# Pairwise Scene Generation Evaluation on Procthor-10K (Easy)

Source: [Pairwise Scene Generation Evaluation on Procthor-10K (Easy)](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy>)

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

Data updated: 2026-04-15T10:22:56.447Z

Evaluation of pairwise scene generation quality on the Easy subset of the Procthor-10K dataset. Performance is assessed by comparing two generated scenes based on Semantic, Orientation, and Overlap criteria, utilizing scores from SceneCritic and VLM evaluators, alongside inter-annotator agreement metrics.

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 | Pairwise Scene Generation Evaluation |
| Dataset (including any stated split) | Procthor-10K (Easy) |
| Metric | Score (SceneCritic) Method A |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 6 of 6 |
| 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

- [Score (SceneCritic) Method A](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=Score%20(SceneCritic)%20Method%20A>): Higher is better; unit: PERCENT.

- [Score (VLM) Method A](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=Score%20(VLM)%20Method%20A>): Higher is better; unit: PERCENT.

- [Score (SceneCritic) Method B](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=Score%20(SceneCritic)%20Method%20B>): Higher is better; unit: PERCENT.

- [Score (VLM) Method B](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=Score%20(VLM)%20Method%20B>): Higher is better; unit: PERCENT.

- [Semantic Score Method A](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=Semantic%20Score%20Method%20A>): Higher is better; unit: PERCENT.

- [Semantic Score Method B](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=Semantic%20Score%20Method%20B>): Higher is better; unit: PERCENT.

- [Orientation Score Method A](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=Orientation%20Score%20Method%20A>): Higher is better; unit: PERCENT.

- [Orientation Score Method B](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=Orientation%20Score%20Method%20B>): Higher is better; unit: PERCENT.

- [Overlap Score Method A](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=Overlap%20Score%20Method%20A>): Higher is better; unit: PERCENT.

- [Overlap Score Method B](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=Overlap%20Score%20Method%20B>): Higher is better; unit: PERCENT.

- [IAA Semantic](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=IAA%20Semantic>): Higher is better; unit: PERCENT.

- [IAA Orientation](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=IAA%20Orientation>): Higher is better; unit: PERCENT.

- [IAA Overlap](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=IAA%20Overlap>): Higher is better; unit: PERCENT.

- [Fleiss' k Semantic](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=Fleiss%27%20k%20Semantic>): Higher is better; unit: SCALAR.

- [Fleiss' k Orientation](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=Fleiss%27%20k%20Orientation>): Higher is better; unit: SCALAR.

- [Fleiss' k Overlap](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=Fleiss%27%20k%20Overlap>): Higher is better; unit: SCALAR.

- [VLM Agreement Semantic](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=VLM%20Agreement%20Semantic>): Higher is better; unit: PERCENT.

- [VLM Agreement Orientation](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=VLM%20Agreement%20Orientation>): Higher is better; unit: PERCENT.

- [VLM Agreement Overlap](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=VLM%20Agreement%20Overlap>): Higher is better; unit: PERCENT.

- [SceneCritic Agreement Semantic](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=SceneCritic%20Agreement%20Semantic>): Higher is better; unit: PERCENT.

- [SceneCritic Agreement Orientation](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=SceneCritic%20Agreement%20Orientation>): Higher is better; unit: PERCENT.

- [SceneCritic Agreement Overlap](<https://www.sota2.com/research/sota/pairwise-scene-generation-evaluation-on-procthor-10k-easy.md?metric=SceneCritic%20Agreement%20Overlap>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Score (SceneCritic) Method A | Result date | Report paper |
| --- | --- | --- | --- |
| heuristic vs holodeck | 75.6 | Not recorded | 69df1bcb7d0172140ec7bc2c |
| heuristic vs layoutgpt | 75.6 | Not recorded | 69df1bcb7d0172140ec7bc2c |
| heuristic vs layoutvlm | 75.6 | Not recorded | 69df1bcb7d0172140ec7bc2c |
| holodeck vs layoutgpt | 71.3 | Not recorded | 69df1bcb7d0172140ec7bc2c |
| holodeck vs layoutvlm | 71.3 | Not recorded | 69df1bcb7d0172140ec7bc2c |
| layoutgpt vs layoutvlm | 53.6 | Not recorded | 69df1bcb7d0172140ec7bc2c |

### heuristic vs holodeck

Score (SceneCritic) Method A: 75.6. Result ID: 69df667247f34ce924c8bee7.

[Reporting paper](<https://www.sota2.com/research/paper/69df1bcb7d0172140ec7bc2c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72fc1e209131cea59c8a.md>)

Recorded evaluation configuration: {"Method A":"heuristic","Method B":"holodeck"}. This record does not assert independent reproduction.

### heuristic vs layoutgpt

Score (SceneCritic) Method A: 75.6. Result ID: 69df667247f34ce924c8bee8.

[Reporting paper](<https://www.sota2.com/research/paper/69df1bcb7d0172140ec7bc2c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71021e209131cea4e814.md>)

Recorded evaluation configuration: {"Method A":"heuristic","Method B":"layoutgpt"}. This record does not assert independent reproduction.

### heuristic vs layoutvlm

Score (SceneCritic) Method A: 75.6. Result ID: 69df667247f34ce924c8bee9.

[Reporting paper](<https://www.sota2.com/research/paper/69df1bcb7d0172140ec7bc2c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77581e209131cea72918.md>)

[Implementation](<https://ai.stanford.edu/~sunfanyun/layoutvlm/>)

Recorded evaluation configuration: {"Method A":"heuristic","Method B":"layoutvlm"}. This record does not assert independent reproduction.

### holodeck vs layoutgpt

Score (SceneCritic) Method A: 71.3. Result ID: 69df667247f34ce924c8beea.

[Reporting paper](<https://www.sota2.com/research/paper/69df1bcb7d0172140ec7bc2c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72fc1e209131cea59c8a.md>)

Recorded evaluation configuration: {"Method A":"holodeck","Method B":"layoutgpt"}. This record does not assert independent reproduction.

### holodeck vs layoutvlm

Score (SceneCritic) Method A: 71.3. Result ID: 69df667247f34ce924c8beeb.

[Reporting paper](<https://www.sota2.com/research/paper/69df1bcb7d0172140ec7bc2c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72fc1e209131cea59c8a.md>)

Recorded evaluation configuration: {"Method A":"holodeck","Method B":"layoutvlm"}. This record does not assert independent reproduction.

### layoutgpt vs layoutvlm

Score (SceneCritic) Method A: 53.6. Result ID: 69df667247f34ce924c8beec.

[Reporting paper](<https://www.sota2.com/research/paper/69df1bcb7d0172140ec7bc2c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71021e209131cea4e814.md>)

Recorded evaluation configuration: {"Method A":"layoutgpt","Method B":"layoutvlm"}. This record does not assert independent reproduction.

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