AI 연구 논문 데일리

Qwen3 Technical Report

GuardReasoner-VL: Safeguarding VLMs via Reinforced Reasoning

MMLongBench: Benchmarking Long-Context Vision-Language Models Effectively and Thoroughly

Visual Planning: Let's Think Only with Images

Group Think: Multiple Concurrent Reasoning Agents Collaborating at Token Level Granularity

Simple Semi-supervised Knowledge Distillation from Vision-Language Models via texttt{D}ual-texttt{H}ead texttt{O}ptimization

Mergenetic: a Simple Evolutionary Model Merging Library

Multi-Token Prediction Needs Registers

MPS-Prover: Advancing Stepwise Theorem Proving by Multi-Perspective Search and Data Curation

Scaling Reasoning can Improve Factuality in Large Language Models

MatTools: Benchmarking Large Language Models for Materials Science Tools

Improving Assembly Code Performance with Large Language Models via Reinforcement Learning

Humans expect rationality and cooperation from LLM opponents in strategic games

InstanceGen: Image Generation with Instance-level Instructions

Unifying Segment Anything in Microscopy with Multimodal Large Language Model

GIE-Bench: Towards Grounded Evaluation for Text-Guided Image Editing

Learning Dense Hand Contact Estimation from Imbalanced Data

CheXGenBench: A Unified Benchmark For Fidelity, Privacy and Utility of Synthetic Chest Radiographs

From Trade-off to Synergy: A Versatile Symbiotic Watermarking Framework for Large Language Models

Improving Inference-Time Optimisation for Vocal Effects Style Transfer with a Gaussian Prior