APIUp to 28% cheaper than official pricesTry the API →
Hermes
APIModelsCases
Back to News
paperReasoningPost Training

Looped Flows Beats Rival Reasoners on ARC-AGI With Just 7M Parameters

A new recurrent reasoning method mixes denoising with looped hidden states, hitting 58.8% on ARC-AGI-1 and 12.2% on ARC-AGI-2 with just 7M parameters.

Looped Flows Beats Rival Reasoners on ARC-AGI With Just 7M Parameters
Source
AlphaSignal
Published
Author
AlphaSignal Newsroom
Read
1 min read

A new recurrent reasoning method mixes denoising with looped hidden states, hitting 58.8% on ARC-AGI-1 and 12.2% on ARC-AGI-2 with just 7M parameters.

Reporting is indexed from AlphaSignal. Rights remain with the original publisher and cited sources.

Read original report

Next reads

Kyutai's Perturb-Select-Vote Beats Neural Thickets With a Single Bias Vector
AlphaSignal · paper

Kyutai's Perturb-Select-Vote Beats Neural Thickets With a Single Bias Vector

OPST Trains AI Models Using Only Topic Names, No Human Labels
AlphaSignal · paper

OPST Trains AI Models Using Only Topic Names, No Human Labels

TU Delft's FMER Trains Robot Policies 7x Faster on Sparse Tasks
AlphaSignal · paper

TU Delft's FMER Trains Robot Policies 7x Faster on Sparse Tasks