Rubric Generation on OpenAI HealthBench OSS EVAL-300 benchmark leaderboard · SOTA2 Research

Rubric Generation on OpenAI HealthBench OSS EVAL-300

4.9BLEURubricRAG

Evaluation Results

Method	Links
RubricRAG2026.03		4.9	33.1	11.5	26.9	0.567
SFT2026.03		4.2	31.1	10.3	25.3	0.481
RubricRAG2026.03		3.7	30.4	9.7	24.5	0.558
GRPO2026.03		3.5	31.1	10.1	24.8	0.551
Few-Shot2026.03		3.3	29.3	9.1	23.7	0.548
Zero-Shot2026.03		2	23.1	6.5	19.2	0.521