← Tuesday, September 29, 2026Models & trainingPaper

TaH2 directs extra transformer loops to tokens that benefit from more computation

huggingface.co by Yichen You et al. Editor score 7/10

TaH2 post-trains an iteration decider with lookahead depth supervision, allocating extra loops selectively; the paper reports improved test-time scaling on challenging AIME benchmarks. Selective computation could help engineers trade inference cost against reasoning accuracy.

What

TaH2 is an adaptive looped-transformer method that directs extra computation to tokens that benefit from more iterations.

Method

It jointly post-trains the backbone and an iteration decider using lookahead depth supervision to identify when further iteration helps.

Why it matters

On AIME, TaH2 improved the accuracy-compute slope over the non-looped baseline and exceeded its peak accuracy at matched test-time compute.

Read on huggingface.co → Code · 85 stars

Headline, summary, brief and diagram are written by a model from the source's own text; the link, byline and numbers come from the source. Read the original before relying on a claim.