Metraipolis › archive › dc_ebe0582ec956

CORE: Conflict-Oriented Reasoning Elimination for Verifiable Language-Model Search — arXiv:2609.39069

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.39069
License (as archived)
CC0-1.0
Archived
2026-10-07T13:13:39.777Z
SHA-256 of the archived text
7dd8c3e60190e69ceef4a8220a7887b1b0b4c0a9cde5c1601d6df672e58b0baa
Stable URL of this page
https://metraipolis.com/documento/dc_ebe0582ec956

How to cite

CORE: Conflict-Oriented Reasoning Elimination for Verifiable Language-Model Search — arXiv:2609.39069 — arXiv (en), https://arxiv.org/abs/2609.39069 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ebe0582ec956 (Metraipolis archive dc_ebe0582ec956, sha256 7dd8c3e60190e69ceef4a8220a7887b1b0b4c0a9cde5c1601d6df672e58b0baa)
BibTeX
@misc{dcebe0582ec956,
  title = {CORE: Conflict-Oriented Reasoning Elimination for Verifiable Language-Model Search — arXiv:2609.39069},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.39069},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ebe0582ec956; sha256 7dd8c3e60190e69ceef4a8220a7887b1b0b4c0a9cde5c1601d6df672e58b0baa},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.39069. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Test-time reasoning systems often respond to failure by restarting or revising the latest step, even when an earlier decision caused the error. We introduce CORE, a search controller that requests a certified conflict core from a verifier, backjumps to the latest decision in that core, and caches the conflict to avoid repeating it. Under sound verification, finite branching and depth, and exhaustive proposals, the uncapped search is complete and never prunes a valid solution. On 2,000 planted graph-coloring instances with matched proposals and an exact verifier, CORE reduces median verifier calls by 39.8% at 30 variables and 35.0% at 36 variables relative to chronological repair; caching further improves on backjumping alone. Across five reasoning tasks, CORE achieves 75.9% mean success with Qwen2.5-7B-Instruct and 84.2% with Qwen3-8B, compared with 72.5% and 81.8% for Tree of Thoughts. It also uses fewer verifier calls and generated tokens on both backbones. These results show the value of using certified failure explanations to direct language-model search.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-01 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.39069 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T12:44:51.790Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-10-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).