MemeBridge: leaderboard
MemeBridge (Zhu et al., 2026) tests cross-cultural understanding of 621 U.S.-originated memes with explanations, anticipated misunderstandings and sentiment and emotion labels from U.S. crowdworkers. The performance score, out of 5, adds the cosine similarity of the model explanation to the original and to a GPT-4-rewritten explanation and the multiple-choice, sentiment and emotion accuracies scaled so that chance performance on each task scores about 1, under the default prompt; higher is better.
Metric: Performance score (out of 5) summing the cosine similarity of the generated explanation to the original and to the GPT-4-rewritten explanation plus the multiple-choice, sentiment and emotion accuracies scaled so that each chance-level task scores about 1 (621 U.S.-originated memes with U.S. crowd labels, default prompt (no role-play), the same test given to Chinese participants). Source: arxiv.org. Saturation forecast: Around February 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 3.24 | #741 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=memebridge · How It Works · Data refreshed daily, snapshot 2026-11-12.