Interactive Visual Grounding (Static): leaderboard
The Interactive Visual Grounding benchmark (2026) casts a vision-language model as the Matcher in a reference game: a simulated Director describes target items drawn from human-grounded image sets of dogs, baskets and tangrams, and the model must pick each target from a grid, asking clarification questions where the protocol allows. In the Static protocol the Matcher gets the full human-produced description, no interaction. Score is the share of targets matched correctly, in percent; higher is better.
Metric: Matcher accuracy (%; share of target items the model selects correctly given the full human-produced description, no interaction; micro-averaged over 536 target items of four human-grounded image sets (dogs, two basket sets, tangrams), with GPT-5.4 as the simulated Director answering from the human description; a 10-turn budget per target, utterances capped at 50 words). Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 87.5 | #94 |
| 2 | Gemini 3.1 Flash Lite (Preview) | 67.8 | #418 |
| 3 | GPT-5.4 (2026-03-05) | 66.1 | #87 |
| 4 | GPT-5.5 (2026-04-23) | 65.8 | #33 |
| 5 | Gemma 4 31B (IT) | 56 | #342 |
| 6 | Qwen 3 VL 32B Instruct | 48.8 | #585 |
| 7 | GPT-5.4 Mini (2026-03-17) | 47.9 | #287 |
| 8 | Qwen 3 VL 8B Instruct | 33.6 | #981 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=interactive-visual-grounding-static · How It Works · Data refreshed daily, snapshot 2026-11-13.