AI研究論文每日精選

Reflect, Retry, Reward: Self-Improving LLMs via Reinforcement Learning

UniWorld: High-Resolution Semantic Encoders for Unified Visual Understanding and Generation

VS-Bench: Evaluating VLMs for Strategic Reasoning and Decision-Making in Multi-Agent Environments

SynthRL: Scaling Visual Reasoning with Verifiable Data Synthesis

CSVQA: A Chinese Multimodal Benchmark for Evaluating STEM Reasoning Capabilities of VLMs

GUI-Actor: Coordinate-Free Visual Grounding for GUI Agents

FinMME: Benchmark Dataset for Financial Multi-Modal Reasoning Evaluation

OmniSpatial: Towards Comprehensive Spatial Reasoning Benchmark for Vision Language Models

OThink-R1: Intrinsic Fast/Slow Thinking Mode Switching for Over-Reasoning Mitigation

Visual Embodied Brain: Let Multimodal Large Language Models See, Think, and Control in Spaces

Sparse-vDiT: Unleashing the Power of Sparse Attention to Accelerate Video Diffusion Transformers

DINGO: Constrained Inference for Diffusion LLMs

Robot-R1: Reinforcement Learning for Enhanced Embodied Reasoning in Robotics

MotionSight: Boosting Fine-Grained Motion Understanding in Multimodal LLMs

Co-Evolving LLM Coder and Unit Tester via Reinforcement Learning

AnimeShooter: A Multi-Shot Animation Dataset for Reference-Guided Video Generation

Negative-Guided Subject Fidelity Optimization for Zero-Shot Subject-Driven Generation

LumosFlow: Motion-Guided Long Video Generation

Native-Resolution Image Synthesis

RelationAdapter: Learning and Transferring Visual Relation with Diffusion Transformers

DCM: Dual-Expert Consistency Model for Efficient and High-Quality Video Generation

FlowMo: Variance-Based Flow Guidance for Coherent Motion in Video Generation

Datasheets Aren't Enough: DataRubrics for Automated Quality Metrics and Accountability

PCoreSet: Effective Active Learning through Knowledge Distillation from Vision-Language Models

Ctrl-Crash: Controllable Diffusion for Realistic Car Crashes

Training Language Models to Generate Quality Code with Program Analysis Feedback

Self-Challenging Language Model Agents

Motion-Aware Concept Alignment for Consistent Video Editing

ORV: 4D Occupancy-centric Robot Video Generation

Accelerating Diffusion LLMs via Adaptive Parallel Decoding

MERIT: Multilingual Semantic Retrieval with Interleaved Multi-Condition Query

FuseLIP: Multimodal Embeddings via Early Fusion of Discrete Tokens

Multimodal DeepResearcher: Generating Text-Chart Interleaved Reports From Scratch with Agentic Framework

One Missing Piece for Open-Source Reasoning Models: A Dataset to Mitigate Cold-Starting Short CoT LLMs in RL

Angles Don't Lie: Unlocking Training-Efficient RL Through the Model's Own Signals

Hanfu-Bench: A Multimodal Benchmark on Cross-Temporal Cultural Understanding and Transcreation

ReFoCUS: Reinforcement-guided Frame Optimization for Contextual Understanding

SHARE: An SLM-based Hierarchical Action CorREction Assistant for Text-to-SQL

How Much Backtracking is Enough? Exploring the Interplay of SFT and RL in Enhancing LLM Reasoning

Deep Video Discovery: Agentic Search with Tool Use for Long-form Video Understanding

Controllable Human-centric Keyframe Interpolation with Generative Prior

TL;DR: Too Long, Do Re-weighting for Effcient LLM Reasoning Compression

M^3FinMeeting: A Multilingual, Multi-Sector, and Multi-Task Financial Meeting Understanding Evaluation Dataset

QARI-OCR: High-Fidelity Arabic Text Recognition through Multimodal Large Language Model Adaptation

Control-R: Towards controllable test-time scaling

R^2ec: Towards Large Recommender Models with Reasoning

ByteMorph: Benchmarking Instruction-Guided Image Editing with Non-Rigid Motions

Revisiting LRP: Positional Attribution as the Missing Ingredient for Transformer Explainability

Beyond In-Context Learning: Aligning Long-form Generation of Large Language Models via Task-Inherent Attribute Guidelines

Knowing Before Saying: LLM Representations Encode Information About Chain-of-Thought Success Before Completion