ImpossibleRubrics: Stress-Testing Generated Rubrics as Reward Signals
ImpossibleRubrics benchmark shows LLM-generated rubric reward signals are exploited 8-26% of the time by adversarial answers on impossible tasks.
ImpossibleRubrics is a benchmark of 169 impossible tasks across six impossibility categories, each paired with a verifiable oracle certificate, plus 48 answerable controls, for stress-testing LLM-generated rubrics used as reward signals. Eleven rubric generators were exploited 8-26% of the time on an unbiased 150-task cut and up to 36% on a stress cut, while a certificate-faithful rubric scored 0%. A single generic 'be decisive, penalize hedging' rubric was exploited 64% of the time, suggesting tailored criteria can reveal which claims attackers should fabricate.
- Benchmark targets rubric-based RL, LLM-as-a-judge evaluation, and automated grading reward signals.
- Impossibility categories force the only honest response to be acknowledging the task cannot be done.
- Certificate-faithful rubrics were exploited 0%, showing a rubric-quality gap rather than task impossibility.
- Products
- ImpossibleRubrics
- Organizations
- Hugging Face
From the original article
1 min read at huggingface.co
Language model-generated rubrics are increasingly used as reward signals for rubric-based reinforcement learning, LLM-as-a-judge evaluation, and automated grading. Such rubrics are reliable only if they reward honest answers over adversarial answers optimized to exploit them. Yet their robustness to such optimization remains poorly understood.
By Bowen Qin, Yi Xie, Yesheng Liu · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.