MMLU: High School Mathematics

MMLU subject subset: Algebra, geometry and trigonometry at high-school level.

unassessed

This page is a discovery lead. Nobody has yet assessed it against the catalogue contract, so it carries no disposition. Absence of evidence here is not evidence of staleness.
Categoryknowledge
Subcategorymath
Page statusactive
Metricaccuracy
Directionhigher_is_better
Unit%
Dataset size270
Dataset licenceMIT
PublisherUC Berkeley (original); Center for AI Safety (current host)

What it measures

Algebra, geometry and trigonometry at high-school level. Questions are four-option multiple-choice, drawn from the MMLU test set's "math" subcategory within the benchmark's "STEM" top-level group, and are graded on the single correct labelled option.

Task format

Four-option multiple-choice questions, graded on the single correct labelled option; commonly evaluated 5-shot, consistent with the rest of MMLU.

Models reporting this benchmark

These figures come from the model cards, which carry one collection date per card and no per-score attribution. They are shown as reported, not as verified evidence.
ModelProviderScoreCard as of
Yi 1.5 34B Chat01.AI55.62026-04
Yi 1.5 34B Chat 16K01.AI55.62024-07
Meta Llama 3 70BMeta54.42024-07
Meta Llama 3 70B InstructMeta54.42026-04
Meta Llama 3 70B InstructNous Research54.42026-04
Yi 1.5 34B 32K01.AI51.92024-07
Yi 1.5 34B01.AI51.12026-04
Mixtral 8x22B Instruct v0.1Mistral AI50.72026-04
Yi 1.5 9B Chat01.AI47.02024-07
Yi 1.5 9B Chat 16K01.AI45.62024-07
Yi 1.5 9B01.AI43.72024-07
gemma 7B itGoogle DeepMind43.02024-07
Yi 1.5 9B 32K01.AI43.02024-07
Yi 1.5 6B Chat01.AI42.62024-07
Nous Hermes 2 Yi 34BNous Research41.52024-07
Yi 1.5 6B01.AI40.72024-07
Meta Llama 3 8BMeta40.02024-07
Meta Llama 3 8BNous Research40.02024-07
Meta Llama 3 8B InstructMeta40.02024-07
Meta Llama 3 8B InstructNous Research40.02024-07
Phi 3 mini 128K instructMicrosoft40.02024-07
Yi 9B01.AI40.02024-07
Phi 3 mini 4K instructMicrosoft39.62024-07
Mixtral 8x7B Instruct v0.1Mistral AI38.92026-04
Hermes 2 Pro Llama 3 8BNous Research38.52024-07
Mixtral 8x7B v0.1Mistral AI38.52026-04
Hermes 2 Theta Llama 3 8BNous Research38.12024-07
Yi 34B 200K01.AI38.12024-07
Nous Hermes 2 SOLAR 10.7BNous Research36.32024-07
Mistral 7B v0.3Mistral AI35.92024-07
mistral 7B v0.3 bnb 4bitUnsloth35.92024-07
Qwen2 1.5B InstructAlibaba / Qwen Team35.92024-07
Yi 34B Chat01.AI35.62024-07
Nous Hermes 2 Mixtral 8x7B DPONous Research34.82024-07
Qwen2 0.5B InstructAlibaba / Qwen Team33.72024-07
phi 2Microsoft33.02024-07
Yi 6B01.AI32.62024-07
Yi 6B Chat01.AI32.62024-07
falcon 40BTII31.52024-07
deepseek llm 7B baseDeepSeek30.02024-07
deepseek llm 7B chatDeepSeek30.02024-07
Mistral 7B Instruct v0.2Mistral AI30.02024-07
deepseek coder 6.7B baseDeepSeek28.92024-07
deepseek coder 6.7B instructDeepSeek27.82024-07
chatglm2 6BZhipu AI26.32024-07
gemma 2BGoogle DeepMind25.92024-07
OLMo 1B hfAllen AI25.92024-07
deepseek coder 1.3B baseDeepSeek24.42024-07
deepseek coder 1.3B instructDeepSeek24.42024-07
gemma 2B itGoogle DeepMind20.02024-07

Data

This page as JSON · Edit on GitHub