XHotpotQA (Gold Supporting Facts): leaderboard

Metric: Answer F1 (%; one-shot reader given the question and the annotated supporting sentences of the audited V1 validation set, 7,405 mixed-language HotpotQA-derived items, legacy Hotpot answer normalization). Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.

Top models

#ModelScore
1GPT-4o50.82
2Llama 3.1 405B Instruct48.32
3GPT-4o Mini45.13
4Gemma 2 27B (IT)39.54
5Qwen 2 72B Instruct37.95
6aya-23-35B36.92
7Gemma 2 9B (IT)35.32
8Llama 3.1 8B Instruct32.12
9Qwen 2 7B Instruct28.12
10aya-23-8B22.29

Interactive version: theaggregate.ai/benchmark?slug=xhotpotqa-gold-supporting-facts · How It Works · Data refreshed daily, snapshot 2026-09-26.