Metraipolis › archive › dc_250eeb24e398
SpatialOPSD: Self-Distilling Spatial Intelligence from Verified Coding Agent Traces — arXiv:2610.11366
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.11366
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-09T06:23:18.102Z
- SHA-256 of the archived text
aa1ad0c7829bf37056dcc95d079a156a772dcda8921f8775e0a5aa39496a4185- Stable URL of this page
https://metraipolis.com/documento/dc_250eeb24e398
How to cite
SpatialOPSD: Self-Distilling Spatial Intelligence from Verified Coding Agent Traces — arXiv:2610.11366 — arXiv (en), https://arxiv.org/abs/2610.11366 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_250eeb24e398 (Metraipolis archive dc_250eeb24e398, sha256 aa1ad0c7829bf37056dcc95d079a156a772dcda8921f8775e0a5aa39496a4185)
BibTeX
@misc{dc250eeb24e398,
title = {SpatialOPSD: Self-Distilling Spatial Intelligence from Verified Coding Agent Traces — arXiv:2610.11366},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.11366},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_250eeb24e398; sha256 aa1ad0c7829bf37056dcc95d079a156a772dcda8921f8775e0a5aa39496a4185},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.11366. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Spatial coding agents significantly improve spatial reasoning in Multimodal Large Language Models (MLLMs) by using external tools to generate verified execution traces. However, this paradigm inherently suffers from prohibitive inference-time overhead and external dependencies. In this paper, we explore whether an MLLM can internalize this agentic capability to operate entirely tool-free. We begin with a simple observation: prompting an MLLM with summarized execution traces of a spatial coding agent naturally unlocks the model's internal spatial Chain-of-Thought (CoT). Motivated by this, we introduce SpatialOPSD, an on-policy self-distillation framework that internalizes spatial reasoning into a standalone MLLM by formulating verified agent traces as privileged information. To mitigate privileged-information leakage during distillation, we introduce Repetition-Aware Distillation, which combines repetition masking with unlikelihood regularization. Experiments across multiple benchmarks demonstrate that self-distilling SpatialOPSD achieves higher average accuracy than SFT and GRPO on both spatial and OOD datasets, exhibiting superior performance and generalization. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2610.11366 — 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 2026-10-09T05:50:31.288Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).