List your company

Language Understanding on MMLU-Pro OOD (test)

80.92Biology Avg@16CAST
78.621679.218379.81580.4117May 29, 2026

Evaluation results

MethodLinks
CASTBackbone=Qwen3-4B, Generation token budget=8K2026.0580.9294.2873.7289.7375.4393.1170.988.7873.8889.4555.1389.4759.5979.3448.1670.0830.8164.8585.2195.1955.176.5253.0875.3575.2792.366.7282.3365.7485.43
RLRTBackbone=Qwen3-4B, Generation token budget=8K2026.0579.3893.0370.5490.1171.3691.4368.9987.5671.987.5653.7787.6257.9478.8547.7770.3429.5362.4981.394.353.4576.6251.1972.9571.4391.6966.1781.0863.4184.32
GRPOBackbone=Qwen3-4B, Generation token budget=8K2026.0579.3193.0370.3288.9770.5691.1768.6388.7872.0388.9853.5988.3457.9379.5847.2870.8729.6264.0380.8393.9353.2774.8951.0173.3571.2191.7666.0381.763.1884.52
RLSDBackbone=Qwen3-4B, Generation token budget=8K2026.0579.1593.0370.5689.3570.7692.468.5787.5672.0488.5154.3788.2457.6478.8547.4770.0829.5862.3180.7594.4553.3876.0850.5672.5570.7891.6965.8881.0863.1784.41
BaseBackbone=Qwen3-4B, Generation token budget=8K2026.0579.0293.0370.0889.4870.5692.1467.7988.7872.0788.3954.3587.7258.217847.8568.7729.6762.480.4694.0853.3475.8751.5874.3570.7891.6165.9782.0863.1684.37
OPSDBackbone=Qwen3-4B, Generation token budget=8K2026.0578.7192.6170.6688.2170.4391.6168.0587.5671.9188.2753.9387.7257.8478.6147.969.5529.7962.5880.6794.353.2475.4350.7473.5570.9891.6166.1581.3363.1384.15