IAAIFRM
Flow Reasoning Models: Turning Flows Into Efficient Recurrent Reasoners
Flow Reasoning Models turn a non-causal flow denoiser into a recurrent reasoner by feeding its clean-solution prediction back as a conditioning state. Fixed-Point Forcing trains the model on states produced by its own recurrent rollouts, enabling deeper iterative refinement while retaining the standard flow-matching objective.
Inside the method
See paper for architecture
Simplified conceptual schematic. Consult the paper for the complete architecture.
- Recurrence family
- Core recurrence
- Depth control
- See paper
- KV / state strategy
- See paper
Reading note
Community-submitted work. Consult the linked paper for architecture details and experimental conditions.
Sources checked 2026-09-26. This catalog does not imply independent reproduction.
Related works
Cite this work
@misc{arxiv2606291502026,
title = {Flow Reasoning Models: Turning Flows Into Efficient Recurrent Reasoners},
author = {Alec Helbling and Andrey Bryutkin and Mauro Martino and Duen Horng Chau and Nima Dehmamy and Hendrik Strobelt},
year = {2026},
eprint = {2606.29150},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2606.29150}
}