Metraipolis › archive › dc_b35596b91567

Highlight-Then-Summarize: Learning to Compress Evidence for Long-Context Understanding — arXiv:2609.31382

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.31382
License (as archived)
CC0-1.0
Archived
2026-10-08T01:39:54.206Z
SHA-256 of the archived text
656dcd9ddb925c5c03cd67eb6a2d1df2f02a7c8261f5eb980421694703c6b8c2
Stable URL of this page
https://metraipolis.com/documento/dc_b35596b91567

How to cite

Highlight-Then-Summarize: Learning to Compress Evidence for Long-Context Understanding — arXiv:2609.31382 — arXiv (en), https://arxiv.org/abs/2609.31382 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_b35596b91567 (Metraipolis archive dc_b35596b91567, sha256 656dcd9ddb925c5c03cd67eb6a2d1df2f02a7c8261f5eb980421694703c6b8c2)
BibTeX
@misc{dcb35596b91567,
  title = {Highlight-Then-Summarize: Learning to Compress Evidence for Long-Context Understanding — arXiv:2609.31382},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.31382},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_b35596b91567; sha256 656dcd9ddb925c5c03cd67eb6a2d1df2f02a7c8261f5eb980421694703c6b8c2},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.31382. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Long-context understanding requires large language models (LLMs) to reason over lengthy documents, conversations, and code, yet task-relevant evidence is often sparse and scattered amid substantial irrelevant and redundant content. We propose Highlight-Then-Summarize (H2S), a compress-then-reason paradigm that first identifies source-grounded, question-relevant evidence and then integrates it into a compact, question-conditioned summary before producing the final answer. To train this behavior, we construct H2S-Dataset, comprising 6,647 examples from 11 benchmark families with an average context length of 43.9K tokens, and introduce H2S-RL, which provides process-level rewards for evidence selection and summary construction in addition to final-answer correctness. We evaluate on H2S-Bench, a seven-task long-context suite. Under a shared 128K input and 4K output budget, H2S-14B achieves an average score of 32.60, outperforming Qwen3.8-27B by 10.17 points and obtaining the strongest overall result among the evaluated open-source models. H2S-14B also achieves the highest Evidence-Summary Quality score and retains 97.1% of its 16K-budget performance with only a 4K output budget. These results show that explicitly selecting and integrating evidence improves long-context reasoning while enabling more compact generation.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-28 in cs.AI).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.31382 — 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 2026-10-08T01:12:14.583Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).