List your company

General Language Intelligence on MMLU, GSM8K, BBH, TriviaQA, NQ (test)

83.54MMLUGAC (Qwen2-72B + Llama-3)
63.083268.394173.70579.0159Jun 18, 2024

Evaluation results

MethodLinks
GAC (Qwen2-72B + Llama-3)Date=~2024/06/07, Latency=151.56ms/token2024.0683.5490.9163.9979.2937.6571.084.06
GAC (Task-specific Top-2 Ensemble)Date=~2024/06/072024.0683.5490.9163.9980.8437.9571.454.61
Qwen2-72B-InstructDate=2024/06/07, Latency=113.91ms/token2024.0682.389.762.5773.5833.1168.25—
GAC (Llama-3 + Qwen1.5-72B)Date=~2024/04/18, Latency=153.96ms/token2024.0681.4987.0656.7378.636.0167.98-0.47
GAC (Task-specific Top-2 Ensemble)Date=~2024/04/182024.0681.4987.9658.6480.8437.9569.381.58
GAC (Qwen1.5-72B + Yi)Date=~2024/02/04, Latency=103.69ms/token2024.0679.8377.2752.0570.8833.862.773.65
Llama-3-70B-InstructDate=2024/04/18, Latency=150.32ms/token2024.0679.689057.1379.1235.5768.3—
GAC (Qwen1.5-72B + Mixtral)Date=~2024/02/04, Latency=112.83ms/token2024.0679.5575.7654.1975.7131.0963.264.47
Qwen1.5-72B-ChatDate=2024/02/04, Latency=102.11ms/token2024.0677.7983.3348.9465.6927.0260.55—
GAC (Yi + Mixtral)Date=~2023/12/11, Latency=98.13ms/token2024.0674.8371.2152.6475.633.5261.563.13
GACLatency=98.13ms/token, Ensemble Model Combination=Mixtral and Yi, Ensemble Frequency=every generation step2024.0674.8371.2152.6475.633.5261.56—
OAssistRMLatency=99.75ms/token, Ensemble Model Combination=Mixtral and Yi2024.0673.3470.1551.9172.7930.6959.78—
Yi-34B-ChatDate=2023/11/08, Latency=67.96ms/token2024.0672.7568.7650.8870.0129.8158.44—
Yi-34B-ChatLatency=67.96ms/token2024.0672.7568.7650.8870.0129.8158.44—
LLM Blender (PairRanker)Latency=105.21ms/token, Ensemble Model Combination=Mixtral and Yi2024.0672.6969.5951.772.3732.2459.72—
Mixtral-8x7B-Instruct-v0.1Date=2023/12/11, Latency=96.64ms/token2024.0670.8966.8249.8476.5434.3559.69—
Mixtral-8x7B-Instruct-v0.1Latency=96.64ms/token2024.0670.8966.8249.8476.5434.3559.69—
UltraRMLatency=114.57ms/token, Ensemble Model Combination=Mixtral and Yi2024.0669.4971.0952.2773.8232.3659.81—
GACLatency=51.32ms/token, Ensemble Model Combination=openchat and SOLAR, Ensemble Frequency=every generation step2024.0666.5174.351.1972.533.8259.66—
GACLatency=31.34ms/token, Ensemble Model Combination=openchat and SOLAR, Ensembled token proportion=7.68%2024.0665.1473.1850.3269.6831.7558.01—
UltraRMLatency=74.51ms/token, Ensemble Model Combination=openchat and SOLAR2024.0664.9475.5150.6571.2332.0358.87—
Nous-Hermes-2-SOLAR-10.7BLatency=50.89ms/token2024.0664.8872.8649.9271.3332.2158.24—
OAssistRMLatency=55.62ms/token, Ensemble Model Combination=openchat and SOLAR2024.0664.8772.9349.0270.0631.6457.7—
LLM Blender (PairRanker)Latency=57.86ms/token, Ensemble Model Combination=openchat and SOLAR2024.0664.2374.0750.1870.5532.5558.32—
FuseLLMLatency=28.31ms/token, Ensemble Model Combination=openchat and SOLAR2024.0663.9465.546.3264.5729.0653.88—
openchat_3.5Latency=28.01ms/token2024.0663.8768.4647.9668.1229.7555.63—