2025-12-02 | Reproducing Neo et al.

Goal: Reproduce Neo et al. 2024

Summary: Fix Tokenizer to use left truncation and padding to keep max activating prompt right context; reach out to AI Safety folks

Work sessions

In	Out
14:45	15:50
22:45	23:50

Reading List

Added Rohin Shah's Google DeepMind: An Approach to Technical AGI Safety and Security to reading list