Junyang Lin

Language Confusion Gate: Language-Aware Decoding Through Model Self-Distillation

VideoAgentTrek: Computer-Use Pretraining from Unlabeled Videos

PlotCraft: Pushing the Limits of LLMs for Complex and Interactive Data Visualization

Omni-Captioner: Data Pipeline, Models, and Benchmark for Omni Detailed Perception

From Narrow to Panoramic Vision: Attention-Guided Cold-Start Reshapes Multimodal Reasoning

MegaFlow: Large-Scale Distributed Orchestration System for the Agentic Era

A$^2$Search: Ambiguity-Aware Question Answering with Reinforcement Learning

Revisiting Multimodal Positional Encoding in Vision–Language Models

Overcoming Joint Intractability with Lossless Hierarchical Speculative Decoding

RefineX: Learning to Refine Pre-training Data at Scale from Expert-Guided Programs

SWE-RM: Execution-free Feedback for Software Engineering Agents

Winning the Pruning Gamble: A Unified Approach to Joint Sample and Token Pruning for Efficient Supervised Fine-Tuning

TriSpec: Ternary Speculative Decoding via Lightweight Proxy Verification

WavReward: Spoken Dialogue Models With Generalist Reward Evaluators

WorldPM: Understanding Scaling Patterns in Human Preference Modeling

C$^3$-Bench: Evaluating and Achieving Controllable Code Completion in Code LLM

One Model to Critique Them All: Rewarding Agentic Tool-Use via Efficient Reasoning

Understanding DeepResearch via Reports

RefCritic: Training Long Chain-of-Thought Critic Models with Refinement Feedback

Beyond Turn Limits: Training Deep Search Agents with Dynamic Context Window

ReAlign: Safety-Aligning Reasoning Models with Verifier-Guided Reinforcement Learning

Reinforcement Learning for Symbolic Graphics Code with Visual Feedback

Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free

Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning

Chain of Execution Supervision Promotes General Reasoning in Large Language Models

Parallel Scaling Law for Language Models

OpenHands: An Open Platform for AI Software Developers as Generalist Agents

MARGE: Improving Math Reasoning with Guided Exploration

Teaching Language Models to Reason with Tools

Self-Evolving Critique Abilities in Large Language Models

Efficient Long Context Fine-tuning with Chunk Flow

A Spark of Vision-Language Intelligence: 2-Dimensional Autoregressive Transformer for Efficient Finegrained Image Generation

CateKV: On Sequential Consistency for Long-Context LLM Inference Acceleration

Synthesizing Software Engineering Data in a Test-Driven Manner

DataMan: Data Manager for Pre-training Large Language Models

CARE: Decoding-Time Safety Alignment via Rollback and Introspection Intervention

Rotated Runtime Smooth: Training-Free Activation Smoother for accurate INT4 inference

Disentangling Reasoning Tokens and Boilerplate Tokens For Language Model Fine-tuning

Analyzing and Mitigating Inconsistency in Discrete Audio Tokens for Neural Codec Language Models

Rethinking Data Selection at Scale: Random Selection is Almost All You Need

Language Models can Self-Lengthen to Generate Long Texts

Aligning Large Language Models via Self-Steering Optimization