Hand-writing all weights of a 100k-param model beats training it at spotting fake legal citations
AaronBergman18 · x · 2026-11-04
Developer jacsebl showed a model called fable hand-writing the weights of a 100k-parameter neural network to recognize real versus fake legal citations — reportedly one-shotting a full NN that outperforms training it. A striking demo of weight-space programming.
More from Research
- Artificial Analysis ships CritPt v2 grading server, regrades all models with max 1-point shifts — ArtificialAnlys · 2026-11-04
- Vector and Waterloo's Chrysalis makes students teach the AI to finish assignments — VectorInst · 2026-11-04
- Neuro for AI & AI for Neuro workshop at AAAI 2027 opens call for papers, due Nov 20, 2026 — naturecomputes · 2026-11-04
- Multimodal benchmarks should undergo blind filtering by default, researcher argues — cwolferesearch · 2026-11-04
- Larger tokenizers suffer more multi-epoch regret, small vocabs survive repeats — SonglinYang4 · 2026-11-04
- Reddit debates claim that 3-SAT and #3-SAT are solvable in subexponential time — burny_tech · 2026-11-04