Interpretability researcher cites weight-space evidence against RL convergence claim
voooooogel · x · 2026-09-29
Interpretability researcher voooooogel pushed back on a claim by @zetalyrae: if it were true, RL updates across independent domains should be highly aligned in weight space (i.e., mostly touching similar motivational circuits). He notes this is empirically not the case. A substantive but brief technical exchange lacking the full context of the original claim.
Related event: Researchers Debate Whether AI Models Are More Capable Than They Appear(4 posts)→
More from Research
- Google Research open-sources PaperOrchestra, a multi-agent framework for writing AI papers — google-research · 2026-11-16
- Fable 5.2's Puzzle Skills Look Suspiciously Good, Reciting Outdated MazeBench Solution — patience_cave · 2026-11-16
- Kind ICLR Reviewer Raises Score Despite Unresolved Concerns, Citing the Paper's Value — MarkSchmidtUBC · 2026-11-16
- Hand-editing 0.2% of weights plus 100 CPU-minutes turns an English TTS model into an 18MB Sanskrit one — kastnerkyle · 2026-11-16
- Why chirality matters for AI for Science: enantiomers behave differently in biology — madeofAjala · 2026-11-16
- RSS 2027 submissions open with FAQ clarifying two-stage review and negative-result bar — GeorgiaChal · 2026-11-16