Danqi Chen

Goedel-Prover-V2: Scaling Formal Theorem Proving with Scaffolded Data Synthesis and Self-Correction

Retaining by Doing: The Role of On-Policy Data in Mitigating Forgetting

Language Models That Think, Chat Better

Lost in the Maze: Overcoming Context Limitations in Long-Horizon Information-Seeking

Extracting Rule-based Descriptions of Attention Features in Transformers

LongProc: Benchmarking Long-Context Language Models on Long Procedural Generation

BRIGHT: A Realistic and Challenging Benchmark for Reasoning-Intensive Retrieval

Unintentional Unalignment: Likelihood Displacement in Direct Preference Optimization

Goedel-Prover: A Frontier Model for Open-Source Automated Theorem Proving

Precise Information Control in Long-Form Text Generation

SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal

Metadata Conditioning Accelerates Language Model Pre-training

Organize the Web: Constructing Domains Enhances Pre-Training Data Curation

HELMET: How to Evaluate Long-context Models Effectively and Thoroughly

The Surprising Effectiveness of Negative Reinforcement in LLM Reasoning

How to Train Long-Context Language Models (Effectively)

Continual Memorization of Factoids in Large Language Models

Fantastic Copyrighted Beasts and How (Not) to Generate Them

Certifiably Robust RAG against Retrieval Corruption Attacks