Pinned Loading

1. emergent-lying-sales-finetuning emergent-lying-sales-finetuning Public

   If you fine-tune a language model on sales conversations (enthusiastic, persuasive, but never factually wrong) does it start lying on its own?

   Python 1

2. in-context-trajectory-poisoning in-context-trajectory-poisoning Public

   Bypassing LLM-based agent monitors with natural language — no model access, no GPUs, no gibberish. Adapts PAIR for monitor bypass. Apart Research AI Control Hackathon 2026.

   Python 1 1

3. ThoughtGuards ThoughtGuards Public

   A real-time dashboard monitoring AI chain-of-thought traces for manipulative patterns, deception, and reward hacking.

   Python 4

4. Endless-Range/klaviyo-campaign-analysis Endless-Range/klaviyo-campaign-analysis Public

   Analyzing campaigns and subject lines in Klaviyo

   Python 3 1