Hazy Research open-sources NCM, a coarse-to-fine framework that generates text in parallel per level
andrew_n_carr · x · 2026-10-27
Hazy Research's new Next-Concept Modeling (NCM) framework represents sequences as a hierarchy of discrete codes via a multiscale residual VQ tokenizer, with a transformer predicting the hierarchy coarse-to-fine. Positions within each level generate in parallel, requiring only one forward pass per level. NCM beats or matches strong diffusion baselines at a fraction of the cost. Andrew Carr shared a Claude-assisted explainer video, calling it "almost autoregressive." MIT-licensed repo already has 1.1k stars.
More from Research
- Pessimistic complexity theory hypothesis: faster algorithms must get uglier — Elijah_Meeks · 2026-10-27
- UMAP arcing explained: one giant coordinate dominates neural network activation maps — leland_mcinnes · 2026-10-27
- NeurIPS benchmark: tree models still beat deep learning on 45 tabular datasets — tak3sh8 · 2026-10-27
- Kevin Bryan: maximizing raw accuracy is rarely the right way to train AI assistants — Afinetheorem · 2026-10-27
- Szegedy: a flaw in common training losses went unnoticed for over a decade — ChrSzegedy · 2026-10-27
- Claude research model claims to refute Exponential Time Hypothesis with 14M-token autonomous run — scaling01 · 2026-10-27