Metraipolis › archive › dc_9ccaa4754e8d

From Weeks to Hours: Fast and Principled SFT Data Generation for Reasoning LLMs — arXiv:2611.00554

Source: arXiv (en)
Original: https://arxiv.org/abs/2611.00554
License (as archived): CC0-1.0
Archived: 2026-11-03T11:19:27.025Z
SHA-256 of the archived text: d75b2fb0706971179bfeea107d42de5afee8883b095a6c78a3225005cab4cbbb
Stable URL of this page: https://metraipolis.com/documento/dc_9ccaa4754e8d

How to cite

From Weeks to Hours: Fast and Principled SFT Data Generation for Reasoning LLMs — arXiv:2611.00554 — arXiv (en), https://arxiv.org/abs/2611.00554 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_9ccaa4754e8d (Metraipolis archive dc_9ccaa4754e8d, sha256 d75b2fb0706971179bfeea107d42de5afee8883b095a6c78a3225005cab4cbbb)

BibTeX

@misc{dc9ccaa4754e8d,
  title = {From Weeks to Hours: Fast and Principled SFT Data Generation for Reasoning LLMs — arXiv:2611.00554},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.00554},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_9ccaa4754e8d; sha256 d75b2fb0706971179bfeea107d42de5afee8883b095a6c78a3225005cab4cbbb},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.00554. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Reasoning-oriented post-training enables large language models (LLMs) to solve complex tasks via multi-step inference. Supervised fine-tuning (SFT) on high-quality reasoning traces is a particularly efficient approach, but its effectiveness depends critically on careful data curation. Existing pipelines rely on a costly generate-then-filter paradigm: large pools of long reasoning traces are produced by multiple teacher models and subsequently filtered for difficulty and diversity using additional LLMs, often requiring weeks of computation. In this work, we propose SFTProbe, a fundamentally different approach that bypasses this pipeline by directly predicting the quality of reasoning data. We show that a brief LoRA-based adaptation, combined with evaluating loss on only 5% of each reasoning trace, suffices to estimate both difficulty and diversity. SFTProbe is grounded in a geometric view of post-training, where pretrained LLMs lie in a low-rank, anisotropic loss basin. By probing this basin via low-rank perturbations, we construct compact loss signatures whose clustering captures diversity, while loss at curvature inflection points provides a robust proxy for difficulty. We also provide a simple criterion for teacher selection. Empirically, our approach reduces data curation time from weeks to hours while maintaining or improving performance. Remarkably, on mathematical reasoning, 5k generated examples match the highly filtered 89k-example in OpenThoughts-Math dataset, and 7.5k generated examples outperforms it by 2.5% in accuracy. We further introduce a high-quality physics reasoning dataset, and use it to train a model that outperforms an IPhO 2025 silver-medal-winning model. The code and the dataset are released at this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2611.00554 — 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 2026-11-03T11:02:37.451Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).