List your company

Instruction Tuning on TyDiQA, MMLU, and BBH Aggregate

57.6TyDiQA ScoreSEED
31.80838.50445.251.896May 15, 2026

Evaluation results

MethodLinks
SEEDModel Backbone=LLAMA-3-8B, Selection Ratio=5%2026.0557.665.166.957.3
SEEDModel Backbone=LLAMA-3-8B, Selection Ratio=1%2026.0556.166.669.958.2
NAITModel Backbone=LLAMA-3-8B, Selection Ratio=5%2026.0555.964.466.156.2
INFOMAXModel Backbone=LLAMA-3-8B, Selection Ratio=5%2026.0555.764.566.156.4
LESSModel Backbone=LLAMA-3-8B, Selection Ratio=5%2026.0555.464.465.755.9
NAITModel Backbone=LLAMA-3-8B, Selection Ratio=1%2026.0555.465.167.456.8
TAROTModel Backbone=LLAMA-3-8B, Selection Ratio=5%2026.0555.264.76656.2
FULLModel Backbone=LLAMA-3-8B, Selection Ratio=100%2026.0555.165.167.357.3
MONAModel Backbone=LLAMA-3-8B, Selection Ratio=1%2026.055565.468.157.2
D2PRUNINGModel Backbone=LLAMA-3-8B, Selection Ratio=5%2026.0554.864.26555.6
D2PRUNINGModel Backbone=LLAMA-3-8B, Selection Ratio=1%2026.0554.765.567.356.6
INFOMAXModel Backbone=LLAMA-3-8B, Selection Ratio=1%2026.0554.665.868.257.1
MONAModel Backbone=QWEN3-4B, Selection Ratio=1%2026.0554.46974.357.2
SEEDModel Backbone=QWEN3-4B, Selection Ratio=1%2026.0554.470.475.958.2
LESSModel Backbone=LLAMA-3-8B, Selection Ratio=1%2026.0554.165.667.856.5
NAITModel Backbone=QWEN3-4B, Selection Ratio=1%2026.0553.369.273.656.8
INFOMAXModel Backbone=QWEN3-4B, Selection Ratio=1%2026.0553.169.374.257.1
LESSModel Backbone=QWEN3-4B, Selection Ratio=1%2026.0552.66973.856.5
RANDOMModel Backbone=LLAMA-3-8B, Selection Ratio=1%2026.0552.365.366.356
D2PRUNINGModel Backbone=QWEN3-4B, Selection Ratio=1%2026.0552.368.97456.6
RANDOMModel Backbone=QWEN3-4B, Selection Ratio=1%2026.055268.773.956
RANDOMModel Backbone=LLAMA-3-8B, Selection Ratio=5%2026.0551.66463.154.6
SEEDModel Backbone=QWEN3-4B, Selection Ratio=5%2026.0550.669.572.657.3
NAITModel Backbone=QWEN3-4B, Selection Ratio=5%2026.0550.268.570.856.2
INFOMAXModel Backbone=QWEN3-4B, Selection Ratio=5%2026.0549.969.171.356.4
LESSModel Backbone=QWEN3-4B, Selection Ratio=5%2026.0549.668.870.655.9
TAROTModel Backbone=QWEN3-4B, Selection Ratio=5%2026.0549.6697156.2
FULLModel Backbone=QWEN3-4B, Selection Ratio=100%2026.0549.57072.957.3
D2PRUNINGModel Backbone=QWEN3-4B, Selection Ratio=5%2026.0549.368.570.455.6
RANDOMModel Backbone=QWEN3-4B, Selection Ratio=5%2026.0548.568.770.354.6
BASEModel Backbone=QWEN3-4B, Selection Ratio=0%2026.0547.467.97053
BASEModel Backbone=LLAMA-3-8B, Selection Ratio=0%2026.0547.363.362.253
SEEDModel Backbone=LLAMA-2-7B, Selection Ratio=1%2026.0542.546.941.458.2
SEEDModel Backbone=LLAMA-2-7B, Selection Ratio=5%2026.0541.949.641.957.3
MONAModel Backbone=LLAMA-2-7B, Selection Ratio=1%2026.0540.9474157.2
INFOMAXModel Backbone=LLAMA-2-7B, Selection Ratio=1%2026.0540.846.541.557.1
FULLModel Backbone=LLAMA-2-7B, Selection Ratio=100%2026.0540.550.444.857.3
LESSModel Backbone=LLAMA-2-7B, Selection Ratio=1%2026.0540.345.840.156.5
NAITModel Backbone=LLAMA-2-7B, Selection Ratio=1%2026.0540.146.240.756.8
D2PRUNINGModel Backbone=LLAMA-2-7B, Selection Ratio=1%2026.054046.340.656.6
NAITModel Backbone=LLAMA-2-7B, Selection Ratio=5%2026.0539.94941.356.2
TAROTModel Backbone=LLAMA-2-7B, Selection Ratio=5%2026.0539.749.540.956.2
INFOMAXModel Backbone=LLAMA-2-7B, Selection Ratio=5%2026.0539.650.141.556.4
RANDOMModel Backbone=LLAMA-2-7B, Selection Ratio=1%2026.0539.646.139.856
LESSModel Backbone=LLAMA-2-7B, Selection Ratio=5%2026.0539.349.340.655.9
D2PRUNINGModel Backbone=LLAMA-2-7B, Selection Ratio=5%2026.053948.940.355.6
RANDOMModel Backbone=LLAMA-2-7B, Selection Ratio=5%2026.0538.546.84054.6
BASEModel Backbone=LLAMA-2-7B, Selection Ratio=0%2026.0532.845.739.553