- Rating
- 3
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
The paper introduces VIGOR-his, containing 43,653 temporal street-view and satellite quadruplets across 11 cities. CrossTimeEdit uses recent street views and satellite-derived instructions to generate historical appearances through supervised fine-tuning and multi-reward reinforcement learning. The value of a dataset may be greater than the work presented in the paper itself.
Critical strengths
- This dataset fills a gap in historical cross-view observation.
- The authors' editing scheme accounts for the use of historical information and the preservation of context, employing a reward mechanism to evaluate the execution of edits, content retention, and editing quality.
- The ablation studies are comprehensive.
Critical weaknesses
-
In the context of buildings, human activity accounts for a significant portion and inherently involves a degree of randomness. Consequently, the images generated by the model may appear plausible while failing to align with reality—as illustrated by the results in Figure 3.
-
It appears that the model used to provide training rewards is the same as the one used for final evaluation. This raises concerns as to whether the reinforcement learning (RL) process is being employed solely to boost specific metrics.
-
Regarding the baseline methods, some of the training-based approaches cited are somewhat dated—for instance, ControlNet based on SD1.5. It would be more appropriate to include comparisons based on the same foundational model.
Questions for authors
Do existing human judgments or independent evaluators confirm the SFT-to-RL improvement on requested historical changes?
5 comments
Response to Weakness 2 and Question
Independent confirmation of SFT-to-RL improvements (Weakness 2 and Question)
We thank the reviewer for requesting human or independent evaluation of the improvements from SFT to RL. We agree that verification beyond the training reward model is important for establishing whether RL improves historical editing performance. We therefore conducted evaluations using GPT-5.6-Terra and Gemini-3.6-Flash, both separate from the Gemini-3.1-Flash-Lite reward model, and provide complementary qualitative ablation comparisons.
The table below reports the independent ablation results, including a direct comparison between SFT alone (w/o RL) and the full SFT-plus-RL model. Both evaluators assess the same generated outputs using the original IA/BP/QP criteria and Overall weights of (0.45, 0.30, 0.25).
GPT-5.6-Terra
| Method | IA ↑ | BP ↑ | QP ↑ | Overall ↑ |
|---|---|---|---|---|
| Baseline | 5.827 | 8.467 | 7.318 | 6.992 |
| w/o RL | 6.292 | 6.487 | 7.361 | 6.618 |
| w/o SFT | 6.038 | 8.542 | 7.383 | 7.126 |
| w/o GT | 6.665 | 6.690 | 7.433 | 6.865 |
| w/o DGN | 6.750 | 6.897 | 7.392 | 6.954 |
| Full | 6.893 | 7.493 | 7.548 | 7.237 |
Gemini-3.6-Flash
| Method | IA ↑ | BP ↑ | QP ↑ | Overall ↑ |
|---|---|---|---|---|
| Baseline | 4.561 | 6.503 | 6.499 | 5.628 |
| w/o RL | 5.053 | 5.206 | 5.997 | 5.335 |
| w/o SFT | 4.613 | 6.388 | 6.410 | 5.595 |
| w/o GT | 4.788 | 5.154 | 6.382 | 5.296 |
| w/o DGN | 4.958 | 5.310 | 6.448 | 5.436 |
| Full | 5.838 | 5.920 | 6.647 | 6.065 |
The results show consistent improvements from SFT to RL across all three criteria under both independent evaluators. Under GPT-5.6-Terra, IA increases from 6.292 to 6.893, BP from 6.487 to 7.493, QP from 7.361 to 7.548, and Overall from 6.618 to 7.237. Under Gemini-3.6-Flash, the corresponding improvements are 5.053 to 5.838 for IA, 5.206 to 5.920 for BP, 5.997 to 6.647 for QP, and 5.335 to 6.065 for Overall. Both evaluators also rank the full model highest in Overall among all ablation variants. These results demonstrate that RL improves historical edit execution, background preservation, and visual quality beyond the preferences of the training reward model.
The qualitative ablation comparisons provide complementary visual evidence. Across matched locations, the full model executes the requested structural edits while better preserving surrounding road and building appearance than the SFT-only and other ablated variants. Together, the independent scores and qualitative comparisons support consistent improvements in historical editing from SFT to RL.
The table below further compares CrossTimeEdit with general image editing models on VIGOR-his. CrossTimeEdit achieves the highest IA, QP, and Overall among the evaluated open-source image editing models under both independent evaluators, showing that its competitive performance also persists beyond the original evaluator.
GPT-5.6-Terra
| Model | IA ↑ | BP ↑ | QP ↑ | Overall ↑ |
|---|---|---|---|---|
| Closed-source models | ||||
| Nano Banana 2 Lite | 8.317 | 9.066 | 7.609 | 8.365 |
| Qwen-Image-3.0-Pro | 7.629 | 9.408 | 7.658 | 8.170 |
| GPT-Image-2 | 7.949 | 9.377 | 7.804 | 8.341 |
| Open-source models | ||||
| FLUX.2 [Klein] 4B | 5.827 | 8.467 | 7.318 | 6.992 |
| FLUX.2 [Klein] 9B | 6.480 | 7.839 | 7.240 | 7.078 |
| OmniGen2 | 5.320 | 8.115 | 6.542 | 6.464 |
| LongCat-Image-Edit | 6.569 | 8.027 | 7.127 | 7.146 |
| CrossTimeEdit | 6.893 | 7.493 | 7.548 | 7.237 |
Gemini-3.6-Flash
| Model | IA ↑ | BP ↑ | QP ↑ | Overall ↑ |
|---|---|---|---|---|
| Closed-source models | ||||
| Nano Banana 2 Lite | 5.699 | 6.601 | 6.738 | 6.229 |
| Qwen-Image-3.0-Pro | 5.779 | 6.957 | 6.856 | 6.401 |
| GPT-Image-2 | 6.178 | 7.109 | 7.046 | 6.674 |
| Open-source models | ||||
| FLUX.2 [Klein] 4B | 4.561 | 6.503 | 6.499 | 5.628 |
| FLUX.2 [Klein] 9B | 4.869 | 5.939 | 6.336 | 5.557 |
| OmniGen2 | 3.720 | 6.115 | 5.941 | 4.994 |
| LongCat-Image-Edit | 4.992 | 6.154 | 6.451 | 5.705 |
| CrossTimeEdit | 5.838 | 5.920 | 6.647 | 6.065 |
We have added the independent evaluation protocol and analysis to Section 5.5, with the main results summarized in Fig. 4. The complete model comparison and ablation results are reported in Tables 5 and 6 of Appendix B.2, respectively. The qualitative ablation comparisons and accompanying analysis are provided in Appendix C.2.
Response to Weakness 3
Baseline selection and comparison with the same backbone (Weakness 3)
We thank the reviewer for this suggestion. Recent cross-view methods such as CrossViewDiff [1] and SatDreamer360 [2] build on Stable Diffusion-based architectures. To include a stronger backbone in the comparison, we fine-tuned FLUX.2 [Klein] 4B with LoRA on VIGOR-his for satellite-to-street generation, using the earlier satellite image as input and the same pretrained foundation model as CrossTimeEdit.
Starting from the pretrained distilled model, we trained Transformer LoRA adapters with rank and alpha both set to 32, while freezing the backbone weights, VAE, and text encoder. Training used AdamW with a learning rate of 3 × 10⁻⁵ and a global effective batch size of 8 on four NVIDIA RTX 3090 GPUs for 10 epochs. The final checkpoint generated 1024 × 512 panoramas using a fixed satellite-to-street prompt, four denoising steps, and a guidance scale of 1.0.
The table below reports the updated comparison with cross-view generation models on VIGOR-his.
| Model | PSNR ↑ | SSIM ↑ | LPIPS ↓ | SD ↑ | CV-C ↑ | CV-VR ↑ | CV-PQ ↑ | Overall ↑ |
|---|---|---|---|---|---|---|---|---|
| Sat2Density | 14.685 | 0.532 | 0.634 | 20.040 | 2.048 | 1.973 | 2.006 | 2.015 |
| ControlS2S | 14.119 | 0.519 | 0.657 | 20.015 | 2.027 | 2.180 | 2.071 | 2.084 |
| ControlNet | 12.142 | 0.344 | 0.680 | 17.885 | 1.119 | 2.118 | 2.404 | 1.740 |
| FLUX.2 [Klein] 4B | 14.716 | 0.526 | 0.592 | 19.325 | 2.160 | 1.495 | 1.887 | 1.892 |
| CrossTimeEdit | 12.589 | 0.463 | 0.530 | 18.705 | 3.005 | 3.980 | 3.998 | 3.546 |
CrossTimeEdit outperforms the fine-tuned FLUX.2 [Klein] 4B baseline in all three cross-view evaluation criteria, achieving CV-C of 3.005 versus 2.160, CV-VR of 3.980 versus 1.495, and CV-PQ of 3.998 versus 1.887. Overall increases from 1.892 to 3.546, while LPIPS decreases from 0.592 to 0.530. These results show that CrossTimeEdit’s advantages persist when the pretrained foundation model is shared, supporting the effectiveness of using recent street views to preserve persistent scene content while applying historical edits guided by temporal satellite evidence.
We have added this baseline to Table 3, introduced its cross-view fine-tuning setup in Section 5.1, and updated the quantitative comparison in Section 5.2. Appendix B.8 now documents its training and inference configurations.
[1] Weijia Li et al. “CrossViewDiff: A Cross-View Diffusion Model for Satellite-to-Street View Synthesis.” Pattern Recognition, 114735, 2026.
[2] Xianghui Ze et al. “SatDreamer360: Geometry Consistent Street-View Video Generation from Satellite Imagery.” arXiv preprint, 2025.
Response to Weakness 1
Visual plausibility and historical reality (Weakness 1)
We agree with the reviewer that human activity introduces variability that makes the exact recovery of historical street scenes challenging, particularly for transient objects such as pedestrians and vehicles. Historical satellite imagery is limited by spatial resolution, overhead viewing geometry, and occlusion, and its acquisition time generally differs from that of the corresponding street view. Consequently, it cannot reliably establish which pedestrians or vehicles were present at a particular street-level moment, their locations, or their appearance. Without sufficient observational evidence, a VLM cannot reliably describe these transient states or translate them into historically grounded editing instructions and learning signals.
CrossTimeEdit therefore focuses on persistent building and road changes that can be supported by temporal satellite evidence. Its annotation pipeline disregards transient objects when identifying structural changes, while the recent street view provides a reference for viewpoint and persistent scene content. This formulation directs learning toward restoring the historical semantic layout of buildings and roads, rather than reconstructing the precise configuration of human activity at a particular moment.
Despite the uncertainty in transient human activity, the qualitative results in Fig. 3 show that CrossTimeEdit can recover the building and road configurations that characterize the earlier scene. The relevant evidence is the restoration of changed structures and the retention of surrounding scene content, rather than the presence of particular pedestrians or vehicles. Compared with the direct cross-view generation methods illustrated in Appendix C.3, using the recent street view as a reference helps retain observed spatial context and concentrate generation on the historical changes supported by satellite evidence. This enables CrossTimeEdit to recover a semantic layout more consistent with the earlier street view, even when moment-specific human activity cannot be reconstructed.
We have clarified in Section 4.1 that the task targets persistent building and road changes rather than the exact reconstruction of transient objects such as pedestrians and vehicles. Section 5.3 now emphasizes that the qualitative results demonstrate visually plausible generation together with recovery of the historical semantic layout. We have also expanded Appendix E to discuss the observational constraints on historical reconstruction, and Appendix F to consider archival photographs, temporally and geographically grounded textual records, and uncertainty modeling as directions for incorporating additional historical evidence and representing plausible alternatives.