AI研究论文每日精选

Revisiting Multi-Agent Debate as Test-Time Scaling: A Systematic Study of Conditional Effectiveness

PixelThink: Towards Efficient Chain-of-Pixel Reasoning

Table-R1: Inference-Time Scaling for Table Reasoning

Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligence

The Climb Carves Wisdom Deeper Than the Summit: On the Noisy Rewards in Learning to Reason

VF-Eval: Evaluating Multimodal LLMs for Generating Feedback on AIGC Videos

ZeroGUI: Automating Online GUI Learning at Zero Human Cost

VideoReasonBench: Can MLLMs Perform Vision-Centric Complex Video Reasoning?

Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding

D-AR: Diffusion via Autoregressive Models

AnySplat: Feed-forward 3D Gaussian Splatting from Unconstrained Views

cadrille: Multi-modal CAD Reconstruction with Online Reinforcement Learning

Are Reasoning Models More Prone to Hallucination?

LoRAShop: Training-Free Multi-Concept Image Generation and Editing with Rectified Flow Transformers

Satori-SWE: Evolutionary Test-Time Scaling for Sample-Efficient Software Engineering

UniRL: Self-Improving Unified Multimodal Models via Supervised and Reinforcement Learning

ATLAS: Learning to Optimally Memorize the Context at Test Time

Train Sparse Autoencoders Efficiently by Utilizing Features Correlation

Multi-Domain Explainability of Preferences

SWE-bench Goes Live!

VidText: Towards Comprehensive Evaluation for Video Text Understanding

FAMA: The First Large-Scale Open-Science Speech Foundation Model for English and Italian

StressTest: Can YOUR Speech LM Handle the Stress?

Towards Safety Reasoning in LLMs: AI-agentic Deliberation for Policy-embedded CoT Data Creation

REOrdering Patches Improves Vision Models

DeepTheorem: Advancing LLM Reasoning for Theorem Proving Through Natural Language and Reinforcement Learning

Muddit: Liberating Generation Beyond Text-to-Image with a Unified Discrete Diffusion Model

On-Policy RL with Optimal Reward Baseline

SafeScientist: Toward Risk-Aware Scientific Discoveries by LLM Agents

System-1.5 Reasoning: Traversal in Language and Latent Spaces with Dynamic Shortcuts

GeoDrive: 3D Geometry-Informed Driving World Model with Precise Action Control

PatientSim: A Persona-Driven Simulator for Realistic Doctor-Patient Interactions

Differentiable Solver Search for Fast Diffusion Sampling

Breaking Down Video LLM Benchmarks: Knowledge, Spatial Perception, or True Temporal Understanding?

MAGREF: Masked Guidance for Any-Reference Video Generation

KVzip: Query-Agnostic KV Cache Compression with Context Reconstruction

ToMAP: Training Opponent-Aware LLM Persuaders with Theory of Mind

Uni-Instruct: One-step Diffusion Model through Unified Diffusion Divergence Instruction

ZeroSep: Separate Anything in Audio with Zero Training

Afterburner: Reinforcement Learning Facilitates Self-Improving Code Efficiency Optimization

ATI: Any Trajectory Instruction for Controllable Video Generation

Re-ttention: Ultra Sparse Visual Generation via Attention Statistical Reshape

One-shot Entropy Minimization

When Models Reason in Your Language: Controlling Thinking Trace Language Comes at the Cost of Accuracy

CXReasonBench: A Benchmark for Evaluating Structured Diagnostic Reasoning in Chest X-rays

Puzzled by Puzzles: When Vision-Language Models Can't Take a Hint

To Trust Or Not To Trust Your Vision-Language Model's Prediction

UniTEX: Universal High Fidelity Generative Texturing for 3D Shapes

CLIPGaussian: Universal and Multimodal Style Transfer Based on Gaussian Splatting

Concise Reasoning, Big Gains: Pruning Long Reasoning Trace with Difficulty-Aware Prompting

How Animals Dance (When You're Not Looking)

ZPressor: Bottleneck-Aware Compression for Scalable Feed-Forward 3DGS

Can LLMs Deceive CLIP? Benchmarking Adversarial Compositionality of Pre-trained Multimodal Representation via Text Updates

SridBench: Benchmark of Scientific Research Illustration Drawing of Image Generation Model

Lunguage: A Benchmark for Structured and Sequential Chest X-ray Interpretation

ChartLens: Fine-grained Visual Attribution in Charts

A Graph Perspective to Probe Structural Patterns of Knowledge in Large Language Models

MMSI-Bench: A Benchmark for Multi-Image Spatial Intelligence

Differential Information: An Information-Theoretic Perspective on Preference Optimization

Grounded Reinforcement Learning for Visual Reasoning

GSO: Challenging Software Optimization Tasks for Evaluating SWE-Agents

Evaluating Text Creativity across Diverse Domains: A Dataset and Large Language Model Evaluator

TokBench: Evaluating Your Visual Tokenizer before Visual Generation

Unsupervised Word-level Quality Estimation for Machine Translation Through the Lens of Annotators (Dis)agreement

Model-Preserving Adaptive Rounding

Adaptive Classifier-Free Guidance via Dynamic Low-Confidence Masking

Large Language Models Meet Knowledge Graphs for Question Answering: Synthesis and Opportunities

Toward Reliable Biomedical Hypothesis Generation: Evaluating Truthfulness and Hallucination in Large Language Models