Metraipolis › archive › dc_72060e5a5e09
Compositional Self-Improvement: Promises and Pitfalls — arXiv:2611.00452
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.00452
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-03T18:35:12.802Z
- SHA-256 of the archived text
8361f083e0a4d41d1aa6b0692c22c61127cafb53846d9b62686fb85c1eb1476e- Stable URL of this page
https://metraipolis.com/documento/dc_72060e5a5e09
How to cite
Compositional Self-Improvement: Promises and Pitfalls — arXiv:2611.00452 — arXiv (en), https://arxiv.org/abs/2611.00452 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_72060e5a5e09 (Metraipolis archive dc_72060e5a5e09, sha256 8361f083e0a4d41d1aa6b0692c22c61127cafb53846d9b62686fb85c1eb1476e)
BibTeX
@misc{dc72060e5a5e09,
title = {Compositional Self-Improvement: Promises and Pitfalls — arXiv:2611.00452},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.00452},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_72060e5a5e09; sha256 8361f083e0a4d41d1aa6b0692c22c61127cafb53846d9b62686fb85c1eb1476e},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.00452. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Although self-improvement is becoming increasingly important as tasks become more complex, models struggle to generate reliable supervision for harder, out-of-distribution instances. One way to address this challenge is to exploit compositional structure: models often solve simpler subproblems more reliably than full problems, so their subproblem predictions can be composed into supervision for harder ones. We systematically study when this kind of composed supervision supports self-improvement and when it fails. Our controlled experiments identify three failure modes: unreliable subproblem predictions, errors introduced by composition, and limited supervision coverage. Beyond controlled settings, these findings carry over to more realistic coding and reasoning tasks: composed labels are more accurate than direct self-generated labels and improve performance even on compositions longer than those seen during training, while the failure modes remain visible. Further, although these failure modes often keep compositional self-improvement from fully succeeding on its own, a short SFT warmup with compositional self-improvement can bootstrap reinforcement learning on tasks that require composing skills the model already has, particularly when direct exploration rarely succeeds. Together, these findings characterize when compositional self-improvement can provide sufficiently reliable supervision to expand model capabilities. 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/2611.00452 — 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 2026-11-03T18:12:54.422Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).