DDAE based on diffusion models, which learns robust features by adding noise and learning denoising simultaneously in both the time domain and spectral domain;

TS$^2$: Training with Sparsemax+, Testing with Softmax for Accurate and Diverse LLM Fine-Tuning

XuZiyang, Yuangang Pan (Centre for Frontier AI Research, Agency for Science, Technology and Research)

CodeLarge Language ModelSupervised Fine-TuningTextBenchmark

🎯 What it does: This paper proposes a method (TS²) for fine-tuning large language models that first uses Sparsemax+ during training and then employs Softmax for inference, significantly improving generation quality and diversity.

TSM-Bench: Detecting LLM-Generated Text in Real-World Wikipedia Editing Practices

Gerrit Quaremba (King's College London), Elena Simperl (Wikimedia Foundation)

CodeClassificationTransformerLarge Language ModelTextBenchmark

🎯 What it does: This paper constructs a multilingual, multi-generator, multi-task benchmark dataset called TSM-BENCH to evaluate the detection performance of machine-generated text in Wikipedia editing scenarios.

Tucker-FNO: Tensor Tucker-Fourier Neural Operator and its Universal Approximation Theory

Guancheng Zhou (Xi'an Jiaotong University), Deyu Meng (Xi'an Jiaotong University)

CodeRestorationImageVideoPhysics Related

🎯 What it does: Proposed Tucker-FNO, which decomposes high-dimensional Fourier Neural Operator into multiple one-dimensional FNOs via Tucker tensor decomposition, significantly reducing FFT computational complexity while preserving expressive power, and validated on PDE solving and high-dimensional visual signal recovery tasks.

TumorChain: Interleaved Multimodal Chain-of-Thought Reasoning for Traceable Clinical Tumor Analysis

Sijing Li (Zhejiang University), Ling Zhang (DAMO Academy, Alibaba Group)

CodeClassificationSegmentationAnomaly DetectionConvolutional Neural NetworkTransformerPrompt EngineeringVision Language ModelMultimodalityBiomedical DataComputed TomographyBenchmarkChain-of-Thought

🎯 What it does: Constructed the TumorChain framework and the TumorCoT-1.5M dataset, achieving multi-modal chain reasoning from CT images to pathological predictions.

Tuning the burn-in phase in training recurrent neural networks improves their performance

Julian D. Schiller (Leibniz University Hannover), Matthias A. MΓΌller (Leibniz University Hannover)

CodeOptimizationHyperparameter SearchRecurrent Neural NetworkTime Series

🎯 What it does: This paper studies and quantifies the impact of the burn-in phase (the first m steps ignored after network initialization) on training performance when training recurrent neural networks using truncated backpropagation through time (TBPTT), and provides theoretical upper bounds on the performance loss (regret).

TurboBoA: Faster and Exact Attention-aware Quantization without Backpropagation

Junhan Kim (Samsung Research), Yongkweon Jeon (Samsung Research)

CodeComputational EfficiencyTransformerText

🎯 What it does: Proposes TurboBoA, a gradient-free post-training quantization algorithm that can quickly quantize large language models to low bit widths while maintaining accuracy.

TurboQuant: Online Vector Quantization with Near-optimal Distortion Rate

Amir Zandieh (Google Research), Vahab Mirrokni (Google Research)

CodeCompressionOptimizationComputational EfficiencyTextTabular

🎯 What it does: Proposes an online vector quantization method called TURBOQUANT, which achieves near-optimal distortion rates at any bit width.

TusoAI: Agentic Optimization for Scientific Methods

Alistair Turcan (Carnegie Mellon University), Martin Jinye Zhang (Carnegie Mellon University)

CodeOptimizationNeural Architecture SearchTransformerLarge Language ModelAgentic AIBiomedical DataBenchmark

🎯 What it does: Developed TusoAI, an agent-based AI system that leverages large language models and structured domain knowledge trees to automate the development and optimization of scientific computing methods.

Tversky Neural Networks: Psychologically Plausible Deep Learning with Differentiable Tversky Similarity

Moussa Koulako Bala Doumbouya (Stanford University), Christopher D Manning (Stanford University)

CodeClassificationRepresentation LearningConvolutional Neural NetworkTransformerImageText

🎯 What it does: This paper proposes a differentiable Tversky similarity and its projection layer, replacing traditional linear projection layers in image recognition and language modeling tasks, demonstrating higher accuracy and fewer parameters.

TwinFlow: Realizing One-step Generation on Large Models with Self-adversarial Flows

Zhenglin Cheng (Westlake University), Tao Lin (Westlake University)

CodeGenerationData SynthesisSupervised Fine-TuningFlow-based ModelImageTextMultimodality

🎯 What it does: Proposed the TWINFLOW framework, achieving one-step generation for large generative models through self-adversarial dual-track training;

Two-Way Is Better Than One: Bidirectional Alignment with Cycle Consistency for Exemplar-Free Class-Incremental Learning

Hongye Xu (Rochester Institute of Technology), Bartosz Krawczyk (Rochester Institute of Technology)

CodeClassificationKnowledge DistillationRepresentation LearningImageBenchmark

🎯 What it does: Propose a bidirectional cyclic consistency projection (BiCyc) framework to compensate for feature drift in class-incremental learning without sample memory.

Type-Compliant Adaptation Cascades

Chu-Cheng Lin (Google), Eugene Ie (Google)

CodeOptimizationComputational EfficiencyRepresentation LearningData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningText

🎯 What it does: Proposed the Type-Compatible Adaptable Cascading (TACs) framework, viewing large language model workflows as unnormalized probabilistic programs with type constraints, and achieving end-to-end training through gradient optimization.

TyphoonMLA: A Mixed Naive-Absorb MLA Kernel For Shared Prefix

Ahmet Caner YΓΌzΓΌgΓΌler (Huawei), Lukas Cavigelli (Huawei)

CodeComputational EfficiencyText

🎯 What it does: Proposed TyphoonMLA, a hybrid MLA kernel combining naive and absorb implementations, leveraging shared prefixes to enhance attention computation efficiency.

U-MARVEL: Unveiling Key Factors for Universal Multimodal Retrieval via Embedding Learning with MLLMs

Xiaojie Li (Tencent PCG), Xi Chen (Tencent PCG)

CodeRetrievalKnowledge DistillationVision Language ModelContrastive LearningImageVideoTextMultimodalityBenchmark

🎯 What it does: Developed and evaluated a general-purpose retrieval framework called U-MARVEL based on multimodal large language models.

ULD-Net: Enabling Ultra-Low-Degree Fully Polynomial Networks for Homomorphically Encrypted Inference

Xi Xie (University of Connecticut), Caiwen Ding (University of Minnesota - Twin Cities)

CodeComputational EfficiencyConvolutional Neural NetworkTransformerImage

🎯 What it does: This work proposes a training method called ULD-Net, which can train fully low-order (multiplication depth ≀3) polynomial networks on ImageNet and Transformer scales, achieving high accuracy while maintaining HE inference efficiency.

UME-R1: Exploring Reasoning-Driven Generative Multimodal Embeddings

Zhibin Lan (Xiamen University), Jinsong Su (Xiamen University)

CodeRepresentation LearningTransformerSupervised Fine-TuningReinforcement LearningPrompt EngineeringImageVideoTextMultimodalityChain-of-Thought

🎯 What it does: Explore the reasoning-driven generative multi-modal embedding framework UME-R1;

Unbiased Gradient Estimation for Event Binning via Functional Backpropagation

Jinze Chen (University of Science and Technology of China), Zheng-Jun Zha (University of Science and Technology of China)

CodeAutonomous DrivingOptimizationComputational EfficiencySimultaneous Localization and MappingOptical Flow

🎯 What it does: Proposed an unbiased gradient estimation framework based on Functional Backpropagation for gradient computation of discrete packing in event vision.

Unbiased Object Detection Beyond Frequency with Visually Prompted Image Synthesis

Xinhao Cai (Nanjing University of Science and Technology), Wenguan Wang (Zhejiang University)

CodeObject DetectionData SynthesisPrompt EngineeringDiffusion modelImage

🎯 What it does: Proposes a generative unbiased object detection framework that dynamically corrects frequency and diversity biases through representative score-driven layout re-calibration and visual blueprint prompt-based L2I synthesis.

Uncertainty as Feature Gaps: Epistemic Uncertainty Quantification of LLMs in Contextual Question-Answering

Yavuz Faruk Bakman (University of Southern California), Sai Praneeth Karimireddy (University of Southern California)

CodeExplainability and InterpretabilityComputational EfficiencyTransformerLarge Language ModelPrompt EngineeringContrastive LearningText

🎯 What it does: This paper proposes a theoretically grounded confidence measurement method for context-aware question answering tasks. It estimates empirical uncertainty by leveraging the gap between the model's hidden layer and an ideal model, and decomposes and quantifies uncertainty through three semantic features (context dependency, context understanding, and honesty).

Uncertainty Estimation via Hyperspherical Confidence Mapping

Eunseo Choi (KAIST), Heejin Ahn (KAIST)

CodeClassificationDepth EstimationAnomaly DetectionOptimizationImageTabular

🎯 What it does: Propose a sampling-free and distribution-assumption-free uncertainty estimation framework called Hyperspherical Confidence Mapping (HCM), which decomposes network outputs into magnitude and unit vectors, using vector deviation from the unit sphere as an uncertainty metric.

Uncertainty-Aware Gaussian Map for Vision-Language Navigation

Jianzhe Gao (Zhejiang University), Wenguan Wang (Zhejiang University)

CodeAutonomous DrivingExplainability and InterpretabilityTransformerVision-Language-Action ModelGaussian SplattingTextMultimodalityPoint CloudBenchmark

🎯 What it does: Built a Semantic Gaussian Map based on semantic Gaussian point clouds, explicitly modeling geometric, semantic, and appearance uncertainties, which are finally fused into a unified 3D Value Map to guide Vision-Language Navigation.

Uncertainty-driven Embedding Convolution

Sungjun Lim (Yonsei University), Kyungwoo Song (Yonsei University)

CodeClassificationRetrievalConvolutional Neural NetworkContrastive LearningText

🎯 What it does: Propose the Uncertainty-Driven Embedding Convolution (UEC) framework, which first converts pre-trained deterministic word vectors into probabilistic vectors using posterior Laplacian approximation, then adaptively aggregates these probabilistic vectors with uncertainty-driven weights, and employs variance-based similarity estimation for downstream tasks;

Uncover Underlying Correspondence for Robust Multi-view Clustering

Haochen Zhou (Sichuan University), Xi Peng (Sichuan University)

CodeRepresentation LearningContrastive LearningImage

🎯 What it does: This paper proposes a generative framework called CorreGen for multi-view clustering in the presence of noisy correspondences (category-level and sample-level mismatch), automatically discovering potential cross-view correspondences.

Uncovering Semantic Selectivity of Latent Groups in Higher Visual Cortex with Mutual Information-Guided Diffusion

Yule Wang (Georgia Institute of Technology), Anqi Wu (Georgia Institute of Technology)

CodeDiffusion modelAuto EncoderImage

🎯 What it does: Learn neural latent subspaces through a group-level decoupled variational autoencoder, and visualize semantic features of each subspace using a diffusion model guided by mutual information maximization, revealing semantic selectivity in the higher visual cortex.

Understanding and Improving Continuous LLM Adversarial Training via In-context Learning Theory

Shaopeng Fu (King Abdullah University of Science and Technology), Di Wang (King Abdullah University of Science and Technology)

CodeSafty and PrivacyAdversarial AttackTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark

🎯 What it does: Studied the effective mechanisms of Continuous Adversarial Training (CAT) in large language models (LLMs) against jailbreak attacks, and proposed an improved method called ER-CAT;

Understanding and Improving Hyperbolic Deep Reinforcement Learning

Timo Klein (University of Vienna), Sebastian Tschiatschek (University of Vienna)

CodeOptimizationReinforcement LearningImage

🎯 What it does: Proposed HYPER++, an architecture that stabilizes hyperbolic space deep reinforcement learning through regularization, feature scaling, and classification value loss;

Understanding Cross-layer Contributions to Mixture-of-Experts Routing in LLMs

Wengang Li (Institute of Science Tokyo), Mohamed Wahib (RIKEN)

CodeExplainability and InterpretabilityTransformerMixture of ExpertsText

🎯 What it does: This paper proposes a recursive decomposition framework to quantify the contributions of different model components (such as tokens, attention layers, MoE outputs, and attention heads) to Mixture-of-Experts (MoE) routing decisions, and performs cross-layer interpretability analysis on four mainstream MoE LLMs.

Understanding Task Vectors in In-Context Learning: Emergence, Functionality, and Limitations

Yuxin Dong (Ohio State University), Xia Ning (Ohio State University)

CodeExplainability and InterpretabilityMeta LearningTransformerText

🎯 What it does: This paper analyzes the generation mechanism and function of task vectors in ICL, proposing and verifying the hypothesis that 'task vectors represent demonstrations';

Understanding Transformers for Time Series: Rank Structure, Flow-of-ranks, and Compressibility

Annan Yu (Center for Applied Mathematics, Cornell University), Bernie Wang

CodeCompressionComputational EfficiencyTransformerTime Series

🎯 What it does: This paper systematically investigates the rank structures of the embedding layer, attention layer, and deep networks in time series Transformers (TSFM) through numerical rank analysis, introduces the concept of 'flow-of-ranks,' and applies it to compress large-scale TSFM models such as Chronos;

Understanding vs. Generation: Navigating Optimization Dilemma in Multimodal Models

Sen Ye (Peking University), Winston Hu

CodeOptimizationReinforcement LearningVision Language ModelDiffusion modelFlow-based ModelImageTextMultimodalityBenchmark

🎯 What it does: Proposed the Reason-Reflect-Refine (R3) framework, decomposing image generation into three stages: reasoning, reflection, and refinement, while explicitly leveraging the model's understanding capabilities to enhance both generation and comprehension performance.

Uni-CoT: Towards Unified Chain-of-Thought Reasoning Across Text and Vision

Luozheng Qin (Shanghai Academy Of Ai For Science), Hao Li (Fudan University)

CodeGenerationTransformerMixture of ExpertsVision Language ModelMultimodalityChain-of-Thought

🎯 What it does: Developed a unified Chain-of-Thought framework, Uni-CoT, enabling cross-modal reasoning between vision and text, and enhancing performance in image generation and understanding tasks.

Uni-DPO: A Unified Paradigm for Dynamic Preference Optimization of LLMs

Shangpin Peng (Harbin Institute of Technology), Min Zhang (Harbin Institute of Technology)

CodeOptimizationReinforcement Learning from Human FeedbackLarge Language ModelReinforcement LearningText

🎯 What it does: Propose the Uni-DPO framework, achieving dynamic weighted learning based on direct preference optimization.

Uni-X: Mitigating Modality Conflict with a Two-End-Separated Architecture for Unified Multimodal Models

Jitai Hao (Baidu Inc), Jun Yu (Baidu Inc)

CodeGenerationTransformerVision Language ModelAuto EncoderImageTextMultimodality

🎯 What it does: Proposed and implemented the Uni-X unified multimodal model, which employs modality-specific layers in shallow and deep layers, and shares parameters in the middle layer to alleviate gradient conflicts between vision and text.

UniCA: Unified Covariate Adaptation for Time Series Foundation Model

Lu Han (Ant Group), De-Chuan Zhan (Nanjing University)

CodeDomain AdaptationTransformerImageTextMultimodalityTime Series

🎯 What it does: Explored how to make time series foundation models (TSFMs) compatible with general covariate (homogeneous, heterogeneous) prediction tasks, and proposed the UniCA framework to achieve unified assimilation and fusion of covariates.

Uniform Discrete Diffusion with Metric Path for Video Generation

Haoge Deng (National Laboratory of Pattern Recognition), Xinlong Wang (Beijing Academy of Artificial Intelligence)

CodeGenerationTransformerLarge Language ModelDiffusion modelImageVideoTextMultimodality

🎯 What it does: In the paper, the authors propose the URSA framework, which utilizes a unified discrete diffusion model and a metric path to achieve video generation.

UniLiP: Adapting CLIP for Unified Multimodal Understanding, Generation and Editing

Hao Tang (Peking University), Liwei Wang (Peking University)

CodeRestorationGenerationTransformerVision Language ModelDiffusion modelMultimodality

🎯 What it does: Propose the UniLIP framework, extending CLIP into a unified model that supports visual understanding, image reconstruction, generation, and editing

UniOD: A Universal Model for Outlier Detection across Diverse Domains

Dazhi Fu (Chinese University of Hong Kong, Shenzhen), Jicong Fan (Chinese University of Hong Kong, Shenzhen)

CodeAnomaly DetectionGraph Neural NetworkTransformerTabularBenchmarkFinance Related

🎯 What it does: Propose a unified anomaly detection framework called UniOD, which trains a single model using historical labeled data to directly detect anomalies on new datasets (cross-domain, multi-dimensional);

UniQL: Unified Quantization and Low-rank Compression for Adaptive Edge LLMs

Hung-Yueh Chiang (University of Texas at Austin), Diana Marculescu (University of Texas at Austin)

CodeCompressionComputational EfficiencyLarge Language ModelText

🎯 What it does: Developed a unified post-training quantization and low-rank compression framework called UniQL for adaptive deployment of large language models on edge devices.

UniRestorer: Universal Image Restoration via Adaptively Estimating Image Degradation at Proper Granularity

Jingbo Lin (Harbin Institute of Technology), Wangmeng Zuo (Harbin Institute of Technology)

CodeRestorationMixture of ExpertsVision Language ModelImage

🎯 What it does: Designed and implemented UniRestorer, a multi-grained hybrid expert framework for uniformly handling various image restoration tasks;

UniTrack: Differentiable Graph Representation Learning for Multi-Object Tracking

Bishoy Galoaa (Northeastern University), Sarah Ostadabbas (Northeastern University)

CodeObject TrackingGraph Neural NetworkVideo

🎯 What it does: Propose a pluggable graph-theoretic loss function UniTrack, which unifies the optimization of detection accuracy, identity preservation, and spatiotemporal consistency during training, significantly reducing errors such as ID switching, temporal inconsistency, and cross-target ID swapping in multi-object tracking;

Universal Inverse Distillation for Matching Models with Real-Data Supervision (No GANs)

Nikita Maksimovich Kornilov, Alexander Korotin (Applied AI Institute)

CodeGenerationKnowledge DistillationDiffusion modelScore-based ModelFlow-based ModelImage

🎯 What it does: Proposed a general GAN-free reverse distillation framework RealUID for directly utilizing real data to accelerate generation in matching models (diffusion, flow matching, bridge matching, etc.).

Universal Multi-Domain Translation via Diffusion Routers

Duc Kieu (Deakin University), Thin Nguyen (Deakin University)

CodeImage TranslationDiffusion modelRectified FlowAuto EncoderImage

🎯 What it does: This paper proposes a unified multi-domain translation framework called Diffusion Router, which can learn mappings between any two domains, including indirect and direct translations, by utilizing only K-1 central domain paired data.

UniVideo: Unified Understanding, Generation, and Editing for Videos

Cong Wei (University of Waterloo), Wenhu Chen (University of Waterloo)

CodeGenerationTransformerLarge Language ModelPrompt EngineeringVision Language ModelDiffusion modelAuto EncoderVideoTextMultimodalityBenchmarkChain-of-Thought

🎯 What it does: A unified video multimodal model called UniVideo was constructed, capable of accomplishing video understanding, text/image-to-video generation, and editing under multimodal instructions within the same framework.

Unlearning Evaluation through Subset Statistical Independence

Chenhao Zhang (University of Queensland), Miao Xu (University of Queensland)

CodeSafty and PrivacyConvolutional Neural NetworkImage

🎯 What it does: Proposes a subset-level machine forgetting evaluation method called Split-Half Dependence Evaluation (SDE), which uses HSIC to measure the model's output dependency on subsets, enabling subset-level forgetting effect assessment without retraining or auxiliary classifiers.

Unlearning Isn't Invisible: Detecting Unlearning Traces in LLMs from Model Outputs

Yiwei Chen (Michigan State University), Sijia Liu (Michigan State University)

CodeAnomaly DetectionSafty and PrivacyExplainability and InterpretabilityRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningText

🎯 What it does: Investigate whether large language models (LLMs) leave detectable traces after performing a 'forget' operation, and verify that it is possible to determine whether a model has been forgotten by analyzing only model outputs (text or pre-logit activations).

Unleashing Perception-Time Scaling to Multimodal Reasoning Models

Yifan Li (Renmin University of China), Minghui Qiu (ByteDance)

CodeComputational EfficiencySupervised Fine-TuningReinforcement LearningVision Language ModelMultimodalityBenchmark

🎯 What it does: Investigate the impact of reasoning time expansion on the perception capabilities of large vision-language models (LVLMs), proposing and validating the Perception-Time Scaling (PTS) framework.

Unleashing Scientific Reasoning for Bio-experimental Protocol Generation via Structured Component-based Reward Mechanism

Haoran Sun (Shanghai Artificial Intelligence Laboratory), Xiaosong Wang (Shanghai Artificial Intelligence Laboratory)

CodeDrug DiscoveryTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextBiomedical DataBenchmarkChain-of-Thought

🎯 What it does: Automatically generate executable and logically rigorous biological experiment protocols by training large language models;

UnLoc: Leveraging Depth Uncertainties for Floorplan Localization

Matthias WΓΌest, Daniel Barath (ETH Zurich)

CodePose EstimationDepth EstimationConvolutional Neural NetworkSimultaneous Localization and MappingImageVideo

🎯 What it does: Propose UnLoc, a sequential visual localization method that leverages depth uncertainty for floor plan localization, utilizing a pre-trained monocular depth network and modeling the uncertainty of depth prediction through probability distributions, followed by fusing multi-frame observations with a histogram filter to complete pose estimation;

Unlocking the Essence of Beauty: Advanced Aesthetic Reasoning with Relative-Absolute Policy Optimization

Boyang Liu (Fudan University), Xuanjing Huang (Fudan University)

CodeOptimizationExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningImageMultimodalityChain-of-Thought

🎯 What it does: Propose the Aes-R1 framework to achieve bidirectional reasoning and scoring for image aesthetic assessment. The framework includes automatically generating multi-dimensional aesthetic reasoning data (AesCoT) and optimizing with reinforcement learning based on relative-absolute rewards (RAPO), enhancing model interpretability and scoring accuracy through a two-phase training process (SFT + RL).

Unlocking the Value of Text: Event-Driven Reasoning and Multi-Level Alignment for Time Series Forecasting

Siyuan Wang (East China Normal University), Yang Shu (East China Normal University)

CodeTransformerLarge Language ModelContrastive LearningTextMultimodalityTime SeriesAgriculture RelatedFinance RelatedRetrieval-Augmented Generation

🎯 What it does: Developed a dual-branch multi-modal time series prediction framework called VoT, which effectively integrates external text and numerical sequences using event-driven reasoning and multi-level alignment.

Unmasking Backdoors: An Explainable Defense via Gradient-Attention Anomaly Scoring for Pre-trained Language Models

Anindya Sundar Das (Umea University), Monowar Bhuyan (Umea University)

CodeAnomaly DetectionExplainability and InterpretabilityAdversarial AttackTransformerLarge Language ModelText

🎯 What it does: Propose an inference-time anti-backdoor defense framework named X-GRAAD, which detects and neutralizes backdoor triggers by leveraging the abnormal effects of trigger words on attention and gradients in pre-trained language models.

Unsupervised Representation Learning - an Invariant Risk Minimization Perspective

Yotam Norman (Technion Israel Institute of Technology), Ron Meir (Technion Israel Institute of Technology)

CodeDomain AdaptationRepresentation LearningAuto EncoderImage

🎯 What it does: Propose a new unsupervised invariant risk minimization (IRM) framework without labels, and design two methods within this framework: PICA (Principal Invariant Component Analysis) and VIAE (Variational Invariant AutoEncoder), achieving variable decomposition and feature extraction for cross-environment data.

Unsupervised Representation Learning for 3D Mesh Parameterization with Semantic and Visibility Objectives

AmirHossein Zamani (Autodesk Research), Arianna Rampini (Autodesk Research)

CodeSegmentationRepresentation LearningRecurrent Neural NetworkMesh

🎯 What it does: For the UV mapping task on 3D meshes, the authors propose an unsupervised differentiable framework that simultaneously achieves semantic alignment and view-friendly notch layout while preserving geometric fidelity.

Unveiling Perceptual Artifacts: A Fine-Grained Benchmark for Interpretable AI-Generated Image Detection

Yao Xiao (Sun Yat-sen University), Liang Lin (Sun Yat-sen University)

CodeAnomaly DetectionExplainability and InterpretabilityTransformerImageBenchmark

🎯 What it does: This paper proposes X-AIGD, a fine-grained interpretable AI-generated image detection benchmark, providing pixel-level multi-level perceptual defect annotations.

Unveiling Super Experts in Mixture-of-Experts Large Language Models

Zunhai Su (Tsinghua University), Kehong Yuan (Tsinghua University)

CodeExplainability and InterpretabilityTransformerLarge Language ModelMixture of ExpertsText

🎯 What it does: This paper proposes and systematically studies 'Super Experts' (SE) in Mixture-of-Experts LLMs, a small subset of experts critical for inference.

UrbanFeel:A Comprehensive Benchmark for Temporal and Perceptual Understanding of City Scenes through Human Perspective

Jun He (Sun Yat-sen University), Xiang Zhang (Sun Yat-sen University)

CodeExplainability and InterpretabilityLarge Language ModelPrompt EngineeringVision Language ModelImageMultimodalityBenchmark

🎯 What it does: Designed and constructed a comprehensive benchmark named UrbanFeel, aimed at evaluating the capabilities of multimodal large language models (MLLMs) in urban development understanding and subjective environmental perception.

Use the Online Network If You Can: Towards Fast and Stable Reinforcement Learning

Ahmed Hendawy (Technical University of Darmstadt), Carlo D'Eramo (University of WΓΌrzburg)

CodeReinforcement LearningBenchmark

🎯 What it does: Designed and verified a new target calculation method called Minimum of Online and Target Networks (MINTO), which takes the minimum value between the online network and target network estimates to achieve fast and stable reinforcement learning.

Using maximal information auxiliary variables to improve synthetic data generation based on TabPFN foundation models

Elias Chaibub Neto (Sage Bionetworks)

CodeData SynthesisSafty and PrivacyTransformerTabular

🎯 What it does: Proposed and implemented a synthetic data generation method based on Maximum Information Auxiliary Variable (MIAV), leveraging TabPFN (and TabICL) within a context learning framework to generate high-fidelity and privacy-preserving synthetic tabular data.

USTBench: Benchmarking and Dissecting Spatiotemporal Reasoning Capabilities of LLMs as Urban Agents

Siqi Lai (Hong Kong University of Science and Technology (Guangzhou)), Hao Liu (Hong Kong University of Science and Technology (Guangzhou))

CodeLarge Language ModelReinforcement LearningAgentic AIWorld ModelTextGraphTabularBenchmarkChain-of-Thought

🎯 What it does: Propose USTBench and construct UAgentEnv, an interactive urban environment, to evaluate LLMs in four processes of urban spatiotemporal reasoning: understanding, prediction, planning, and reflection, containing 62,466 structured question-answer pairs and nine real-world urban tasks.