List your company

Forgetting-aware Instruction Tuning on Magicoder Stability and Plasticity suites (test)

54.27ARC-CPre-trained
39.418843.274447.1350.9856May 7, 2026

Evaluation results

MethodLinks
Pre-trainedBackbone=Qwen-3.5-9B-Base2026.0554.2780.0972.6286.273.360.9867.14
DualSFTBackbone=Qwen-3.5-9B-Base2026.0554.0179.6470.3384.6172.1573.7572.95
DualSFT-DataBackbone=Qwen-3.5-9B-Base2026.0553.6680.0370.7585.0472.3774.2173.29
DualSFT-ParamBackbone=Qwen-3.5-9B-Base2026.0553.5278.7270.2684.2271.6875.1373.41
TALRBackbone=Qwen-3.5-9B-Base2026.0553.4579.770.7285.7372.475.3673.88
FLOWBackbone=Qwen-3.5-9B-Base2026.0553.279.5670.6285.2872.1773.172.64
NanoAdamBackbone=Qwen-3.5-9B-Base2026.0552.8779.0370.2284.2271.5973.2872.43
FAPMBackbone=Qwen-3.5-9B-Base2026.0552.8679.4870.1883.8871.670.8571.23
STMBackbone=Qwen-3.5-9B-Base2026.0552.879.1870.3884.5571.7373.1272.43
DavIRBackbone=Qwen-3.5-9B-Base2026.0552.478.4270.0283.671.1173.472.26
LoRABackbone=Qwen-3.5-9B-Base2026.0551.5176.6269.4781.9469.8970.2970.09
Pre-trainedBackbone=Gemma-3-4B-PT2026.0551.4579.1659.613756.8135.3746.09
DualSFTBackbone=Gemma-3-4B-PT2026.0551.1479.0159.1635.6256.2356.7656.5
DualSFT-ParamBackbone=Gemma-3-4B-PT2026.0551.1378.5458.1436.3456.0457.5256.78
DualSFT-DataBackbone=Gemma-3-4B-PT2026.0551.0178.5258.3336.5556.156.9456.52
TALRBackbone=Gemma-3-4B-PT2026.0550.9678.8458.9436.756.3657.4456.9
FLOWBackbone=Gemma-3-4B-PT2026.0550.9278.8258.9236.8256.3755.756.04
NanoAdamBackbone=Gemma-3-4B-PT2026.0550.7878.258.3333.1255.1156.3855.74
STMBackbone=Gemma-3-4B-PT2026.0550.6278.4858.5836.155.9555.9255.94
FAPMBackbone=Gemma-3-4B-PT2026.0550.678.558.635.955.952.254.05
DavIRBackbone=Gemma-3-4B-PT2026.0550.2578.0458.0133.75556.2455.62
Standard SFTBackbone=Qwen-3.5-9B-Base2026.0549.2274.3166.1478.6467.0875.5671.32
LoRABackbone=Gemma-3-4B-PT2026.0549.1575.5756.7131.6353.2754.6653.96
Standard SFTBackbone=Gemma-3-4B-PT2026.0547.1776.3953.8724.0350.3758.154.23
Pre-trainedBackbone=Llama-3.2-3B2026.054376.6156.5227.4550.928.6639.78
FLOWBackbone=Llama-3.2-3B2026.0542.9476.2356.9127.2850.8442.2846.56
DualSFTBackbone=Llama-3.2-3B2026.0542.9476.2356.0926.8850.5443.6747.1
DavIRBackbone=Llama-3.2-3B2026.0542.7174.9253.0924.6548.8443.2846.06
NanoAdamBackbone=Llama-3.2-3B2026.0542.6674.6852.7124.3948.6143.5446.08
DualSFT-DataBackbone=Llama-3.2-3B2026.0542.6676.4455.2826.5450.2344.0547.14
LoRABackbone=Llama-3.2-3B2026.0542.6174.4351.8723.2948.0541.4644.76
FAPMBackbone=Llama-3.2-3B2026.0542.5576.0153.8626.0149.6140.9945.3
DualSFT-ParamBackbone=Llama-3.2-3B2026.0542.5476.1954.0126.2749.7544.5247.14
TALRBackbone=Llama-3.2-3B2026.0542.4976.3456.4127.5450.744.1847.44
STMBackbone=Llama-3.2-3B2026.0542.3876.1256.0427.0850.4142.6546.53
Standard SFTBackbone=Llama-3.2-3B2026.0539.9972.0948.1820.6945.2444.6844.96