Dicklesworthstone/llm_introspective_compression_and_metacognition

A novel approach for transformer model introspection that enables saving, compressing, and manipulating internal thought states for advanced capabilities like reasoning backtracking, latent thought optimization, and metacognitive control.

14 best points · 11 comments · 1 Hacker News post · 35 GitHub stars · View on GitHub

Hacker News discussions

• Real-Time Introspective Compression for Transformers (14 points, 11 comments, by eigenvalue, 2025-04-02)