{
  "schema_version": "1.1",
  "id": "s13:https://arxiv.org/abs/2610.42128v1",
  "slug": "2610-42128v1-0sepil9",
  "url": "https://feed7.dev/p/2610-42128v1-0sepil9",
  "title": "DecisionRank: On-Policy Distillation for Decision-Model Rerankers",
  "why_included": "DecisionRank transfers a generative ranker's preferences into a score-based reranker through on-policy distillation. The paper reports competitive retrieval results with substantially faster ranking at 8B parameters.",
  "summary": "DecisionRank sorts a decision model's document scores into a ranking, then uses a generative teacher's conditional feedback to supervise those scores. The paper reports that **1.7B** DecisionRank beats its **27B** teacher on TREC DL19–23; its 8B model runs **24.5×** faster on TREC DL and **16.7×** faster on BRIGHT than a generative listwise counterpart.",
  "practical_implication": "Builders using retrieval in coding agents could evaluate a score-based reranker when ranking latency matters. The method combines item-space on-policy distillation with offline ranking supervision, and the paper reports results on TREC DL, BEIR and BRIGHT.",
  "agent_context": "DecisionRank sorts a decision model's document scores into a ranking, then uses a generative teacher's conditional feedback to supervise those scores. The paper reports that **1.7B** DecisionRank beats its **27B** teacher on TREC DL19–23; its 8B model runs **24.5×** faster on TREC DL and **16.7×** faster on BRIGHT than a generative listwise counterpart.\n\nBuilders using retrieval in coding agents could evaluate a score-based reranker when ranking latency matters. The method combines item-space on-policy distillation with offline ranking supervision, and the paper reports results on TREC DL, BEIR and BRIGHT.\n\nThese are benchmark results, not evidence about end-to-end agent quality or latency in a specific retrieval stack. The speed comparisons apply to the named benchmarks and generative listwise counterpart.",
  "source": {
    "name": "arXiv",
    "url": "https://arxiv.org/abs/2610.42128v1",
    "published_at": "2026-10-30T17:56:58.000Z"
  },
  "source_class": "blog_post",
  "content_type": "Paper",
  "layer": "model",
  "domains": [
    "research",
    "data"
  ],
  "topics": [
    "retrieval"
  ],
  "verification": {
    "status": "needs_review",
    "label": "Needs Review",
    "method": "unverified",
    "verified_at": null
  },
  "uncertainty": [
    "These are benchmark results, not evidence about end-to-end agent quality or latency in a specific retrieval stack. The speed comparisons apply to the named benchmarks and generative listwise counterpart."
  ],
  "connected_context": {
    "meaning": "DecisionRank offers a fast score-based reranker trained from generative feedback, giving retrieval systems a way to reduce ranking latency without replacing their candidate-generation stage. DecisionRAG explores the opposite architectural move by navigating a corpus without a first-stage retriever. The benchmark results motivate comparing these routes, but do not establish which improves an end-to-end agent.",
    "corpus_size": 974,
    "generated_at": "2026-11-02T11:02:31.143Z",
    "connections": [
      {
        "title": "Retriever-Free Retrieval: Navigating Documents with Decision Models",
        "source_name": "arXiv",
        "source_url": "https://arxiv.org/abs/2610.35868v1",
        "feed7_url": "https://feed7.dev/p/2610-35868v1-1uzxrmk",
        "reason": "DecisionRank speeds ranking after candidates are available; DecisionRAG removes the first-stage retriever and navigates document branches, so they change different stages of retrieval."
      },
      {
        "title": "Yalda-Embedding: Open Development of a Competitive Persian Retrieval Model",
        "source_name": "arXiv",
        "source_url": "https://arxiv.org/abs/2610.32933v1",
        "feed7_url": "https://feed7.dev/p/2610-32933v1-0kbcu6k",
        "reason": "Yalda provides a compact retrieval model, while DecisionRank focuses on reranking. They could serve complementary stages, though the supplied evidence does not evaluate them together."
      }
    ]
  },
  "lifecycle": "Current",
  "published_at": "2026-10-30T17:56:58.000Z",
  "modified_at": "2026-10-30T17:56:58.000Z",
  "supersedes": [],
  "expires_at": null,
  "formats": {
    "html": "https://feed7.dev/p/2610-42128v1-0sepil9",
    "json": "https://feed7.dev/p/2610-42128v1-0sepil9.json",
    "markdown": "https://feed7.dev/p/2610-42128v1-0sepil9.md"
  }
}