Department of Business, Energy and Industrial Strategy

Popular repositories Loading

1. inspect_ai inspect_ai Public

   Inspect: A framework for large language model evaluations

   Python 2.6k 655

2. inspect_evals inspect_evals Public

   Collection of evals for Inspect AI

   Python 626 401

3. control-arena control-arena Public

   ControlArena is a collection of settings, model organisms and protocols - for running control experiments.

   Python 224 128

4. vllm-lens vllm-lens Public

   Extract residual-stream activations and apply steering vectors (including activation oracles) to any vLLM model during inference.

   Python 120 13

5. hibayes hibayes Public

   Python 53 15

6. sandbox_escape_bench sandbox_escape_bench Public

   Python 39 8