AI研究論文每日精選

Chain-of-Model Learning for Language Model

AdaptThink: Reasoning Models Can Learn When to Think

AdaCoT: Pareto-Optimal Adaptive Chain-of-Thought Triggering via Reinforcement Learning

Delta Attention: Fast and Accurate Sparse Attention Inference by Delta Correction

Scaling Computer-Use Grounding via User Interface Decomposition and Synthesis

Thinkless: LLM Learns When to Think

Faster Video Diffusion with Trainable Sparse Attention

Seek in the Dark: Reasoning via Test-Time Instance-Level Policy Gradient in Latent Space

Model Merging in Pre-training of Large Language Models

MM-PRM: Enhancing Multimodal Mathematical Reasoning with Scalable Step-Level Supervision

Hybrid 3D-4D Gaussian Splatting for Fast Dynamic Scene Representation

FedSVD: Adaptive Orthogonalization for Private Federated Learning with LoRA

CPGD: Toward Stable Rule-based Reinforcement Learning for Language Models

Fractured Chain-of-Thought Reasoning

ChartMuseum: Testing Visual Reasoning Capabilities of Large Vision-Language Models

Neuro-Symbolic Query Compiler

SEED-GRPO: Semantic Entropy Enhanced GRPO for Uncertainty-Aware Policy Optimization

VisionReasoner: Unified Visual Perception and Reasoning via Reinforcement Learning

Through the Looking Glass: Common Sense Consistency Evaluation of Weird Images

ViPlan: A Benchmark for Visual Planning with Symbolic Predicates and Vision-Language Models

When AI Co-Scientists Fail: SPOT-a Benchmark for Automated Verification of Scientific Research

Accelerate TarFlow Sampling with GS-Jacobi Iteration

R3: Robust Rubric-Agnostic Reward Models

Tiny QA Benchmark++: Ultra-Lightweight, Synthetic Multilingual Dataset Generation & Smoke-Tests for Continuous LLM Evaluation

FinePhys: Fine-grained Human Action Generation by Explicitly Incorporating Physical Laws for Effective Skeletal Guidance

MTVCrafter: 4D Motion Tokenization for Open-World Human Image Animation

ExTrans: Multilingual Deep Reasoning Translation via Exemplar-Enhanced Reinforcement Learning

HISTAI: An Open-Source, Large-Scale Whole Slide Image Dataset for Computational Pathology

QVGen: Pushing the Limit of Quantized Video Generative Models

SoftCoT++: Test-Time Scaling with Soft Chain-of-Thought Reasoning

From Grunts to Grammar: Emergent Language from Cooperative Foraging

MedCaseReasoning: Evaluating and learning diagnostic reasoning from clinical case reports

HelpSteer3-Preference: Open Human-Annotated Preference Data across Diverse Tasks and Languages

A Token is Worth over 1,000 Tokens: Efficient Knowledge Distillation through Low-Rank Clone

LLM Context Conditioning and PWP Prompting for Multimodal Validation of Chemical Formulas

TechniqueRAG: Retrieval Augmented Generation for Adversarial Technique Annotation in Cyber Threat Intelligence Text

Learned Lightweight Smartphone ISP with Unpaired Data

AI-Driven Scholarly Peer Review via Persistent Workflow Prompting, Meta-Prompting, and Meta-Reasoning

Fast, Not Fancy: Rethinking G2P with Rich Data and Rule-Based Models

Creating General User Models from Computer Use