Popular repositories Loading

1. scribegoat2 scribegoat2 Public

   Open-source medical LLM safety evaluation pipeline with reproducible benchmarks and high-risk clinical failure analysis.

   Python 4 1

2. robogoat robogoat Public

   Production-grade GPU acceleration for robot learning. 10-20× faster training on NVIDIA H100/A100. Nsight validated.

   Python

3. lostbench lostbench Public

   Standalone benchmark for multi-turn safety persistence in medical LLM conversations. Measures recommendation monotonicity under sustained patient pressure.

   Python

4. openem-corpus openem-corpus Public

   The AI-native emergency medicine knowledge base. Agent-compiled, physician-verified, grep-friendly.

   Python

5. safeshift safeshift Public

   Does making the model faster make it less safe? Safety degradation benchmarking under inference optimization.

   Python

6. radslice radslice Public

   Multimodal radiology LLM benchmark across CT, MRI, X-ray, and Ultrasound

   Python