JaWildText - Dense STVQA: leaderboard
Metric: Accuracy (%) on 1,025 open-ended questions about 745 photos of dense Japanese scene text (signboards, bulletin boards, posters, packaging), answers in \boxed{} judged correct or not by a GPT-5.1 verifier, zero-shot open-weight VLM inference at temperature 0 with up to 2,048 output tokens, one image at its original resolution, a fixed prompt per task and unparseable outputs scored 0; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 3 27B (IT) | 37 | #509 |
| 2 | Gemma 3 12B (IT) | 32 | #655 |
| 3 | Gemma 3 4B (IT) | 12 | #971 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=jawildtext-dense-stvqa · How It Works · Data refreshed daily, snapshot 2026-10-11.