Pinned Loading

1. weight-edit-elicitation weight-edit-elicitation Public

   Can an LLM agent unlock a password-locked model by editing its weights? Code, agent logs and validation results of a six-run case study (preliminary).

   Python

2. sim-eval-awareness sim-eval-awareness Public

   Does an LLM-simulated environment change an AI agent's evaluation awareness or behaviour? Code and experiments for a master's thesis (La Salle).

   Python

3. awesome-ai-safety-tools awesome-ai-safety-tools Public

   A curated list of open-source tools for AI safety: interpretability, evals, red teaming, guardrails, control and training-time safety. When to use each, when not to, and what to use instead.

   TypeScript

4. Haedo-Studio/junsu-claude-plugin Haedo-Studio/junsu-claude-plugin Public

   Junsu plugin for Claude (marketplace, generated from the Junsu monorepo)