# Edit Inspection Questions on Imagen3 edits Source: [Edit Inspection Questions on Imagen3 edits]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:27:00.150Z Evaluation of image edit quality based on specific inspector questions applied to the Imagen3 edits dataset, using metrics like Accuracy, Contextual Consistency, and Technical Precision. 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 | Edit Inspection Questions | | Dataset (including any stated split) | Imagen3 edits | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 7 of 7 | | 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 - [Accuracy](): Higher is better; unit: PERCENT. - [Contextual Consistency](): Higher is better; unit: PERCENT. - [Technical Precision](): Higher is better; unit: PERCENT. - [Artifacts Score](): Lower is better; unit: PERCENT. - [Diff Caption Acc](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | LLaVA | 58.8 | Not recorded | 68fa79651e209131cea7d6a6 | | GPT-4 | 55.3 | Not recorded | 68fa79651e209131cea7d6a6 | | LLaVA | 54.4 | Not recorded | 68fa79651e209131cea7d6a6 | | Qwen2.5 VL | 52.9 | Not recorded | 68fa79651e209131cea7d6a6 | | InternVL3 | 51.5 | Not recorded | 68fa79651e209131cea7d6a6 | | GPT-4o | 49.9 | Not recorded | 68fa79651e209131cea7d6a6 | | GPT-4 Turbo | 49.2 | Not recorded | 68fa79651e209131cea7d6a6 | ### LLaVA Accuracy: 58.8. Result ID: 698b4b8e8d67a41d0a6ce167. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"training":"supervised"}. This record does not assert independent reproduction. ### GPT-4 Accuracy: 55.3. Result ID: 698b4b8e8d67a41d0a6ce161. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### LLaVA Accuracy: 54.4. Result ID: 698b4b8e8d67a41d0a6ce166. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Qwen2.5 VL Accuracy: 52.9. Result ID: 698b4b8e8d67a41d0a6ce164. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### InternVL3 Accuracy: 51.5. Result ID: 698b4b8e8d67a41d0a6ce165. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### GPT-4o Accuracy: 49.9. Result ID: 698b4b8e8d67a41d0a6ce162. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### GPT-4 Turbo Accuracy: 49.2. Result ID: 698b4b8e8d67a41d0a6ce163. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.