List your company

Clinical Contextual Question Answering on Finnish breast cancer EHR dataset N=1,664 1.0 (Full)

95.3V1 AccuracyLlama-3.1-70B
43.40456.87770.3583.823Mar 27, 2026

Evaluation results

MethodLinks
Llama-3.1-70BDeveloper=Meta, VRAM (GB) Min=132.5, VRAM (GB) Max=148.2, Latency (s) Min=1.2, Latency (s) Max=19.72026.0395.394.297.3
Qwen3-30B-A3B-2507Developer=Qwen, VRAM (GB) Min=57.2, VRAM (GB) Max=61.8, Latency (s) Min=2.7, Latency (s) Max=8.92026.0395.194.497.7
Llama-3.3-70BDeveloper=Meta, VRAM (GB) Min=132.5, VRAM (GB) Max=148.2, Latency (s) Min=1.2, Latency (s) Max=19.82026.0394.994.997.9
R1-Distill-Llama-70BDeveloper=DeepSeek, VRAM (GB) Min=132.8, VRAM (GB) Max=148.2, Latency (s) Min=18.7, Latency (s) Max=153.32026.0394.895.197
Gemma-3-27B-itDeveloper=Google, VRAM (GB) Min=52.4, VRAM (GB) Max=71.9, Latency (s) Min=0.7, Latency (s) Max=16.62026.0392.792.896.3
R1-Distill-Qwen-32BDeveloper=DeepSeek, VRAM (GB) Min=62.1, VRAM (GB) Max=77.7, Latency (s) Min=11.0, Latency (s) Max=91.32026.0392.794.694.7
MedGemma-27B-text-itDeveloper=Google, VRAM (GB) Min=51.6, VRAM (GB) Max=71.2, Latency (s) Min=0.7, Latency (s) Max=16.72026.0392.19094
Qwen3-4B-2507Developer=Qwen, VRAM (GB) Min=8.0, VRAM (GB) Max=15.0, Latency (s) Min=0.2, Latency (s) Max=2.72026.0391.189.895.3
Llama-3.1-8BDeveloper=Meta, VRAM (GB) Min=15.5, VRAM (GB) Max=22.7, Latency (s) Min=0.2, Latency (s) Max=2.72026.0379.980.294.2
MedGemma-4B-itDeveloper=Google, VRAM (GB) Min=8.5, VRAM (GB) Max=18.4, Latency (s) Min=0.3, Latency (s) Max=5.92026.0374.573.189.6
Gemma-3-4B-itDeveloper=Google, VRAM (GB) Min=8.5, VRAM (GB) Max=18.4, Latency (s) Min=0.2, Latency (s) Max=5.82026.0372.668.283.2
R1-Distill-Qwen-7BDeveloper=DeepSeek, VRAM (GB) Min=14.7, VRAM (GB) Max=22.6, Latency (s) Min=5.7, Latency (s) Max=41.42026.0350.451.243.4
Rule-based (substring)Developer=Baseline, Latency (s) Min=<0.1, Latency (s) Max=<0.12026.0345.445.4100