List your company

Multiple Choice Question Answering on FrenchBench

82.2AccuracyTokens to Words
44.65654.40364.1573.897May 26, 2025

Evaluation results

MethodLinks
Tokens to WordsBackbone=Qwen3-8B2025.0582.2
Original tokenizationBackbone=Qwen3-8B2025.0581.7
Original tokenizationModel=Qwen3-8B2025.0581.7
Token DistillationBackbone=Qwen3-8B2025.0581.5
Token DistillationModel=Qwen3-8B2025.0581.5
NTPBackbone=Qwen3-8B2025.0581.2
NTPModel=Qwen3-8B2025.0581.2
Original tokenizationBackbone=Avg.2025.0573.2
Original tokenizationModel=Avg.2025.0573.2
Token DistillationBackbone=Llama3-8B-i2025.0572.9
Token DistillationBackbone=Avg.2025.0572.9
Token DistillationModel=Llama3-8B-i2025.0572.9
Token DistillationModel=Avg.2025.0572.9
Original tokenizationBackbone=Llama3-8B-i2025.0572.1
Original tokenizationModel=Llama3-8B-i2025.0572.1
NTPBackbone=Avg.2025.0570.8
NTPModel=Avg.2025.0570.8
NTPBackbone=Llama3-8B-i2025.0570.1
NTPModel=Llama3-8B-i2025.0570.1
MeanBackbone=Qwen3-8B2025.0569.6
MeanModel=Qwen3-8B2025.0569.6
Original tokenizationBackbone=Mistral-7B2025.0569.5
Original tokenizationModel=Mistral-7B2025.0569.5
Original tokenizationBackbone=Llama3-8B2025.0569.4
Original tokenizationModel=Llama3-8B2025.0569.4
Tokens to WordsBackbone=Avg.2025.0569.1
Token DistillationBackbone=Llama3-8B2025.0568.9
Token DistillationModel=Llama3-8B2025.0568.9
Token DistillationBackbone=Mistral-7B2025.0568.5
Token DistillationModel=Mistral-7B2025.0568.5
NTPBackbone=Llama3-8B2025.0567
NTPModel=Llama3-8B2025.0567
Tokens to WordsBackbone=Llama3-8B-i2025.0566.6
ZeTTBackbone=Mistral-7B2025.0566.1
ZeTTModel=Mistral-7B2025.0566.1
ZeTTBackbone=Llama3-8B-i2025.0565.2
ZeTTModel=Llama3-8B-i2025.0565.2
Tokens to WordsBackbone=Mistral-7B2025.0564.8
NTPBackbone=Mistral-7B2025.0564.7
NTPModel=Mistral-7B2025.0564.7
Tokens to WordsBackbone=Llama3-8B2025.0562.6
RandomBackbone=Qwen3-8B2025.0562.1
RandomModel=Qwen3-8B2025.0562.1
MeanBackbone=Llama3-8B-i2025.0561.7
MeanModel=Llama3-8B-i2025.0561.7
MeanBackbone=Avg.2025.0561.5
MeanModel=Avg.2025.0561.5
ZeTTBackbone=Llama3-8B2025.0561.3
ZeTTModel=Llama3-8B2025.0561.3
MeanBackbone=Llama3-8B2025.0558.4
MeanModel=Llama3-8B2025.0558.4
MeanBackbone=Mistral-7B2025.0556.3
MeanModel=Mistral-7B2025.0556.3
RandomBackbone=Avg.2025.0552.7
RandomModel=Avg.2025.0552.7
RandomBackbone=Llama3-8B-i2025.0551.5
RandomModel=Llama3-8B-i2025.0551.5
RandomBackbone=Mistral-7B2025.0551
RandomModel=Mistral-7B2025.0551
RandomBackbone=Llama3-8B2025.0546.1
RandomModel=Llama3-8B2025.0546.1