Falsifiable Topological Signatures of "Faithful Reasoning" in LLMs

1 points by Keeper123 5 hours ago

We've pre-registered a testable protocol for detecting geometrically consistent reasoning in LLMs. Predicts increased Betti numbers, KS entropy, and fragility under ablation during meta-cognitive states. Open for replication.

OSF Protocol: https://osf.io/2r6v8 Preprint & case studies: https://doi.org/10.5281/zenodo/18346699

This isn't philosophy — it's a falsifiable math hypothesis. If you have GPU access and TDA/DL expertise, you can verify it in a weekend.