AI研究论文每日精选

Can 1B LLM Surpass 405B LLM? Rethinking Compute-Optimal Test-Time Scaling

SynthDetoxM: Modern LLMs are Few-Shot Parallel Detoxification Data Annotators

Exploring the Limit of Outcome Reward for Learning Mathematical Reasoning

The Curse of Depth in Large Language Models

Training Language Models for Social Deduction with Multi-Agent Reinforcement Learning

LM2: Large Memory Models

Matryoshka Quantization

CODESIM: Multi-Agent Code Generation and Problem Solving through Simulation-Driven Planning and Debugging

Show-o Turbo: Towards Accelerated Unified Multimodal Understanding and Generation

ReasonFlux: Hierarchical LLM Reasoning via Scaling Thought Templates

Lossless Acceleration of Large Language Models with Hierarchical Drafting based on Temporal Locality in Speculative Decoding

MetaChain: A Fully-Automated and Zero-Code Framework for LLM Agents

Lumina-Video: Efficient and Flexible Video Generation with Multi-scale Next-DiT

EVEv2: Improved Baselines for Encoder-Free Vision-Language Models

History-Guided Video Diffusion

The Hidden Life of Tokens: Reducing Hallucination of Large Vision-Language Models via Visual Information Steering

CustomVideoX: 3D Reference Attention Driven Dynamic Adaptation for Zero-Shot Customized Video Diffusion Transformers

Efficient-vDiT: Efficient Video Diffusion Transformers With Attention Tile

Dual Caption Preference Optimization for Diffusion Models

DreamDPO: Aligning Text-to-3D Generation with Human Preferences via Direct Preference Optimization

Towards Internet-Scale Training For Agents

APE: Faster and Longer Context-Augmented Generation via Adaptive Parallel Encoding

Jakiro: Boosting Speculative Decoding with Decoupled Multi-Head via MoE

Steel-LLM:From Scratch to Open Source -- A Personal Journey in Building a Chinese-Centric LLM

Forbidden Science: Dual-Use AI Challenge Benchmark and Scientific Refusal Tests

Embodied Red Teaming for Auditing Robotic Foundation Models