Metraipolis › archive › dc_0756edee2d3e

When Updating Stops Being Learning: Rethinking LLM Self-Evolution via learnable information gain — arXiv:2609.36535

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.36535
License (as archived)
CC0-1.0
Archived
2026-10-07T16:56:35.002Z
SHA-256 of the archived text
0df9060970f30b59f8b7280a9455900b247f294caa3823cf169b4c05bfc19372
Stable URL of this page
https://metraipolis.com/documento/dc_0756edee2d3e

How to cite

When Updating Stops Being Learning: Rethinking LLM Self-Evolution via learnable information gain — arXiv:2609.36535 — arXiv (en), https://arxiv.org/abs/2609.36535 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0756edee2d3e (Metraipolis archive dc_0756edee2d3e, sha256 0df9060970f30b59f8b7280a9455900b247f294caa3823cf169b4c05bfc19372)
BibTeX
@misc{dc0756edee2d3e,
  title = {When Updating Stops Being Learning: Rethinking LLM Self-Evolution via learnable information gain — arXiv:2609.36535},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.36535},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0756edee2d3e; sha256 0df9060970f30b59f8b7280a9455900b247f294caa3823cf169b4c05bfc19372},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.36535. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Self-evolution lets large language models (LLMs) improve iteratively using their own generated data, but often suffers from self-evolution degeneration: performance improves, plateaus, then declines. Existing methods address this issue at the component level, targeting either the Questioner or the Solver, and overlook that self-evolution is a tightly coupled system. We propose a holistic framework based on learnable information gain, which measures how much novel, parameterizable information a round provides relative to the previous round. Theoretically, this gain equals the Kullback-Leibler divergence between the two rounds' data distributions plus their entropy change. Practically, it is estimated by fitting a small language model to the previous round and scoring new data via negative log-likelihood. Based on this diagnostic, we propose ATRI (Adaptive Training Regulation via Information-gain), which reweights samples within a round and halts training across rounds when information gain remains low. Experiments on popular datasets demonstrate the superiority of our proposal.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.36535 — 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 2026-10-07T16:27:03.216Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).