MUN-vis (5-Shot): leaderboard

Metric: Win rate (%; share of MUN-vis items where a GPT-4o judge, AlpacaEval-style, prefers the model's explanation (explain why a visually odd image leads to an ordinary outcome) over the human-written, GPT-4o-refined reference; five randomly chosen in-context examples). Source: arxiv.org. Saturation forecast: Estimated already saturated. 7 models tracked.

Top models

#ModelScore
1Phi-4 Multimodal Instruct63
2Qwen 2.5 VL 7B Instruct43.9
3Gemma 3 4B (IT)39.3
4Qwen 2 VL 7B Instruct27.2

Interactive version: theaggregate.ai/benchmark?slug=mun-vis-5-shot · How It Works · Data refreshed daily, snapshot 2026-09-26.