- Rating
- 3(+1)
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 1
- Soundness
- 3
Read review
Paper summary
This paper introduces MixOPD, an on-policy distillation framework for vision-centric reasoning that interleaves explicit grounding with the reasoning process. The main idea is that token-level distillation treats different coordinate sequences as different outputs, even when they point to the same visual evidence. As a result, geometrically valid grounding proposals can still be penalized. To address this, MixOPD estimates a teacher-supported box distribution through an efficient teacher-forced constrained search. It then weights each proposal's coordinate-token loss by its expected 1-IoU against this distribution. The paper also provides a theoretical analysis showing that this weighting acts as a selective local gradient scaling under stated assumptions. Experiments show that MixOPD improves both grounding and visual reasoning on top of standard and entropy-aware OPD, without extra teacher calls or search at inference time.
Critical strengths
-
Strong motivation: The paper clearly identifies the mismatch between token-level distillation and spatial grounding, where different coordinate sequences can describe the same visual evidence yet receive unnecessary penalties.
-
Simple and practical method: Spatial calibration integrates naturally with both OPD and EOPD, with no additional search or teacher calls at inference time.
-
Sufficient abaltion studies: Comparisons with relevant distillation methods and representative post-training baselines adequately demonstrate improvements in both grounding and visual reasoning.
Critical weaknesses
-
Presentation is a major weakness. The method section introduces too many formal definitions before clearly explaining the algorithm, making a simple method hard to follow. It should better explain why smaller spatial errors lead to lower loss weights. The search procedure is also too brief, with important details left to the appendix, and Figure 3 does little to clarify the intuition. A clear overview and a concrete example would substantially improve readability.
-
Limited validation across model families: The method is evaluated only on Qwen3-VL students at the 4B and 8B scales. While this demonstrates consistency across model sizes, it does not establish generalization across model families. Experiments with another vision-language model family would strengthen the claim.
Questions for authors
- What is the distribution of bounding-box sizes, aspect ratios, and numbers per image in the training data? How does performance vary across these factors?
- Does MixOPD remain effective on video or multi-image reasoning tasks, where evidence may span multiple frames or images?
- Evaluation is limited within the Qwen3-VL family. Its effectiveness across other model families remains unverified.
4 comments
Response to Reviewer#KSZP (Part2/2)
Thank you for your insightful questions. We provide a point-by-point response below.
Q1: What is the distribution of bounding-box sizes, aspect ratios, and numbers per image in the training data? How does performance vary across these factors?
A: Following your suggestion, we have added statistics on bounding-box sizes, aspect ratios, and counts per training sample, covering about 280k boxes. The full distributions are presented in Figure 16 (page 24 of the revised PDF).
Key statistics are summarized below:
| Factor | Summary |
|---|---|
| Box size (COCO area classes) | 20% small, 31% medium, and 49% large |
| Aspect ratio (width/height) | The distribution peaks near 1:1, with most boxes having aspect ratios between 1:4 and 4:1 and tails extending to very tall (≤1:16) and very wide (≥16:1) boxes |
| Boxes per sample | 53.3% contain 1 box and 33.1% contain 2 boxes; 93.9% contain at most 4 boxes |
Furthermore, we analyze MixOPD-4B/8B performance on TreeBench across evidence-box size, aspect ratio, and count. The tables report task-category-adjusted accuracy (%).
MixOPD-4B
| Box size | <4% | 4–7% | 7–12% | 12–25% | ≥25% |
|---|---|---|---|---|---|
| Accuracy | 51.3 | 52.3 | 42.6 | 55.0 | 62.5 |
| Aspect ratio | ≤1:3 | 1:3–1:1.5 | ~1:1 | 1.5:1–3:1 | ≥3:1 |
|---|---|---|---|---|---|
| Accuracy | 49.8 | 50.3 | 52.4 | 61.3 | 50.7 |
| Boxes per image | 1 | 2 | 3 | ≥4 |
|---|---|---|---|---|
| Accuracy | 57.0 | 53.2 | 47.4 | 38.7 |
MixOPD-8B
| Box size | <4% | 4–7% | 7–12% | 12–25% | ≥25% |
|---|---|---|---|---|---|
| Accuracy | 55.2 | 58.4 | 53.3 | 52.3 | 49.2 |
| Aspect ratio | ≤1:3 | 1:3–1:1.5 | ~1:1 | 1.5:1–3:1 | ≥3:1 |
|---|---|---|---|---|---|
| Accuracy | 51.2 | 54.7 | 52.0 | 61.0 | 54.0 |
| Boxes per image | 1 | 2 | 3 | ≥4 |
|---|---|---|---|---|
| Accuracy | 62.5 | 51.6 | 43.3 | 41.9 |
We have three observations:
-
Box size. Neither model exhibits a consistent monotonic relationship between accuracy and box size. Despite the greater perceptual challenges posed by smaller evidence regions, accuracy shows no consistent decline as box size decreases, suggesting that MixOPD may enhance the base models’ ability to perceive and understand small visual evidence.
-
Aspect ratio. Both models achieve their highest accuracy on moderately wide boxes (1.5:1–3:1), followed by a decline for extremely wide boxes (≥3:1). Notably, the median box aspect ratio in our training data is close to 1:1. We therefore hypothesize that this preference is inherited from the base models rather than introduced by MixOPD.
-
Box count. Accuracy decreases monotonically as the number of target boxes increases, from 57.0% to 38.7% for 4B and from 62.5% to 41.9% for 8B. This is consistent with the greater difficulty of grounding and reasoning over multiple evidence regions.
Q2: Does MixOPD remain effective on video or multi-image reasoning tasks, where evidence may span multiple frames or images?
A: Thank you for raising this question. We added an evaluation on MuirBench [1] to assess multi-image generalization. Appendix D.1, Multi-image Understanding (page 36) of the revised PDF provides comprehensive quantitative results (Table 9) and qualitative examples (Figure 17). Selected results are summarized below:
[1] MuirBench: A Comprehensive Benchmark for Robust Multi-image Understanding. ICLR, 2025.
| Method | Base Model | #Parameters | Accuracy (%) |
|---|---|---|---|
| Teacher | Qwen3-VL‡ | 32B | 72.21 |
| - | Qwen3-VL‡ | 235B | 72.64 |
| Base | Qwen3-VL‡ | 8B | 65.31 |
| OPD | Qwen3-VL | 8B | 67.86 |
| MixOPD | Qwen3-VL | 8B | 74.05 |
| EOPD | Qwen3-VL | 8B | 69.07 |
| MixOPD | Qwen3-VL | 8B | 74.22 |
‡ Evaluated w/o special system prompts. OPD, EOPD, and MixOPD share the same system prompt (Appendix B.5).
We highlight three observations:
-
Spatial calibration improves multi-image performance after single-image post-training. Both OPD and EOPD improve over the base models despite using only single-image data for post-training. MixOPD provides further gains over the corresponding OPD/EOPD baselines by 6.19/5.15 pp.
-
MixOPD achieves competitive multi-image performance with smaller models. Under the EOPD formulation, the 8B variant reaches 74.22% with vision-centric reasoning, exceeding the direct-answering teacher by 2.01 percentage points.
-
Visually grounded reasoning transfers to multi-image inputs. Figure 17 (p.36) illustrates how the reasoning pattern learned through single-image post-training supports cross-image counting and attribute-based comparison through localized visual evidence. Together with the quantitative results, these examples support the transfer of this reasoning pattern to multi-image scenarios, accompanied by consistent gains in answer accuracy.
These results directly support MixOPD's effectiveness in multi-image scenarios.
Sincerely, we hope our responses are helpful to address your concerns.
The author clearly address my concerns about the loss weight intuition, generalization across model families and multi-image evaluation. They also redesigned the Figure 3 and expanded the search algorithm description, which makes the paper easy to follow. So I raise my score the weak accept.
Dear reviewer#KSZP,
We sincerely appreciate that our feedback addresses your concerns. Your insightful comments are very important to further strengthen this work.
Best regards,
Submission 3146 Authors.
Response to Reviewer#KSZP (Part1/2)
Thank you for your helpful comments. We provide a point-by-point response below.
A: Standard OPD encourages token-level agreement, which need not reflect spatial agreement. In Fig. 2, the student proposal (247, 305, 376, 437) and the teacher-preferred box (252, 298, 382, 441) localize the same button decoration, yet several coordinate positions incur substantial token-level penalties. MixOPD therefore calibrates these local signals using the expected spatial error under the teacher-supported box distribution. Smaller errors reduce unnecessary pressure to match exact coordinate tokens when the proposal already agrees spatially with the teacher, while larger errors retain more of the original distillation signal. We have made this intuition explicit in §3.2.
A: To clarify the search procedure, we have added a step-by-step overview with a concrete grounding example in Fig. 3 (please see page 4 in the revised PDF) and reorganized the teacher-forced constrained search paragraph in §3.2 (L.243-L.248) to provide a detailed description of the algorithm. The new figure and the supplied description illustrate how cached teacher predictions are converted into a teacher-supported box distribution, connecting the token-level mismatch illustrated in Fig. 2 to the spatial calibration shown in Fig. 4.
A: Following your suggestion, we added experiments on the InternVL3.5 family. We initialize the student from InternVL3.5-8B and fix InternVL3.5-14B as the teacher model, following the training recipe in Appendix B.3. We compare MixOPD with representative post-training baselines: GRPO, DAPO, OPD, and EOPD. Complete results are provided in Table 2 (page 8) of the revised PDF, with selected results summarized below.
We highlight two observations:
Spatial calibration improves grounding under both objectives. On InternVL3.5-8B, MixOPD raises mIoU from 0.28 to 0.44 under standard OPD and from 0.41 to 0.45 under EOPD.
Reasoning gains extend beyond Qwen3-VL. MixOPD also improves accuracy on all 9 VQA benchmarks over the corresponding OPD/EOPD baselines, supporting the effectiveness of spatial calibration across both evaluated model families.