Metraipolis › archive › dc_a8d3cec85515
Self-Evolving Code-with-Image Reasoning — arXiv:2608.11292
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.11292
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-19T05:49:18.536Z
- SHA-256 of the archived text
94163feeaec083831745963e4483a060a2181839b6e88524f0253206e5457bcb- Stable URL of this page
https://metraipolis.com/documento/dc_a8d3cec85515
How to cite
Self-Evolving Code-with-Image Reasoning — arXiv:2608.11292 — arXiv (en), https://arxiv.org/abs/2608.11292 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a8d3cec85515 (Metraipolis archive dc_a8d3cec85515, sha256 94163feeaec083831745963e4483a060a2181839b6e88524f0253206e5457bcb)
BibTeX
@misc{dca8d3cec85515,
title = {Self-Evolving Code-with-Image Reasoning — arXiv:2608.11292},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.11292},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a8d3cec85515; sha256 94163feeaec083831745963e4483a060a2181839b6e88524f0253206e5457bcb},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.11292. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Multimodal models increasingly reach for tools when solving visual tasks (crop, zoom, rotate, brighten), a paradigm known as thinking-with-images. The central challenge is one of perception: tools mostly serve to expose visual evidence, reasoning over that evidence stays in language, and most targets are ones a human could in principle determine by inspection. Some visual questions, however, are not bottlenecked by perception: recovering their answers requires executing a multi-step visual algorithm over the pixels. On such questions a model often names the correct algorithm at once yet still answers wrong, because language can describe an algorithm without being able to run one. Code-with-Image crosses that line: given nothing but a Python interpreter, the model must implement a genuine visual algorithm in code to solve the task; the program itself becomes the reasoning. The bottleneck then shifts from executing code to deciding which algorithm to implement. So we let the model teach itself: a training-free reflection loop studies its own failed programs, tests repairs against constructive ground truth, and keeps what survives as portable skills. On our Code-with-Image Bench (CwI-Bench), thirty task families induced by hidden visual computations with disjoint learning and evaluation splits, even GPT-5.6-luna stays below 30% with tool-free chain of thought; given a bare interpreter it reaches 43%, and with skills evolved through its own executable reflection, 67%. The open 27B model climbs the same ladder (9% $\rightarrow$ 33% $\rightarrow$ 56%), and the skills are plain text, transferable across scales and families. When code carries the reasoning, debugging code becomes debugging reasoning. Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-08-13 in cs.CV). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2608.11292 — 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 2026-10-19T05:42:25.858Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-08-13?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).