List your company

Scientific Reasoning on GPQA (Subject Scores)

35.7Biology ScoreEasyRL Iter3
-1.4288.21117.8527.489Apr 19, 2026

Evaluation results

MethodLinks
EasyRL Iter3Backbone=Qwen2.5-Math-1.5B2026.0435.717.712.919.4
EasyRL Iter3Backbone=Llama-3.2-3B-Instruct2026.0435.723.822.127.2
Supervised GRPOBackbone=Llama-3.2-3B-Instruct2026.0434.316.219.323.3
EasyRL Iter2Backbone=Llama-3.2-3B-Instruct2026.0434.321.519.325
EasyRL Iter1Backbone=Llama-3.2-3B-Instruct2026.0432.916.923.624.5
EasyRL Iter3Backbone=Qwen2.5-Math-7B2026.0431.424.635.730.6
Unsupervised EMPOBackbone=Llama-3.2-3B-Instruct2026.043017.713.618.5
Vanilla InstructBackbone=Qwen2.5-Math-7B2026.0428.623.831.427.9
EasyRL Iter1Backbone=Qwen2.5-Math-7B2026.0428.623.127.926.2
EasyRL Iter2Backbone=Qwen2.5-Math-7B2026.0428.620.837.129.1
EasyRL Iter1Backbone=Qwen2.5-Math-1.5B2026.0427.114.610.715.6
Vanilla BaseBackbone=Llama-3.2-3B-Instruct2026.0427.117.71518.5
Unsupervised EMPOBackbone=Qwen2.5-Math-7B2026.0425.723.13026.5
EasyRL Iter2Backbone=Qwen2.5-Math-1.5B2026.0424.320.811.417.6
Supervised GRPOBackbone=Qwen2.5-Math-7B2026.0424.323.132.927.4
Unsupervised EMPOBackbone=Qwen2.5-Math-1.5B2026.042016.912.115.6
Vanilla BaseBackbone=Qwen2.5-Math-7B2026.0418.624.626.424.1
Supervised GRPOBackbone=Qwen2.5-Math-1.5B2026.0411.4104.37.9
Vanilla BaseBackbone=Qwen2.5-Math-1.5B2026.0402.31.41.5