DiffCap-Bench - Attribute: leaderboard
DiffCap-Bench (2026) image difference captioning restricted to attribute modifications such as size or colour. Macro-averaged F1* in percent, combining the share of true differences described and the share of described differences that are correct, as judged by Gemini 2.5 Pro; higher is better.
Metric: Macro-averaged F1* (%) on attribute modifications such as size or colour: harmonic mean of recall* (share of key differences described) and precision* (share of described differences that are correct) on the 1,075 image pairs of DiffCap-Bench (6,713 human-verified difference items), one fixed captioning prompt, open models at temperature 0.1 and proprietary models at their defaults, a Gemini 2.5 Pro judge matching each described difference against the key difference list; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 14 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Qwen 3 VL Plus | 82.4 | #514 | |
| 2 | Gemini 3 Pro | 80.8 | #146 | |
| 3 | Kimi K2.5 | Thinking | 80 | #272 |
| 4 | Seed1.6-VL | Thinking | 79.2 | #389 |
| 5 | Qwen 3 VL 32B (Thinking) | 78.6 | #577 | |
| 6 | GPT-5.2 | 77.1 | #190 | |
| 7 | MiMo-VL-7B-RL-2508 | 74.9 | #829 | |
| 8 | Qwen 3 VL 8B (Thinking) | 73.3 | #924 | |
| 9 | Step3 VL 10B | 70.9 | #1058 | |
| 10 | Qwen 3 VL 8B Instruct | 70.4 | #1043 | |
| 11 | Grok 4.20 | 70.3 | #307 | |
| 12 | Qwen 3 VL 32B Instruct | 68.8 | #607 | |
| 13 | InternVL3.5-38B | 63.5 | #1026 | |
| 14 | InternVL3.5-8B | 62 | #1618 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=diffcap-bench-attribute · How It Works · Data refreshed daily, snapshot 2026-11-22.