DataVista
Harbor: pendingA benchmark with 961 real-world data videos and 6,775 multiple-choice questions evaluating MLLMs on data perception, temporal reasoning, and narrative understanding.
- Samples
- 6,775
- Released
- 2026
- Scores
- 13
ScenePercept-Bench
ScenePercept: Scenario-Conditioned Perceptual Assessment of Real-World User-Generated Content Images
A benchmark of 23,100 UGC images across 105 scenes for scenario-conditioned perceptual assessment, featuring visual quality, aesthetic, and semantic annotations.
How to run
ltb · scenepercept-bench
$ npm i -g lettokenburn-cli$ ltb bench info scenepercept-bench
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
9 sourced entries| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| GPT-5.3 | 0.8 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| Gemini 3 Flash Preview | 0.7 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| GPT-4o | 0.6 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| Qwen3-VL-32B | 0.5 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| Qwen2.5-VL-32B | 0.5 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| Qwen2.5-VL-7B | 0.5 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| Intern3.5-VL-8B | 0.5 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| Qwen3-VL-8B-Instruct | 0.5 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| Intern3-VL-8B | 0.5 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsLoading community runs…
About
ScenePercept-Bench comprises 23,100 user-generated content images sampled from a large-scale corpus, spanning 21 categories and 105 fine-grained scenarios. It employs a hierarchical annotation protocol covering visual quality, aesthetics, and scene semantics.
Each image includes an overall score, 15 attribute-level scores, and structured explanatory reasoning. The benchmark supports evaluating models' ability to adapt perceptual criteria to specific scene contexts rather than applying uniform standards.