List your company

Question Answering on MMLU-Pro (full)

61Overall Accuracy (MMLU-Pro QA)MTL
22.93632.81842.752.582May 10, 2026

Evaluation results

MethodLinks
MTLBackbone=Qwen3-4B, Scale=4B2026.056181.667.567.565.170.352.257.250.435.471.351.353.365.769.9
GCWMBackbone=Qwen3-4B, Scale=4B2026.0559.475.166.864.464.268.647.658.242.627.377.547.746.663.463.5
OPCMBackbone=Qwen3-4B, Scale=4B2026.0558.375.265.763.863.567.84757.542.12776.547.24662.663.1
AIMMergingBackbone=Qwen3-4B, Scale=4B2026.0558.174.36663.663.467.846.857.341.726.376.746.945.762.662.7
L&SBackbone=Qwen3-4B, Scale=4B2026.0557.372.764.762.362.266.346.156.341.226.47546.245.161.461.4
EWCBackbone=Qwen3-4B, Scale=4B2026.0555.475.662.957.961.564.547.853.342.525.269.246.545.157.862.9
FOREVERBackbone=Qwen3-4B, Scale=4B2026.0554.772.159.457.458.864.541.253.340.424.77146.745.759.363.5
Seq. SFTBackbone=Qwen3-4B, Scale=4B2026.0551.673.85749.351.961.444.75041.524.662.743.444.553.763
MTLBackbone=Qwen3-0.6B, Scale=0.6B2026.0527.542.133.319.93035.221.122.318.614.736.122.522.728.937.1
GCWMBackbone=Qwen3-0.6B, Scale=0.6B2026.0527.138.929.123.828.334.92221.115.214.34022.221.226.435.1
OPCMBackbone=Qwen3-0.6B, Scale=0.6B2026.0526.842.33318.929.53520.121.417.613.535.921.621.828.436.9
AIMMergingBackbone=Qwen3-0.6B, Scale=0.6B2026.0526.542.232.718.329.234.719.620.91712.835.721.121.32836.8
L&SBackbone=Qwen3-0.6B, Scale=0.6B2026.0526.241.232.218.428.734.119.620.817.113.13521.121.227.736
EWCBackbone=Qwen3-0.6B, Scale=0.6B2026.0524.936.730.922.423.927.819.22116.512.935.921.11825.928.4
FOREVERBackbone=Qwen3-0.6B, Scale=0.6B2026.0524.736.328.421.722.930.919.719.715.811.536.319.82026.429.6
Seq. SFTBackbone=Qwen3-0.6B, Scale=0.6B2026.0524.433.928.120.622.92917.323.518.914.433.221.417.223.434.3