Follow-up: make each token a high-dimensional embedding refined by diffusion in latent space
SubtleGradient · x · 2026-11-17
Expanding on his idea of specialist diffusion models for code, SubtleGradient proposes making each "token" a kind of high-dimensional embedding, so diffusion can nudge representations around in latent space with progressive refinement toward valid programs.
More from Research
- Zuse: decision models for 27 European languages hit 87.5% accuracy at 27B — bineric · 2026-11-17
- Copenhagen postdoc chance at biomaterials × machine intelligence, deadline Dec 6 — _onionesque · 2026-11-17
- $1.1M Templeton grant funds Global Virtues Atlas of moral beliefs for AI alignment — MohammadAtari90 · 2026-11-17
- Dev proposes swarm of specialist diffusion models for code where invalid syntax is unrepresentable — SubtleGradient · 2026-11-17
- Economists warn peer review is collapsing under 99% AI-generated papers and referee reports — Afinetheorem · 2026-11-17
- Meta's Self-Improving Memory: agents discover their own memory skills and distill them into weights — jaseweston · 2026-11-17