Reposted by Maxime Peyrard
🔍 Observable Patterns Are Not Explanations: A Causal-Geometric Analysis of Latent Reasoning Models
Latent thoughts are hidden computation, not explanations. Causality should come first, then we can interpret the geometry of the revealed behavior-influential spaces.
arxiv.org/abs/2606.12689
arxiv.org
Observable Patterns Are Not Explanations: A Causal-Geometric Analysis of Latent Reasoning Models
Latent reasoning models (LRMs) replace explicit chain-of-thought with continuous thoughts. Recent work treats observable latent-state patterns, such as BFS-like frontiers and decodable arithmetic comp...
