ST-Bench (Image Input) - Spatial Entity Identification: leaderboard
Spatial entity identification task of ST-Bench (Ni, Wang, He, Jin and Jin, STReasoner paper, 2026): given the multivariate time series of a node graph, the model picks the semantic role of a node (for example the terminal monitoring point) from four options. The model sees the node series rendered as plot images with the graph structure in text; scenarios are synthetic network simulations and Claude-4.5-Sonnet wrote the questions. Accuracy in percent on 1,194 held-out questions; higher is better.
Metric: Accuracy (%; 1,194 four-option questions). Source: arxiv.org. Saturation forecast: Around May 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 41.85 | #244 |
| 2 | GPT-5.2 | 40.54 | #172 |
| 3 | Qwen 3 VL 8B Instruct | 31.16 | #982 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=st-bench-image-input-spatial-entity-identification · How It Works · Data refreshed daily, snapshot 2026-11-12.