You can deploy this model and start earning money today!

Korea-MES/
Mixtral-Upperbound

Views

No views yet

	Benchmark-Focused MLT Dataset




	
	Overview

Total Samples: 682,707 Train Samples: 680,662 Test Samples: 2,000 (MLT당 200개씩) MLT Labels: 10

Target Benchmarks

Math: GSM8K, MATH Reasoning: BBH, ARC-Challenge Knowledge: MMLU, MMLU-Pro Commonsense: HellaSwag, Winogrande, PIQA Truthfulness: TruthfulQA

	Category Distribution

Category Count Percentage

Instruction 278,647 40.8%

Math 204,485 30.0%

Knowledge 99,792 14.6%

Commonsense 96,413 14.1%… See the full description on the dataset page: https://huggingface.co/datasets/Korea-MES/Mixtral-Upperbound.