Popular repositories Loading

1. Miniserve Miniserve Public

   A lightweight LLM inference and tool-calling agent system with KV cache experiments.

   Python 1

2. agentic-post-training-lab agentic-post-training-lab Public

   Post-training a multi-turn SQLite reconciliation agent with SFT, DPO, and online GRPO, with a focus on action-level credit assignment and error recovery.

   Python 1