Flow Reasoning Models repo out: stable attractors solve Sudoku-Extreme with 44× fewer FLOPs
alec_helbling · x · 2026-09-03
Alec Helbling and colleagues released the repository for their Flow Reasoning Models (FRMs) paper, with implementation code coming soon.
Key claim: FRMs shape correct solutions into stable attractors, reaching EqR's peak Sudoku-Extreme solve rate with 44× fewer inference FLOPs. Co-authors include Mauro Martino, Duen Horng (Polo) Chau, Nima Dehmamy, and Hendrik Strobelt.
Related event: Flow Reasoning Models Turn Flows Into Efficient Recurrent Reasoners(6 posts)→
More from Research
- Stanford's LILIE loss cuts text-to-image gFD by 14 points over pixel-only training — StanfordUniversity · 2026-10-21
- ETH's HumanoidPose recovers full-body robot pose from plain video, no internal sensors — ETHZurich · 2026-10-21
- One L-system rule grows a tree, fern, lightning, river and lungs in a 45s code film — prasenx · 2026-10-21
- MOSS-Image's text encoder is multimodal and accepts image inputs — RandumbRedditor1000 · 2026-10-21
- Skip the open access fees: NIH-funded papers must go on PMC with no embargo — anshulkundaje · 2026-10-21
- Creed: LLM code auditing infers developer beliefs to find semantic security flaws — mboehme_ · 2026-10-21