Xunliang Cai

Expanding Reasoning Potential in Foundation Model by Learning Diverse Chains of Thought Patterns

AMemGym: Interactive Memory Benchmarking for Assistants in Long-Horizon Conversations

R-Horizon: How Far Can Your Large Reasoning Model Really Go in Breadth and Depth?

VitaBench: Benchmarking LLM Agents with Versatile Interactive Tasks in Real-world Applications

Audio Turing Test: Benchmarking the Human-likeness of Large Language Model-based Text-to-Speech Systems in Chinese

Fast Catch-Up, Late Switching: Optimal Batch Size Scheduling via Functional Scaling Laws

Towards Self-Robust LLMs: Intrinsic Prompt Noise Resistance via CoIPO

Autoformalizer with Tool Feedback

CATArena: Evaluation of LLM Agents Through Iterative Tournament Competitions

GradPower: Powering Gradients for Faster Language Model Pre-Training

Bridging Pairwise and Pointwise GRMs: Preference-Aware Reward Mechanism with Dynamic Rubric Adaptation

RepSpec: Structural Re-parameterized Draft Model Training for Speculative Decoding

CoreCodeBench: A Configurable Multi-Scenario Repository-Level Benchmark

InfiniteTalk: Audio-driven Video Generation for Sparse-Frame Video Dubbing

OneCAT: Decoder-Only Auto-Regressive Model for Unified Understanding and Generation

Libra: Assessing and Improving Reward Model by Learning to Think

FuseNorm: Achieving the Best of Both Worlds from PreNorm and PostNorm

C2T: Classifier-based Token Tree Construction in Speculative Decoding

MUA-RL: Multi-turn User-interacting Agent Reinforcement Learning for agentic tool use

CodePlot-CoT: Mathematical Visual Reasoning by Thinking with Code-Driven Images

MoTE: Mixture of Ternary Experts for Memory-efficient Large Multimodal Models

Meeseeks: A Feedback-Driven, Iterative Self-Correction Benchmark evaluating LLMs' Instruction Following Capability

NeedleInATable: Exploring Long-Context Capability of Large Language Models towards Long-Structured Tables

PrefixKV: Adaptive Prefix KV Cache is What Vision Instruction-Following Models Need for Efficient Generation

Let Them Talk: Audio-Driven Multi-Person Conversational Video Generation

Earlier Tokens Contribute More: Learning Direct Preference Optimization From Temporal Decay Perspective

Revisit Self-Debugging with Self-Generated Tests for Code Generation

AgentRefine: Enhancing Agent Generalization through Refinement Tuning

Length Desensitization in Direct Preference Optimization

From Scaling Law to Sub-Scaling Law: Understanding the Diminishing Returns of Larger Models

GameInstruct: Teaching Machines to Reason via Chameleon Game

Co-evolved Self-Critique: Enhancing Large Language Models with Self-Generated Data

RepoFixEval: A Repository-Level Program Repair Benchmark From Issue Discovering to Bug Fixing