Interference as a testable prediction: paper validates theory of distributed neural representations
burny_tech · x · 2026-11-07
Researcher thomasfel calls an upcoming paper "very important": our theory of neural representation (overcomplete, distributed) makes few testable predictions, and interference is one of them. This paper checks that prediction rigorously — "a beautiful piece of science." A cited note adds that a new version is being prepared for ICLR, adding more models and random-seed experiments controlling for non-geometric explanations, with follow-up work expanding beyond the three current settings.
More from Research
- François Fleuret asks: after a loss spike and recovery, is final loss worse, same, or better? — francoisfleuret · 2026-11-08
- Z is Diophantine in Q proof slashes bounds: degree drops from 2↑↑7 to 4 — anderssandberg · 2026-11-08
- In vivo CRISPR base editing of ANGPTL3 shows durable lipid lowering, heading to trials — EricTopol · 2026-11-08
- Testing ICE Constrained Decoding: Enforce Constraints at Generation, Not After — manvigupta03 · 2026-11-08
- Near-identical neural net weights still diverge on ~3% of inputs after training — bravo_abad · 2026-11-08
- Removing negative-advantage samples drops robot zip-tie success from 86% to 60% — micoolcho · 2026-11-08