Yeyun Gong

Beyond Pass@ 1: Self-Play with Variational Problem Synthesis Sustains RLVR

Beyond Length: Quantifying Long-Range Information for Long-Context LLM Pretraining Data

Mixture of Neuron Experts

Training Matryoshka Mixture-of-Experts for Elastic Inference-Time Expert Utilization

Recycling Pretrained Checkpoints: Orthogonal Growth of Mixture-of-Experts for Efficient Large Language Model Pre-Training

TL;DR: Too Long, Do Re-weighting for Efficient LLM Reasoning Compression

Learning from the Best, Differently: A Diversity-Driven Rethinking on Data Selection

Behind RoPE: How Does Causal Mask Encode Positional Information?

PeRL: Permutation-Enhanced Reinforcement Learning for Interleaved Vision-Language Reasoning

Automated Proof Generation for Rust Code via Self-Evolution

Routing Mamba: Scaling State Space Models with Mixture-of-Experts Projection

Optimizing Large Language Model Training Using FP4 Quantization

Alchemy: Amplifying Theorem-Proving Capability Through Symbolic Mutation

Integrative Decoding: Improving Factuality via Implicit Self-consistency

SwS: Self-aware Weakness-driven Problem Synthesis in Reinforcement Learning for LLM Reasoning

Overcoming Vocabulary Mismatch: Vocabulary-agnostic Teacher Guided Language Modeling