Popular repositories Loading

1. VEF VEF Public archive

   VEF — A fully interpretable language model built from closed-form statistics with zero gradient descent. Every answer is traceable. Generalization is provable. Trained in 35 minutes on a single GPU.

   Python

2. compute-retrieve compute-retrieve Public

   Compute-Retrieve Axis: Detecting Reasoning Mode from Transformer Hidden States

   Python

3. compounding-fingerprint compounding-fingerprint Public

   Python

4. ouro-chosen-effort ouro-chosen-effort Public

   Code, preregistrations and results for "Chosen Effort in a Looped Language Model: A Preregistered Study of Ouro's Exit Gate"

   Python

5. injected-thought-reports injected-thought-reports Public

   Code, results and paper for "The Injected Word Arrives but Is Not Read": injected-thought reports in Qwen3 (1.7B–14B)

   Python

6. head-start-or-process head-start-or-process Public

   Code, preregistrations and results for "Head Start or Process?": does an experience-acquired abstraction change how a later one is learned? Output-matched learners on Qwen3-1.7B LoRA adapters. Weig…

   Python