Daniel Kang·Aug 3Noisy Data Breaks RLVRReinforcement learning with verifiable rewards (RLVR) is a widely used post-training paradigm to improve the reasoning capabilities of…
Daniel Kang·Jul 27Pervasive Annotation Errors Break Text-to-SQL Benchmarks and LeaderboardsText-to-SQL is increasingly important in data analytics and database-driven applications. It reduces the effort required for experienced…
Daniel Kang·Jul 20Provable Generalization Bounds for RLVR at the Billion-Parameter ScaleReinforcement learning with verifiable rewards (RLVR) has become a critical tool for modern large language model (LLM) development. RLVR is…
