List your company

Language Model Evaluation on BenchPress short-context (test)

68.84AccuracyQwen3-8B
32.065641.612851.1660.7072Oct 23, 2025

Evaluation results

MethodLinks
Qwen3-8BCompression Ratio=Original2025.1068.84
Qwen3-4BCompression Ratio=Original2025.1067.69
Gemma2-2BCompression Ratio=Original2025.1066.14
Mean-PoolingBackbone=Qwen3-8B, Compression Ratio=4x, Aggregation Mode=Single2025.1065.91
Mean-PoolingBackbone=Qwen3-8B, Compression Ratio=4x, Aggregation Mode=Multi2025.1065.06
Qwen3-1.7BCompression Ratio=Original2025.1064.21
Mean-PoolingBackbone=Qwen3-4B, Compression Ratio=4x, Aggregation Mode=Single2025.1064.05
Compression-Tokens (Bidirectional)Backbone=Qwen3-8B, Compression Ratio=4x, Aggregation Mode=Multi2025.1063.58
Mean-PoolingBackbone=Qwen3-4B, Compression Ratio=4x, Aggregation Mode=Multi2025.1062.94
Mean-PoolingBackbone=Gemma2-2B, Compression Ratio=4x, Aggregation Mode=Single2025.1062.51
Compression-Tokens (Bidirectional)Backbone=Qwen3-8B, Compression Ratio=4x, Aggregation Mode=Single2025.1062.5
Mean-PoolingBackbone=Gemma2-2B, Compression Ratio=4x, Aggregation Mode=Multi2025.1061.28
Compression-Tokens (Bidirectional)Backbone=Qwen3-4B, Compression Ratio=4x, Aggregation Mode=Multi2025.1060.99
Llama3.2-1BCompression Ratio=Original2025.1060.3
Mean-PoolingBackbone=Qwen3-1.7B, Compression Ratio=4x, Aggregation Mode=Single2025.1060.03
Compression-Tokens (Causal)Backbone=Qwen3-8B, Compression Ratio=4x, Aggregation Mode=Single2025.1059.79
Qwen3-0.6BCompression Ratio=Original2025.1059.67
Compression-Tokens (Bidirectional)Backbone=Qwen3-4B, Compression Ratio=4x, Aggregation Mode=Single2025.1059.35
Compression-Tokens (Causal)Backbone=Qwen3-8B, Compression Ratio=4x, Aggregation Mode=Multi2025.1058.58
Compression-Tokens (Bidirectional)Backbone=Gemma2-2B, Compression Ratio=4x, Aggregation Mode=Multi2025.1057.54
Mean-PoolingBackbone=Qwen3-1.7B, Compression Ratio=4x, Aggregation Mode=Multi2025.1057.28
Compression-Tokens (Causal)Backbone=Qwen3-4B, Compression Ratio=4x, Aggregation Mode=Single2025.1057.12
Compression-Tokens (Bidirectional)Backbone=Gemma2-2B, Compression Ratio=4x, Aggregation Mode=Single2025.1056.94
Mean-PoolingBackbone=Qwen3-8B, Compression Ratio=16x, Aggregation Mode=Multi2025.1056.77
Mean-PoolingBackbone=Llama3.2-1B, Compression Ratio=4x, Aggregation Mode=Single2025.1056.62
Mean-PoolingBackbone=Qwen3-8B, Compression Ratio=16x, Aggregation Mode=Single2025.1055.68
Compression-Tokens (Causal)Backbone=Gemma2-2B, Compression Ratio=4x, Aggregation Mode=Single2025.1055.5
Compression-Tokens (Bidirectional)Backbone=Qwen3-8B, Compression Ratio=16x, Aggregation Mode=Multi2025.1055.26
PCC LargeBackbone=Llama3.1-8B, Compression Ratio=4x, Aggregation Mode=Single2025.1055.17
Compression-Tokens (Bidirectional)Backbone=Qwen3-1.7B, Compression Ratio=4x, Aggregation Mode=Multi2025.1054.7
Mean-PoolingBackbone=Qwen3-0.6B, Compression Ratio=4x, Aggregation Mode=Single2025.1054.36
Compression-Tokens (Bidirectional)Backbone=Qwen3-1.7B, Compression Ratio=4x, Aggregation Mode=Single2025.1054.31
Compression-Tokens (Causal)Backbone=Qwen3-4B, Compression Ratio=4x, Aggregation Mode=Multi2025.1054.17
PCC LiteBackbone=GPT2-Large & Llama3.1-8B, Compression Ratio=4x, Aggregation Mode=Single2025.1054.05
PISCOBackbone=Llama3.1-8B, Compression Ratio=16x, Aggregation Mode=Single2025.1053.62
Mean-PoolingBackbone=Llama3.2-1B, Compression Ratio=4x, Aggregation Mode=Multi2025.1053.5
Compression-Tokens (Causal)Backbone=Gemma2-2B, Compression Ratio=4x, Aggregation Mode=Multi2025.1052.94
Mean-PoolingBackbone=Qwen3-4B, Compression Ratio=16x, Aggregation Mode=Multi2025.1052.82
Mean-PoolingBackbone=Qwen3-4B, Compression Ratio=16x, Aggregation Mode=Single2025.1052.77
Mean-PoolingBackbone=Gemma2-2B, Compression Ratio=16x, Aggregation Mode=Single2025.1052.55
Compression-Tokens (Bidirectional)Backbone=Qwen3-4B, Compression Ratio=16x, Aggregation Mode=Multi2025.1052.23
Compression-Tokens (Bidirectional)Backbone=Qwen3-8B, Compression Ratio=16x, Aggregation Mode=Single2025.1052.22
Mean-PoolingBackbone=Gemma2-2B, Compression Ratio=16x, Aggregation Mode=Multi2025.1051.9
Mean-PoolingBackbone=Qwen3-0.6B, Compression Ratio=4x, Aggregation Mode=Multi2025.1051.16
Compression-Tokens (Bidirectional)Backbone=Llama3.2-1B, Compression Ratio=4x, Aggregation Mode=Single2025.1050.89
Compression-Tokens (Bidirectional)Backbone=Llama3.2-1B, Compression Ratio=4x, Aggregation Mode=Multi2025.1050.1
Compression-Tokens (Causal)Backbone=Qwen3-8B, Compression Ratio=16x, Aggregation Mode=Multi2025.1049.99
Compression-Tokens (Bidirectional)Backbone=Gemma2-2B, Compression Ratio=16x, Aggregation Mode=Multi2025.1049.52
Compression-Tokens (Causal)Backbone=Qwen3-1.7B, Compression Ratio=4x, Aggregation Mode=Multi2025.1049.31
ICAEBackbone=Mistral-7B, Compression Ratio=4x, Aggregation Mode=Single2025.1049.18
Compression-Tokens (Bidirectional)Backbone=Qwen3-4B, Compression Ratio=16x, Aggregation Mode=Single2025.1048.92
Compression-Tokens (Causal)Backbone=Llama3.2-1B, Compression Ratio=4x, Aggregation Mode=Single2025.1048.85
Compression-Tokens (Bidirectional)Backbone=Qwen3-0.6B, Compression Ratio=4x, Aggregation Mode=Multi2025.1048.84
Compression-Tokens (Bidirectional)Backbone=Qwen3-0.6B, Compression Ratio=4x, Aggregation Mode=Single2025.1048.04
Compression-Tokens (Causal)Backbone=Qwen3-8B, Compression Ratio=16x, Aggregation Mode=Single2025.1047.07
Compression-Tokens (Bidirectional)Backbone=Gemma2-2B, Compression Ratio=16x, Aggregation Mode=Single2025.1046.9
Mean-PoolingBackbone=Qwen3-1.7B, Compression Ratio=16x, Aggregation Mode=Single2025.1046.63
Compression-Tokens (Causal)Backbone=Qwen3-0.6B, Compression Ratio=4x, Aggregation Mode=Single2025.1046.36
Compression-Tokens (Causal)Backbone=Gemma2-2B, Compression Ratio=16x, Aggregation Mode=Single2025.1046.13
Compression-Tokens (Causal)Backbone=Qwen3-4B, Compression Ratio=16x, Aggregation Mode=Single2025.1046.06
Compression-Tokens (Causal)Backbone=Gemma2-2B, Compression Ratio=16x, Aggregation Mode=Multi2025.1045.59
Compression-Tokens (Causal)Backbone=Qwen3-4B, Compression Ratio=16x, Aggregation Mode=Multi2025.1045.47
Compression-Tokens (Causal)Backbone=Llama3.2-1B, Compression Ratio=4x, Aggregation Mode=Multi2025.1045.41
Compression-Tokens (Bidirectional)Backbone=Qwen3-1.7B, Compression Ratio=16x, Aggregation Mode=Multi2025.1045.13
Mean-PoolingBackbone=Qwen3-1.7B, Compression Ratio=16x, Aggregation Mode=Multi2025.1045.07
PCC LiteBackbone=GPT2-Large & Llama3.1-8B, Compression Ratio=16x, Aggregation Mode=Single2025.1043.67
Compression-Tokens (Causal)Backbone=Qwen3-0.6B, Compression Ratio=4x, Aggregation Mode=Multi2025.1043.62
Compression-Tokens (Causal)Backbone=Qwen3-1.7B, Compression Ratio=4x, Aggregation Mode=Single2025.1043.01
Mean-PoolingBackbone=Llama3.2-1B, Compression Ratio=16x, Aggregation Mode=Multi2025.1042.76
Compression-Tokens (Bidirectional)Backbone=Qwen3-1.7B, Compression Ratio=16x, Aggregation Mode=Single2025.1042.66
Compression-Tokens (Bidirectional)Backbone=Llama3.2-1B, Compression Ratio=16x, Aggregation Mode=Multi2025.1041.91
PCC LargeBackbone=Llama3.1-8B, Compression Ratio=16x, Aggregation Mode=Single2025.1041.79
Compression-Tokens (Causal)Backbone=Qwen3-1.7B, Compression Ratio=16x, Aggregation Mode=Single2025.1041.31
Compression-Tokens (Causal)Backbone=Qwen3-1.7B, Compression Ratio=16x, Aggregation Mode=Multi2025.1040.45
Compression-Tokens (Bidirectional)Backbone=Llama3.2-1B, Compression Ratio=16x, Aggregation Mode=Single2025.1040.44
Mean-PoolingBackbone=Qwen3-8B, Compression Ratio=128x, Aggregation Mode=Single2025.1039.95
Compression-Tokens (Bidirectional)Backbone=Qwen3-0.6B, Compression Ratio=16x, Aggregation Mode=Multi2025.1039.5
Compression-Tokens (Bidirectional)Backbone=Qwen3-8B, Compression Ratio=128x, Aggregation Mode=Multi2025.1039.36
Compression-Tokens (Causal)Backbone=Llama3.2-1B, Compression Ratio=16x, Aggregation Mode=Single2025.1039.16
Compression-Tokens (Causal)Backbone=Qwen3-8B, Compression Ratio=128x, Aggregation Mode=Single2025.1039.09
Mean-PoolingBackbone=Llama3.2-1B, Compression Ratio=16x, Aggregation Mode=Single2025.1038.84
Compression-Tokens (Bidirectional)Backbone=Qwen3-8B, Compression Ratio=128x, Aggregation Mode=Single2025.1038.72
Mean-PoolingBackbone=Qwen3-0.6B, Compression Ratio=16x, Aggregation Mode=Multi2025.1038.68
Compression-Tokens (Causal)Backbone=Llama3.2-1B, Compression Ratio=16x, Aggregation Mode=Multi2025.1038.51
Mean-PoolingBackbone=Qwen3-0.6B, Compression Ratio=16x, Aggregation Mode=Single2025.1038.3
Mean-PoolingBackbone=Qwen3-8B, Compression Ratio=128x, Aggregation Mode=Multi2025.1037.8
Compression-Tokens (Causal)Backbone=Qwen3-8B, Compression Ratio=128x, Aggregation Mode=Multi2025.1037.05
Mean-PoolingBackbone=Gemma2-2B, Compression Ratio=128x, Aggregation Mode=Single2025.1036.61
Compression-Tokens (Causal)Backbone=Gemma2-2B, Compression Ratio=128x, Aggregation Mode=Single2025.1036.14
Compression-Tokens (Bidirectional)Backbone=Gemma2-2B, Compression Ratio=128x, Aggregation Mode=Single2025.1036.14
Compression-Tokens (Bidirectional)Backbone=Qwen3-0.6B, Compression Ratio=16x, Aggregation Mode=Single2025.1036.11
Mean-PoolingBackbone=Qwen3-4B, Compression Ratio=128x, Aggregation Mode=Single2025.1035.49
Mean-PoolingBackbone=Gemma2-2B, Compression Ratio=128x, Aggregation Mode=Multi2025.1034.93
Compression-Tokens (Causal)Backbone=Qwen3-4B, Compression Ratio=128x, Aggregation Mode=Single2025.1034.83
Compression-Tokens (Bidirectional)Backbone=Gemma2-2B, Compression Ratio=128x, Aggregation Mode=Multi2025.1034.75
Compression-Tokens (Bidirectional)Backbone=Qwen3-4B, Compression Ratio=128x, Aggregation Mode=Multi2025.1034.33
Compression-Tokens (Causal)Backbone=Qwen3-0.6B, Compression Ratio=16x, Aggregation Mode=Multi2025.1034.3
Compression-Tokens (Causal)Backbone=Gemma2-2B, Compression Ratio=128x, Aggregation Mode=Multi2025.1033.94
LLMLingua2Backbone=Qwen3-8B, Compression Ratio=4x, Aggregation Mode=Multi2025.1033.63
Compression-Tokens (Causal)Backbone=Qwen3-4B, Compression Ratio=128x, Aggregation Mode=Multi2025.1033.48