{
  "id": 4927,
  "url": "https://arxiv.org/abs/2605.04539v3",
  "title": "RLearner-LLM: Balancing Logical Grounding and Fluency in Large Language Models via Hybrid Direct Preference Optimization",
  "summary": "Direct Preference Optimization (DPO), the efficient alternative to PPO-based RLHF, falls short on knowledge-intensive generation: standard preference signals from human annotators or LLM judges exhibit a systematic verbosity bias that rewards fluency over logical correctness. This blindspot leaves a logical alignment gap -- SFT models reach NLI entailment of only 0.05-0.22 despite producing fluent text. We propose RLearner-LLM with Hybrid-DPO: an automated preference pipeline that fuses a DeBERT",
  "authors": "Qiming Bao, Juho Leinonen, Paul Denny, Michael J. Witbrock",
  "category": "research",
  "topics": "bias-fairness,safety-alignment",
  "orgs": null,
  "regions": null,
  "published_at": "2026-05-06T06:36:09.000Z",
  "fetched_at": "2026-07-14T16:31:21.934Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/4927",
  "original_url": "https://arxiv.org/abs/2605.04539v3",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}