We introduced ImpossibleRubrics, a benchmark for testing whether generated rubrics reward unsupported answers over honest responses to impossible tasks. Announcement.
Citation
If you found this post useful, please cite it as:
@misc{qin2026impossible,
title = {Impossible Rubrics},
author = {Qin, Bowen},
year = {2026},
howpublished = {Blog post},
url = {https://eyuansu62.github.io/news/2026-09-17-impossible-rubrics/}
}