Metraipolis › archive › dc_dd05914e2e36
Damage Predicts Recovery: When Calibration Data Matters in Compressing Financial LLMs — arXiv:2609.26241
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.26241
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T14:36:41.345Z
- SHA-256 of the archived text
b00708b8ac33683d7b6f13e9670a0baf306fce158d032f507c41f3fe17da267b- Stable URL of this page
https://metraipolis.com/documento/dc_dd05914e2e36
How to cite
Damage Predicts Recovery: When Calibration Data Matters in Compressing Financial LLMs — arXiv:2609.26241 — arXiv (en), https://arxiv.org/abs/2609.26241 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_dd05914e2e36 (Metraipolis archive dc_dd05914e2e36, sha256 b00708b8ac33683d7b6f13e9670a0baf306fce158d032f507c41f3fe17da267b)
BibTeX
@misc{dcdd05914e2e36,
title = {Damage Predicts Recovery: When Calibration Data Matters in Compressing Financial LLMs — arXiv:2609.26241},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.26241},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_dd05914e2e36; sha256 b00708b8ac33683d7b6f13e9670a0baf306fce158d032f507c41f3fe17da267b},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.26241. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Post-training quantization and pruning rely on a small calibration corpus. Whether specialized domains such as finance require domain-matched calibration data remains unsettled. We argue that the answer depends on the task-level damage caused by compression rather than on domain mismatch. If compression preserves the target capability, changing the calibration corpus has little effect. If compression causes large losses, task-formatted calibration can recover part of the loss. We test this hypothesis across two model families, six compression configurations, three token-matched calibration corpora, and ten financial classification and numerical question-answering tasks. The results support this hypothesis. Quantization largely preserves task performance, and calibration choice has little effect in this case. Pruning reduces numerical QA accuracy by over 40 points. In these damaged settings, another generic corpus does not help, while FinMix, a mixture of financial task examples, recovers a large part of the loss. The link between damage and recovery holds across model families and scales. These findings support a practical rule. Measure task-specific compression damage first, and construct specialized calibration data only when the damage is large. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-23 in cs.CL). Categorie arXiv: Computation and Language (cs.CL) ; Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.26241 — 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 2026-10-08T13:46:15.731Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-23?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).