XCR-Bench - CSI Prediction: leaderboard

Metric: Exact Match (x100; case-insensitive exact matches of the masked culture-specific items, summed per sentence and averaged over sentences). Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked.

Top models

#ModelScore
1Claude 3.7 Sonnet46.76
2GPT-4o43.7
3Qwen 2.5 Max43.7
4Gemini 2.0 Flash42.94
5Llama 3.3 70B Instruct34.54
6OLMo 2 32B31.87
7DeepSeek R1 Distill Qwen 32B29.96

Interactive version: theaggregate.ai/benchmark?slug=xcr-bench-csi-prediction · How It Works · Data refreshed daily, snapshot 2026-09-25.