RewardBench 1&2

Support Vector Rubrics: Closing the Gap Between Self-Generated and Human Rubrics

arXiv:2606.08077v1 Announce Type: new Abstract: Rubric-based evaluation is a promising paradigm for judging large language model (LLM) outputs, yet self-generated rubrics lag human-annotated criteria on hard instances. We argue this discriminative gap reflects an objective mismatch: self-generated rubrics describe good responses, whereas effective criteria must discriminate between close candidates. To close this gap, we introduce SVR (Support Vector Rubrics), a framework that recasts rubric...