Citation and Evidence Recall on RepliQA benchmark leaderboard · SOTA2 Research

Citation and Evidence Recall on RepliQA

99.5RkGPT-5.2

Evaluation Results

Method	Links
GPT-5.22025.10		99.5	82.8	100
Oracle-DRAG2025.10		99.4	100	99.4
Mistr-S-3.2-24B2025.10		99.2	85.2	99.9
GPT-5-Mini2025.10		99	81.1	99.6
Llama-3.1-8B2025.10		98.9	78.2	99.6
Qwen3-32B2025.10		98.7	80.5	99.8
GPT-OSS-120B2025.10		98.6	78.2	99.6
Oracle-BM252025.10		98.5	100	98.5
Qwen3-14B2025.10		98.1	81.2	99.6
Phi-42025.10		97.8	80.6	99.2
Llama-3.3-70B2025.10		97	79.4	98.2
Qwen3-4B2025.10		97	71.2	99
Ministr-8B2025.10		96.9	73.2	98.3
Qwen3-8B2025.10		96.9	80.2	98.6
GPT-OSS-20B2025.10		96.1	77	97.6
Llama-3.2-3B2025.10		96	73	97.1
Phi-4-Mini2025.10		93.7	68.9	95.4
Qwen3-1.7B2025.10		91.5	70.5	94.5
Qwen3-0.6B2025.10		75.9	14.2	83.5
Llama-3.2-1B2025.10		39.1	63.1	41.4