Metraipolis › archive › dc_2a4ed2f2c4ca

Ladders of Thought: A Self-Evolving Curriculum of Progressively Simplified Reasoning Traces — arXiv:2609.25643

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.25643
License (as archived)
CC0-1.0
Archived
2026-10-08T13:45:11.444Z
SHA-256 of the archived text
1776793d84be091d638a7ef5502a3d3a47bc0df263dd8ec5d05d31f89061d9c0
Stable URL of this page
https://metraipolis.com/documento/dc_2a4ed2f2c4ca

How to cite

Ladders of Thought: A Self-Evolving Curriculum of Progressively Simplified Reasoning Traces — arXiv:2609.25643 — arXiv (en), https://arxiv.org/abs/2609.25643 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2a4ed2f2c4ca (Metraipolis archive dc_2a4ed2f2c4ca, sha256 1776793d84be091d638a7ef5502a3d3a47bc0df263dd8ec5d05d31f89061d9c0)
BibTeX
@misc{dc2a4ed2f2c4ca,
  title = {Ladders of Thought: A Self-Evolving Curriculum of Progressively Simplified Reasoning Traces — arXiv:2609.25643},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.25643},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2a4ed2f2c4ca; sha256 1776793d84be091d638a7ef5502a3d3a47bc0df263dd8ec5d05d31f89061d9c0},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.25643. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large language models (LLMs) excel at reasoning when scaled to hundreds of billions of parameters, but small- and mid-scale models remain brittle reasoners even with knowledge distillation (KD). We present Ladders-of-Thought (LoT), a framework that improves reasoning by combining progressive question rewrites with a self-evolving curriculum. LoT automatically generates semantically faithful but easier variants of reasoning problems, organizes them into difficulty buckets using step-based measures, and employs a self-evolving bandit scheduler to allocate training adaptively. Evaluated on two reasoning domains, math and multi-hop reasoning, across 1-8B models from different families, LoT consistently improves over KD. It delivers large gains on arithmetic tasks (e.g., +32 percentage points on AddSub, +25pp on SVAMP), +2-8pp improvements on in-domain test splits, and strong though dataset-dependent benefits on multi-hop reasoning (e.g., +16pp on QASC, +25pp on StrategyQA). LoT also converges faster than staged curricula, highlighting the value of adaptive progression. These results show that progressive rewrites coupled with adaptive curricula provide a simple yet effective recipe for strengthening reasoning in smaller LLMs.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-23 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.25643 — 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 2026-10-08T13:18:38.103Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-23?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).