LLM Archetype Sentinel
8.1
A tool that monitors large language models for the emergence of archetypal patterns ('goblins', etc.) during training or inference, and provides structured reflection prompts to mitigate their influence, leading to more controlled and predictable model behavior.
240h
mvp estimate
8.1
viability grade
1
views
technology stack
Python
Difficult
inspired by
Semantic drift in LLMs addressed with structured reflection