Metraipolis › archive › dc_bc0f54941a50

SciWalker: Synthesizing Scientific Coding Problems with Operator Graphs and Execution Feedback — arXiv:2609.30054

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.30054
License (as archived)
CC0-1.0
Archived
2026-10-08T04:26:31.990Z
SHA-256 of the archived text
44f6355f2878d6edabc9235ee9aee721170d7ee94d6dde06c8537aa30337c383
Stable URL of this page
https://metraipolis.com/documento/dc_bc0f54941a50

How to cite

SciWalker: Synthesizing Scientific Coding Problems with Operator Graphs and Execution Feedback — arXiv:2609.30054 — arXiv (en), https://arxiv.org/abs/2609.30054 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_bc0f54941a50 (Metraipolis archive dc_bc0f54941a50, sha256 44f6355f2878d6edabc9235ee9aee721170d7ee94d6dde06c8537aa30337c383)
BibTeX
@misc{dcbc0f54941a50,
  title = {SciWalker: Synthesizing Scientific Coding Problems with Operator Graphs and Execution Feedback — arXiv:2609.30054},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.30054},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_bc0f54941a50; sha256 44f6355f2878d6edabc9235ee9aee721170d7ee94d6dde06c8537aa30337c383},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.30054. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Improving the scientific coding capabilities of large language models (LLMs) requires high-quality training data. However, such data remain scarce because manually authoring realistic problems is costly and time-consuming, while systematically covering diverse scientific domains and algorithmic combinations remains challenging. To address this, we introduce SciWalker, a framework for synthesizing scientific coding problems through operator-chain sampling and execution feedback. The framework combines scientific library interfaces with operation modes to instantiate operators, organizes them into operator graphs, and samples operator chains as computational workflow cues. Guided by these cues, we adopt LLMs to generate scientifically grounded problem statements, reference solutions, and tests, with failed generations iteratively repaired using execution feedback. By combining structured workflow composition with verification and quality review, SciWalker enables scalable task generation while promoting scientific grounding, computational diversity, and executability. Using this framework, we construct 8,178 high-quality problems spanning 5 scientific domains and 32 subdomains. To evaluate their training utility, we conduct reinforcement learning on Qwen3.5-9B using the GSPO algorithm. This training improves SciCode subproblem accuracy by 9.9 percentage points, from 29.3% to 39.2%, with gains across scientific code generation, code repair, and reasoning benchmarks. The code for SciWalker is available at this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-25 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.30054 — 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 2026-10-08T03:50:57.871Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-25?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).