- Rating
- 3
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 3
Read review
Paper summary
This paper introduces GroundPO, a two-stage framework for reasoning-intensive visual grounding. It first uses GRPO to stabilize a teacher with privileged textual information, then applies on-policy distillation to train a student without privileged inputs. The authors also construct DeepGround, covering five reasoning tracks. Experiments show improved grounding performance and substantially reduced privilege leakage and excessive reasoning.
Critical strengths
- The paper is well presented, with a clear motivation.
- The proposed dataset DeepGround provides both training data and a challenging benchmark covering five complementary reasoning tracks. Results show strong improvements in reasoning-intensive and fine-grained grounding.
- No additional inference inputs or architectural changes. The proposed method is easy to follow.
Critical weaknesses
- The method is validated only on Qwen3.5 models. Testing multiple model sizes does not establish generality across model families.
- Although DeepGround includes screenshots and web images, GUI grounding performance is not separately evaluated, leaving its effectiveness in this setting unclear.
- Gains on conventional grounding are limited, and overall performance on V* Bench remains unchanged.
Questions for authors
- See weakness above.
- Could you provide the training-set distribution across reasoning tracks and more details on filtering and human verification?
- During Stage II, the internalized teacher still receives privileged textual information. Is this input necessary after policy internalization? Could you compare distillation with and without privileged teacher inputs to clarify its contribution to grounding performance and behavioral stability?
5 comments
W2: Generalization to GUI Grounding
Thanks for suggesting a separate evaluation of GUI grounding. We have added four GUI grounding benchmarks, ScreenSpot-Pro, ScreenSpot-V2, UI-Vision, and MMBenchGUI (Tab. R3-2, also included in Tab. 5 of the revised manuscript). GroundPO improves the average GUI accuracy on Qwen3.5-4B, 9B, and 27B by 1.53%, 1.55%, and 1.05%, respectively. GUI grounding also requires identifying a specific visual element from a language description and localizing it precisely. Although GroundPO is trained only on DeepGround-Train, its gains carry over to GUI grounding.
Tab. R3-2. Accuracy comparison of GroundPO on V* Bench and four GUI grounding benchmarks (%).
| Model | V* Bench | ScreenSpot-Pro | ScreenSpot-V2 | UI-Vision | MMBenchGUI | GUI Avg. |
|---|---|---|---|---|---|---|
| Qwen3.5-4B | 81.68 | 54.27 | 91.82 | 30.04 | 80.88 | 64.25 |
| GroundPO-4B | 83.25 (+1.57) | 57.18 (+2.91) | 91.98 (+0.16) | 32.98 (+2.94) | 80.96 (+0.08) | 65.78 (+1.53) |
| Qwen3.5-9B | 85.86 | 60.47 | 93.08 | 38.46 | 83.69 | 68.92 |
| GroundPO-9B | 86.39 (+0.53) | 62.05 (+1.58) | 93.87 (+0.79) | 40.74 (+2.28) | 85.22 (+1.53) | 70.47 (+1.55) |
| Qwen3.5-27B | 86.91 | 66.79 | 94.34 | 51.98 | 86.84 | 74.99 |
| GroundPO-27B | 88.48 (+1.57) | 68.12 (+1.33) | 94.34 (+0.00) | 53.91 (+1.93) | 87.79 (+0.95) | 76.04 (+1.05) |
W3: Conventional Grounding and General Visual Reasoning
Thank you for the attention to conventional grounding and general visual understanding capabilities.
V* Bench. We rechecked the evaluation protocol for V* Bench and evaluated it using the same GPT-OSS-120B judge as Vision-OPD. With this protocol, GroundPO-4B, 9B, and 27B improve V* Bench overall accuracy by 1.57%, 0.53%, and 1.57%, respectively (Tab. R3-2). Tab. 5 of the revised manuscript further reports the Direct Attributes and Relative Position dimensions: all three models improve on Relative Position, while Direct Attributes is maintained or improved. GroundPO thus improves grounding without degrading general visual reasoning.
Conventional grounding. This paper focuses on reasoning-intensive visual grounding. GroundPO improves over the corresponding base models on GroundingME, PR-Bench, and Ref-Adv at all scales, while on RefCOCO and RefCOCOg its performance stays close to the base models. Some RefCOCO/RefCOCOg annotations are coarser than the referring expression or do not match its target granularity. GroundPO tends to produce finer boxes, so its predictions can match the expression better yet receive lower IoU against the original annotation (App. C.5). Consistent with this, GroundPO achieves clear gains on the fine-grained benchmark RefCOCOm at all scales (Tab. 2).
W1: Generalization Across Model Families
Thank you for suggesting validation across model families. In addition to the original Qwen3.5-4B/9B/27B experiments, we have applied GroundPO to Qwen3-VL-4B-Thinking, Qwen3-VL-8B-Thinking, Qwen3.6-27B, and Qwen3.8-27B, with results shown in Tab. R3-1. F1@mIoU improves for all four models, by 12.16%, 10.01%, 19.07%, and 7.32%, respectively. These results extend the scope of validation from different parameter scales of Qwen3.5 to Qwen3-VL and other Qwen model families, supporting the applicability of GroundPO to different base models.
Tab. R3-1. DeepGround-Test performance of GroundPO across model families