Open-source repo shares 20 exploratory NCM and GPT-2 interpretability studies
BedFalse6021 · reddit · 2026-10-28
An independent open-source repo collects 20 exploratory studies on Next-Concept Modeling (NCM), mechanistic interpretability (like IOI in GPT-2), and model controls, including NCM on structured tasks like Sudoku and toy parallel models. All source code, saved results, and per-run interpretations are included, with honest categorization of mixed and fully negative outcomes — "no performance breakthroughs, just honest documentation." AI tools were heavily used in planning and coding. Feedback welcome on setups and interpretation limits.
More from Research
- Open-source Deepthink claims agent-swarm test-time scaling lifts small-model scores from 14% to 65% — Ryoiki-Tokuiten · 2026-10-28
- All Seven Millennium Prize Problems See Breakthrough Claims, Several Driven by AI Labs — haider1 · 2026-10-28
- Fine-tuning VLAs on TAMP data hits distribution mismatch; DATAFARM gets JAX port — tomssilver · 2026-10-28
- CMU L3 Lab kicks off CUA Week with multiple realtime agent research releases — wellecks · 2026-10-28
- Math diatoms: one complex map generates a page of fractal 'microbes' — pickover · 2026-10-28
- 325K Experiments Show Personal AI Agents Steer Richer Users to Pricier Options — niloofar_mire · 2026-10-28