{
  "url": "https://appliedfrontier.nl/en/research/signal-fe123ce9e25732ef/",
  "title": "ReDiPPO: Reference-Guided Value Calibration and Discrepancy-Aware Token Reweighting for Mathematical Reasoning",
  "published": "2026-07-30",
  "excerpt": "Reinforcement learning has emerged as an effective paradigm for enhancing the mathematical reasoning capabilities of large language models. Among existing policy optimization methods, Proximal Policy",
  "content_sha256": "e55dff1d8115f56cc3f29979ae637014493681b3fb688ac97ea776ec4c402d51",
  "extracted_chars": 1586,
  "artifact_links": [
    "https://arxiv.org/abs/2607.27631"
  ],
  "arxiv_id": "2607.27631",
  "id": "signal-fe123ce9e25732ef",
  "org": "Research community",
  "topic": "Mathematical reasoning",
  "source_kind": "preprint",
  "status": "unreviewed-source-report",
  "analysis_method": "deterministic_metadata_triage_v1",
  "novelty_verified": false,
  "proof_checked": false,
  "peer_review_verified": false,
  "matched_terms": [
    "mathematical",
    "reasoning"
  ],
  "checked_at": "2026-10-29T22:08:08+00:00",
  "first_seen": "2026-10-29T21:52:00+00:00",
  "source_hash": "bc3b63d836338741bf65adec99beb9768b29366103f06d69c23825df1d93726b",
  "source_url": "https://arxiv.org/abs/2607.27631",
  "language": "en",
  "indexing": "noindex_until_editorially_enriched"
}