# Target Viewpoint Reproduction on TVRBench (test)

Source: [Target Viewpoint Reproduction on TVRBench (test)](<https://www.sota2.com/research/sota/target-viewpoint-reproduction-on-tvrbench-test>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-02T10:46:49.017Z

Evaluation of model performance on the test split of the TVRBench dataset, focusing on Target Viewpoint Reproduction capabilities.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Target Viewpoint Reproduction |
| Dataset (including any stated split) | TVRBench (test) |
| Metric | Overall Performance |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 8 of 8 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Overall Performance](<https://www.sota2.com/research/sota/target-viewpoint-reproduction-on-tvrbench-test.md?metric=Overall%20Performance>): Higher is better; unit: PERCENT.

- [S-e Score](<https://www.sota2.com/research/sota/target-viewpoint-reproduction-on-tvrbench-test.md?metric=S-e%20Score>): Higher is better; unit: PERCENT.

- [S-h Score](<https://www.sota2.com/research/sota/target-viewpoint-reproduction-on-tvrbench-test.md?metric=S-h%20Score>): Higher is better; unit: PERCENT.

- [M-e Score](<https://www.sota2.com/research/sota/target-viewpoint-reproduction-on-tvrbench-test.md?metric=M-e%20Score>): Higher is better; unit: PERCENT.

- [M-h Score](<https://www.sota2.com/research/sota/target-viewpoint-reproduction-on-tvrbench-test.md?metric=M-h%20Score>): Higher is better; unit: PERCENT.

- [Average Steps](<https://www.sota2.com/research/sota/target-viewpoint-reproduction-on-tvrbench-test.md?metric=Average%20Steps>): Lower is better; unit: SCALAR.

- [Stop Success Rate](<https://www.sota2.com/research/sota/target-viewpoint-reproduction-on-tvrbench-test.md?metric=Stop%20Success%20Rate>): Higher is better; unit: PERCENT.

- [F-Stop Score](<https://www.sota2.com/research/sota/target-viewpoint-reproduction-on-tvrbench-test.md?metric=F-Stop%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Overall Performance | Result date | Report paper |
| --- | --- | --- | --- |
| Multi-turn GRPO | 51.4 | Not recorded | 6a1e6408dbd7150506f8562a |
| VA-SFT | 50.8 | Not recorded | 6a1e6408dbd7150506f8562a |
| AO-SFT | 44.2 | Not recorded | 6a1e6408dbd7150506f8562a |
| VA-CoT-SFT | 35.6 | Not recorded | 6a1e6408dbd7150506f8562a |
| Single-turn GRPO | 26.2 | Not recorded | 6a1e6408dbd7150506f8562a |
| AO-CoT-SFT | 24.8 | Not recorded | 6a1e6408dbd7150506f8562a |
| Qwen3.5-9B | 2.8 | Not recorded | 6a1e6408dbd7150506f8562a |
| Qwen3.5-9B | 0 | Not recorded | 6a1e6408dbd7150506f8562a |

### Multi-turn GRPO

Overall Performance: 51.4. Result ID: 6a1eb38afbc373592c199dd4.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6408dbd7150506f8562a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e6408dbd7150506f8562a.md>)

[Implementation](<https://github.com/aim-uofa/TVRBench>)

Recorded evaluation configuration: {"Mem":"VA","CoT":"no","Init":"VA-SFT"}. This record does not assert independent reproduction.

### VA-SFT

Overall Performance: 50.8. Result ID: 6a1eb38afbc373592c199dd2.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6408dbd7150506f8562a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e6408dbd7150506f8562a.md>)

[Implementation](<https://github.com/aim-uofa/TVRBench>)

Recorded evaluation configuration: {"Mem":"VA","CoT":"no","Init":"Qwen3.5-9B"}. This record does not assert independent reproduction.

### AO-SFT

Overall Performance: 44.2. Result ID: 6a1eb38afbc373592c199dd0.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6408dbd7150506f8562a.md>)

Recorded evaluation configuration: {"Mem":"AO","CoT":"no","Init":"Qwen3.5-9B"}. This record does not assert independent reproduction.

### VA-CoT-SFT

Overall Performance: 35.6. Result ID: 6a1eb38afbc373592c199dd1.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6408dbd7150506f8562a.md>)

Recorded evaluation configuration: {"Mem":"VA","CoT":"yes","Init":"Qwen3.5-9B"}. This record does not assert independent reproduction.

### Single-turn GRPO

Overall Performance: 26.2. Result ID: 6a1eb38afbc373592c199dd3.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6408dbd7150506f8562a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78121e209131cea76a45.md>)

Recorded evaluation configuration: {"Mem":"AO","CoT":"no","Init":"AO-SFT"}. This record does not assert independent reproduction.

### AO-CoT-SFT

Overall Performance: 24.8. Result ID: 6a1eb38afbc373592c199dcf.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6408dbd7150506f8562a.md>)

Recorded evaluation configuration: {"Mem":"AO","CoT":"yes","Init":"Qwen3.5-9B"}. This record does not assert independent reproduction.

### Qwen3.5-9B

Overall Performance: 2.8. Result ID: 6a1eb38afbc373592c199dcd.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6408dbd7150506f8562a.md>)

Recorded evaluation configuration: {"Mem":"AO"}. This record does not assert independent reproduction.

### Qwen3.5-9B

Overall Performance: 0. Result ID: 6a1eb38afbc373592c199dce.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6408dbd7150506f8562a.md>)

Recorded evaluation configuration: {"Mem":"VA"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.