HISD：基于提示自蒸馏的推理分割方法
Hint-Informed Self-Distillation for Reasoning Segmentation
AI 导读
研究者提出提示自蒸馏方法 HISD，利用外部模型离线生成的 Segment Hints 进行 token 级信用分配，改进推理分割。在统一训练协议下，HISD 在 ReasonSeg 测试集上较 DPAD 复现基线的 gIoU 和 cIoU 分别提升 3.00 和 3.55 个百分点，推理时无需提示或教师模型，并将有效推理 span 的平均 token 数减少 80.17% 至 91.83%。
正文 · 原文
Abstract:Reasoning segmentation requires identifying the target referred to by a query among visually similar objects, yet trajectory-level rewards cannot directly indicate which tokens provide relevant visual evidence. We propose Hint-Informed Self-Distillation (HISD), which uses target-disambiguation evidence for token-level credit assignment. An external model generates Segment Hints offline from training images, queries, and ground-truth target annotations, organizing target identity, discriminative attributes, confusable objects, and anchoring paths. A hint-conditioned exponential moving average teacher scores the student's own trajectories, and HISD computes the teacher-to-student evidence ratio over a token support set to adjust only the magnitudes of positive group-relative advantages, leaving nonpositive advantages unchanged. This design encourages concise reasoning while retaining the information needed for target disambiguation, and the student requires neither hints nor a teacher at inference. Under a unified training protocol, HISD improves ReasonSeg test gIoU and cIoU over our reproduction of DPAD by 3.00 and 3.55 percentage points, respectively. Across the five evaluated splits of ReasonSeg and the RefCOCO family, HISD achieves higher segmentation metrics than this baseline and reduces the average token count of valid reasoning spans by 80.17 to 91.83 percent.
| Comments: | 11 pages, 5 figures, 7 tables. The first two authors contributed equally |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.14931 [cs.CV] |
| (or arXiv:2611.14931v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2611.14931 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Kunze Jiang [view email]
[v1]
Wed, 11 Nov 2026 02:34:13 UTC (6,236 KB)
来源：arXiv cs.AI · arxiv.org