MMLU: Conceptual Physics

MMLU subject subset: Qualitative, non-calculus physics concepts, as opposed to the calculation-heavy College Physics subject.

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
Subcategoryphysics
Page statusactive
Metricaccuracy
Directionhigher_is_better
Unit%
Dataset size235
Dataset licenceMIT
PublisherUC Berkeley (original); Center for AI Safety (current host)

What it measures

Qualitative, non-calculus physics concepts, as opposed to the calculation-heavy College Physics subject. Questions are four-option multiple-choice, drawn from the MMLU test set's "physics" 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 32K01.AI84.72024-07
Yi 1.5 34B01.AI83.82026-04
Meta Llama 3 70BMeta81.72024-07
Meta Llama 3 70B InstructMeta81.72026-04
Meta Llama 3 70B InstructNous Research81.72026-04
Yi 1.5 34B Chat01.AI81.72026-04
Yi 1.5 34B Chat 16K01.AI81.72024-07
Mixtral 8x22B Instruct v0.1Mistral AI79.12026-04
Nous Hermes 2 Yi 34BNous Research79.12024-07
Yi 34B 200K01.AI77.42024-07
Yi 34B Chat01.AI76.62024-07
Yi 1.5 9B01.AI73.62024-07
Yi 1.5 9B Chat01.AI72.32024-07
Yi 1.5 9B 32K01.AI71.92024-07
Yi 9B01.AI71.92024-07
Yi 1.5 9B Chat 16K01.AI71.12024-07
Phi 3 mini 4K instructMicrosoft70.62024-07
Phi 3 mini 128K instructMicrosoft69.82024-07
Nous Hermes 2 Mixtral 8x7B DPONous Research69.42024-07
Mixtral 8x7B v0.1Mistral AI68.12026-04
Mixtral 8x7B Instruct v0.1Mistral AI66.82026-04
Yi 1.5 6B01.AI66.82024-07
Yi 1.5 6B Chat01.AI64.72024-07
gemma 7B itGoogle DeepMind62.12024-07
Meta Llama 3 8B InstructMeta60.42024-07
Meta Llama 3 8B InstructNous Research60.42024-07
Yi 6B01.AI60.42024-07
Yi 6B Chat01.AI60.42024-07
Nous Hermes 2 SOLAR 10.7BNous Research59.62024-07
Meta Llama 3 8BMeta59.12024-07
Meta Llama 3 8BNous Research59.12024-07
Mistral 7B v0.3Mistral AI58.32024-07
mistral 7B v0.3 bnb 4bitUnsloth58.32024-07
Hermes 2 Pro Llama 3 8BNous Research56.62024-07
Hermes 2 Theta Llama 3 8BNous Research55.72024-07
Mistral 7B Instruct v0.2Mistral AI53.22024-07
phi 2Microsoft52.32024-07
Qwen2 1.5B InstructAlibaba / Qwen Team50.22024-07
deepseek llm 7B baseDeepSeek43.42024-07
deepseek llm 7B chatDeepSeek43.42024-07
gemma 2BGoogle DeepMind41.72024-07
falcon 40BTII41.32024-07
chatglm2 6BZhipu AI40.02024-07
Qwen2 0.5B InstructAlibaba / Qwen Team37.02024-07
deepseek coder 6.7B baseDeepSeek35.32024-07
gemma 2B itGoogle DeepMind35.32024-07
deepseek coder 6.7B instructDeepSeek34.92024-07
deepseek coder 1.3B baseDeepSeek31.12024-07
deepseek coder 1.3B instructDeepSeek31.12024-07
OLMo 1B hfAllen AI23.02024-07

Data

This page as JSON · Edit on GitHub