Sakana AI’s UnMaskFork uses multiple diffusion LMs to scale inference at test time
SakanaAILabs · x · 2026-07-29
Sakana AI says its ICML-covered work on UnMaskFork explores how to use the collective intelligence of multiple diffusion language models to enable inference-time scaling.
According to the cited Nikkei xTECH piece, the team argues that inference-time scaling will remain useful even as models improve, because the method can keep applying to future generations of diffusion language models. The research is still early, but the article frames it as a foundational theory effort for a still-maturing field.
More from Research
- Meta paper revives two-tower retrieval with LLM semantic encoders — _reachsumit · 2026-07-29
- Kuaishou’s TWICE splits ad conversion learning into two clocks — _reachsumit · 2026-07-29
- Spotify uses LLM-generated hypotheses to build personalized shelves — _reachsumit · 2026-07-29
- LLM-generated CVE labels fail to improve ATT&CK mapping, study finds — CIRCL · 2026-07-29
- Math paper tightens the theory behind diffusion samplers and PF-ODEs — chaumian · 2026-07-29
- Practical local image-to-video workflows for Blender on a GTX 1650 4GB — Haziq12345 · 2026-07-29