Preprint
Aug 2026
On the Robustness of LLMs'Internal Representation of Code Correctness
This work studies an internal signal of code correctness that is able to judge candidate solutions better than the model's token-level or stated confidence, leaving open an important question: whether it reflects a robust property of the model or an artifact of that choice.
Francisco Ribeiro, Sohaila Abdulsattar, R. Gonzalez et al.
· 1 citation