Popular repositories Loading
-
inspect_ai
inspect_ai PublicInspect: A framework for large language model evaluations
-
-
control-arena
control-arena PublicControlArena is a collection of settings, model organisms and protocols - for running control experiments.
-
Repositories
-
- control-arena Public
ControlArena is a collection of settings, model organisms and protocols - for running control experiments.
- dsit-dvs-register Public
- reward-hacking-misalignment Public
Reproducing "Natural Emergent Misalignment from Reward Hacking" (MacDiarmid et al., Anthropic 2025) with open-source models. Includes reward-hackable RL environments, misalignment evaluations, training configs, and evaluation scripts. Models trained on OLMo (7B, 32B) and GPT-OSS (20B, 120B).
- vllm-lens Public
Extract residual-stream activations and apply steering vectors (including activation oracles) to any vLLM model during inference.
- ukets-maritime Public
- transparency-db-admin-portal Public
People
This organization has no public members. You must be a member to see who’s a part of this organization.
Most used topics
Loading…