{
  "id": 7517,
  "url": "https://arxiv.org/abs/2603.08035v1",
  "title": "CDRRM: Contrast-Driven Rubric Generation for Reliable and Interpretable Reward Modeling",
  "summary": "Reward modeling is essential for aligning Large Language Models(LLMs) with human preferences, yet conventional reward models suffer from poor interpretability and heavy reliance on costly expert annotations. While recent rubric-based approaches enhance evaluation transparency, they lack systematic quality control, yielding noisy and redundant criteria, failing to mitigate persistent biases (e.g., verbosity, position) in LLM evaluators, and creating a scalability-reliability trade-off. To address",
  "authors": "Dengcan Liu, Fengkai Yang, Xiaohan Wang, Shurui Yan, Jiajun Chai, Jiahao Li et al.",
  "category": "research",
  "topics": "safety-alignment,transparency",
  "orgs": null,
  "regions": null,
  "published_at": "2026-03-09T07:15:23.000Z",
  "fetched_at": "2026-07-14T16:33:16.669Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/7517",
  "original_url": "https://arxiv.org/abs/2603.08035v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}