← All threads

Latent Reasoning Estimation

Methods for estimating and evaluating the amount of sequential computation that language models can perform internally before producing language output.

2 papers

Where this stands

The written synthesis of this thread is for subscribers. Subscribe.