Model

LLaMA 2 7B

Meta · open weights

ai.meta.com

Knowledge & Reasoning

Knowledge & Reasoning
BenchmarkScoreMetricHarnessSourceConfidenceLink
BBH39.2averageN/AIndependentmediumarxiv.org
BBH32.6averageN/ASelf-reportedmediumarxiv.org
BoolQ82.6accuracy_(%_)N/AIndependentmediumarxiv.org
BoolQ82.6accuracy_(%_)N/AIndependentmediumarxiv.org
BoolQ81.9accuracy_(%_)N/AIndependentmediumarxiv.org
BoolQ80.3accuracy_(%_)N/AIndependentmediumarxiv.org
BoolQ77.9scoreN/AIndependentmediumarxiv.org
BoolQ77.4accuracyN/AIndependentmediumarxiv.org
BoolQ77.4scoreN/ASelf-reportedmediumarxiv.org
EconLogicQA9.2accuracyN/AIndependentmediumarxiv.org
HellaSwag77.2accuracyN/AIndependentmediumarxiv.org
HellaSwag77.2overall_accuracyN/AIndependentmediumarxiv.org
HellaSwag76.7accuracy_(%_)N/AIndependentmediumarxiv.org
HellaSwag76.7accuracy_(%_)N/AIndependentmediumarxiv.org
HellaSwag76.3accuracy_(%_)N/AIndependentmediumarxiv.org
LAMBADA73.3scoreN/AIndependentmediumarxiv.org
MMLU45.8exact_matchN/AIndependentmediumcrfm.stanford.edu
MMLU45.3average_(%)N/AIndependentmediumarxiv.org
MMLU45.3exact_matchN/ASelf-reportedmediumarxiv.org
OpenBookQA58.6accuracyN/ASelf-reportedmediumarxiv.org
OpenBookQA31.4accuracyN/AIndependentmediumarxiv.org
PIQA78.8accuracyN/AIndependentmediumarxiv.org
PIQA78.8scoreN/ASelf-reportedmediumarxiv.org
PIQA78.8scoreN/AIndependentmediumarxiv.org
PubChemQA0.1bleu-2N/AIndependentmediumarxiv.org
ScienceQA43.1scoreN/AIndependentmediumscienceqa.github.io
Trivia QA73.7exact_matchN/ASelf-reportedmediumarxiv.org
Trivia QA72.1exact_matchN/AIndependentmediumarxiv.org
UniProtQA0.0bleu-2N/AIndependentmediumarxiv.org
WinoGrande70.8accuracy_(%_)N/AIndependentmediumarxiv.org
WinoGrande70.1accuracy_(%_)N/AIndependentmediumarxiv.org
WinoGrande69.8accuracy_(%_)N/AIndependentmediumarxiv.org
WinoGrande69.2accuracyN/ASelf-reportedmediumarxiv.org
WinoGrande69.2accuracyN/AIndependentmediumarxiv.org

Math

Math
BenchmarkScoreMetricHarnessSourceConfidenceLink
GSM8K26.3exact_matchN/AIndependentmediumarxiv.org
GSM8K14.6exact_matchN/ASelf-reportedmediumarxiv.org

Coding

Coding
BenchmarkScoreMetricHarnessSourceConfidenceLink
MBPP20.8accuracyN/AIndependentmediumarxiv.org

Agentic / Tool Use

Agentic / Tool Use
BenchmarkScoreMetricHarnessSourceConfidenceLink
WebLINX24.6overall_scoreN/AIndependentmediumarxiv.org

Safety & Alignment

Safety & Alignment
BenchmarkScoreMetricHarnessSourceConfidenceLink
SynthPAI33.0average_accuracy_in_%N/AIndependentmediumarxiv.org

Reproduced by the community

Loading community runs…