| title | Learning Where Outcomes Change:Credit-Addressable Reasoning for Multimodal Geometry |
|---|---|
| authors | Jiani Guo, Junjie Wang, Jie Wu, Pengxiang Zhao, Dongdong Zhang, Shaohan Huang, Yujiu Yang, Furu Wei |
| date | 2026-08-31 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2608.30457 |
| affiliation | Tsinghua IIGroup |
| upvotes | 4 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf |
| abstract | Multimodal geometry reasoning requires VLMs to extract precise visual relations and preserve them through multi-step deduction. Existing free-form traces obscure the decisions that determine the answer, and trajectory-level reinforcement learning distributes a single terminal signal across the entire response. We introduce credit-addressable reasoning, in which the semantic units exposed during inference also define where learning compares alternatives and assigns credit. We instantiate this principle with Code-CoT, which retains the diagram, represents visual relations as line-addressable executable code, and organizes reasoning into typed events, and CE-GRPO, which selects event boundaries using structural priors and type-normalized entropy, samples complete continuations from shared prefixes, and converts outcome differences into localized advantages. Across nine geometry benchmarks, CE-GRPO achieves an average accuracy of 76.04, outperforming Qwen3-VL-8B and trajectory-level GRPO by 8.09 and 3.43 points, respectively. Its relative advantage increases with the number of intermediate events, demonstrating the value of representation--optimization co-design for long, dependency-heavy multimodal reasoning. |
cite
@misc{guo2026learning,
title={Learning Where Outcomes Change:Credit-Addressable Reasoning for Multimodal Geometry},
author={Jiani Guo and Junjie Wang and Jie Wu and Pengxiang Zhao and Dongdong Zhang and Shaohan Huang and Yujiu Yang and Furu Wei},
year={2026},
eprint={2608.30457},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2608.30457}
}