Open-source PyPI package for benchmarking LLM reasoning over scientific literature, with hybrid retrieval, evaluation harnesses, and agentic research pipelines.
Jan 1, 2026