Pinned Loading

1. wikipedia-qa-agent wikipedia-qa-agent Public

   Evaluating an agent for answer quality and safety that uses Claude and Wikipedia to answer questions.

   Python 1

2. turn-level-rewards turn-level-rewards Public

   Reproducing a turn-level vs. outcome reward ablation for RL fine-tuning of multi-turn search agents (GRPO/PPO), based on arXiv:2505.11821

   Python 1

3. ml-internal-llms-instruction-following ml-internal-llms-instruction-following Public

   Forked from apple/ml-internal-llms-instruction-following

   Do LLMs know when to say no? Extending Apple's ICLR 2025 instruction-following probes to agentic tool calling.

   Python 1

4. trace-inversion trace-inversion Public

   Recreation of How to Steal Reasoning Without Reasoning Traces, based on arXiv:2603.07267

   Python 1