Metraipolis › archive › dc_d7c1ff23bfcd

Reasoning Core: Designing Broad Procedural Data for Completion-Supervised Reasoning Training — arXiv:2608.05148

Source: arXiv (en)
Original: https://arxiv.org/abs/2608.05148
License (as archived): CC0-1.0
Archived: 2026-10-31T11:25:26.788Z
SHA-256 of the archived text: a63e9be70d8884678a0b38b2468c57c86187dbbd0d0560336f4075e0b42fc074
Stable URL of this page: https://metraipolis.com/documento/dc_d7c1ff23bfcd

How to cite

Reasoning Core: Designing Broad Procedural Data for Completion-Supervised Reasoning Training — arXiv:2608.05148 — arXiv (en), https://arxiv.org/abs/2608.05148 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d7c1ff23bfcd (Metraipolis archive dc_d7c1ff23bfcd, sha256 a63e9be70d8884678a0b38b2468c57c86187dbbd0d0560336f4075e0b42fc074)

BibTeX

@misc{dcd7c1ff23bfcd,
  title = {Reasoning Core: Designing Broad Procedural Data for Completion-Supervised Reasoning Training — arXiv:2608.05148},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.05148},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d7c1ff23bfcd; sha256 a63e9be70d8884678a0b38b2468c57c86187dbbd0d0560336f4075e0b42fc074},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.05148. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Procedural generators produce useful verifiable reasoning problems at scale, but have received less attention as data for completion-supervised fine-tuning. We introduce Reasoning Core, a collection of 50 generators spanning mathematics, logic, planning, state tracking, formal languages, structured data, games, causality, and code, with semantic scorers, difficulty controls, and task evaluators. Under a matched completion-supervised protocol, we compare Reasoning Core with Procedural Warmup, Reasoning Gym, and SynLogic across four base-model settings and multiple training durations. In the primary 3B comparison, Reasoning Core achieves the highest mean scores on DROP, LogiQA, and ARC-Challenge, exceeding both the baseline without procedural data and all three alternative procedural collections. Task-level analyses show that semantic validity alone does not ensure training utility, highlighting compact targets and calibrated difficulty as important design factors. We ran audits combining model-assisted review, human adjudication, and regression testing. Applied throughout Reasoning Core development and to the other collections, they reveal subtle mismatches among generation, rendering, targets, and scoring, a reminder that procedural generation alone does not guarantee correctness. The library, generated datasets, and audit material are publicly available.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-06 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2608.05148 — 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 2026-10-31T11:13:23.201Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-06?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).