U-Space
A research framework for revealing when and why uncertainty arises in language models. It combines hidden-state directions with the average entropy of reasoning tokens to provide interpretable uncertainty signals.
Timeline (last 90 days)
The U-Space framework is used to reveal when and why uncertainty arises in language models, and can provide an interpretable model certainty signal through a single forward inference.
The U-Space base does not have cross-model transferability, and each model needs to be built separately once.
The U-Space code is currently research code and a reproduction pipeline, and has not yet become a pip package.
By cloning https://github.com/s2labres/U-Space and running ./run_smoke.sh, you can run the end-to-end pipeline for Qwen3.5-4B on 10 TriviaQA questions (single GPU).
All Facts (4)
The U-Space framework is used to reveal when and why uncertainty arises in language models, and can provide an interpretable model certainty signal through a single forward inference.
50%UnverifiedThe U-Space base does not have cross-model transferability, and each model needs to be built separately once.
50%UnverifiedThe U-Space code is currently research code and a reproduction pipeline, and has not yet become a pip package.
50%UnverifiedBy cloning https://github.com/s2labres/U-Space and running ./run_smoke.sh, you can run the end-to-end pipeline for Qwen3.5-4B on 10 TriviaQA questions (single GPU).
50%