Metraipolis › archive › dc_25b4bef3d570
SciDocBench: A Workflow-Centered Benchmark and Data Pipeline for Scientific Document Understanding — arXiv:2609.05141
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.05141
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T22:21:16.620Z
- SHA-256 of the archived text
3538618eb3e13e7961fb8a2b62caf23f31cdcca8a7b3c40306fdc85bb2314999- Stable URL of this page
https://metraipolis.com/documento/dc_25b4bef3d570
How to cite
SciDocBench: A Workflow-Centered Benchmark and Data Pipeline for Scientific Document Understanding — arXiv:2609.05141 — arXiv (en), https://arxiv.org/abs/2609.05141 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_25b4bef3d570 (Metraipolis archive dc_25b4bef3d570, sha256 3538618eb3e13e7961fb8a2b62caf23f31cdcca8a7b3c40306fdc85bb2314999)
BibTeX
@misc{dc25b4bef3d570,
title = {SciDocBench: A Workflow-Centered Benchmark and Data Pipeline for Scientific Document Understanding — arXiv:2609.05141},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.05141},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_25b4bef3d570; sha256 3538618eb3e13e7961fb8a2b62caf23f31cdcca8a7b3c40306fdc85bb2314999},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.05141. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Scientific papers require models to integrate evidence across text, equations, figures, tables, code, and datasets while preserving its provenance. Beyond answer correctness, scientific reading requires verifiable outputs from operations such as evidence localization, definition extraction, and consistency checking. We introduce SciDocBench, a workflow-centered benchmark targeting these operations through 124 expert-authored and difficulty-screened questions across seven research-assistant capability groups, 19 subtasks, and five scientific domains. Each question is instantiated in four matched settings formed by pairing its bilingual variants with the All Images First and Markdown Interleaved document representations, yielding 496 evaluation instances. The strongest evaluated model, Claude-Opus-5, scores 62.6 out of 100, with remaining gaps in evidence localization, structured information extraction, cross-document synthesis, and robustness to document representation. To convert these diagnostics into scalable training signals, we introduce SciDocIR, a structured representation of scientific document objects, layout and cross-reference relations, and provenance. Using SciDocIR, we construct SciDocDataset, which contains 4K supervised fine-tuning instances and 10K reinforcement-learning instances, built on 14 verifiable training subtasks. Post-training Qwen3.6-27B on task-aligned data improves its SciDocBench score from 40.03 to 45.33 with supervised fine-tuning and to 45.74 with subsequent reinforcement learning. Both adapted models preserve DocVQA and InfoVQA performance and improve ChartQA accuracy over the original model by 0.80 and 3.40 points, respectively. Together, SciDocBench, SciDocIR, and SciDocDataset connect capability diagnosis with verifiable training-data construction for scientific-document assistants. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-07 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.05141 — 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 2026-10-16T22:18:11.439Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-07?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).