M3R-Bench - Sentiment Classification: leaderboard
M3R-Bench (Jiang et al., 2026) contains 1,000 human-annotated image-text pairs from social media and advertisements for multimodal metaphor understanding, where each model jointly predicts metaphor presence, target and source concepts and sentiment. This board scores three-way sentiment classification by macro-F1 in percent; higher is better.
Metric: Macro-F1 (%; positive, neutral or negative sentiment of the image-text pair, jointly predicted). Source: arxiv.org. Saturation forecast: Around March 2028. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 64.69 | #741 |
| 2 | GPT-5.5 | 61.77 | #48 |
| 3 | Claude Sonnet 4.6 | 57.11 | #158 |
| 4 | Kimi K2.5 | 52.96 | #252 |
| 5 | Qwen 3 VL 32B Instruct | 52.49 | #582 |
| 6 | Qwen 3 VL 8B Instruct | 48.22 | #982 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=m3r-bench-sentiment-classification · How It Works · Data refreshed daily, snapshot 2026-11-12.