Skip to content

Benchmark

HELM MMLU: Logical Fallacies - EM

Scores are percentages. Higher is better.

Models measured
64
Scores recorded
68
Sources
1
IQ weight
2.10

Score distribution

81.3%
Mean
82.8%
Median
10%
Std dev
73.6%
10th percentile
90.2%
90th percentile
26.4%
Min
26.4%92.6%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.10 for chat models, measured over 64 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs92.6%
2Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs92%
3Claude 3.5 Sonnetsetting not named by the source· Stanford HELM specialized suites data docs91.4%
4DeepSeek V3· Stanford HELM specialized suites data docs91.4%
5Gemini 2.0 Flash Experimental· Stanford HELM specialized suites data docs91.4%
6Qwen2 72B Instruct· Stanford HELM specialized suites data docs91.4%
7gemini-1.5-pro-002· Stanford HELM specialized suites data docs90.2%
8GPT 4osetting not named by the source· Stanford HELM specialized suites data docs90.2%
9Claude 3 Opus· Stanford HELM specialized suites data docs89.6%
10gemini-1.5-pro-001· Stanford HELM specialized suites data docs89.6%
11Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs89%
12GPT 4osetting not named by the source· Stanford HELM specialized suites data docs88.3%
13Yi-34B· Stanford HELM specialized suites data docs88.3%
14mixtral-8x22b· Stanford HELM specialized suites data docs87.7%
15text-unicorn· Stanford HELM specialized suites data docs87.7%
16GPT 4· Stanford HELM specialized suites data docs87.1%
17GPT-4 Turbo· Stanford HELM specialized suites data docs87.1%
18GPT-4o mini· Stanford HELM specialized suites data docs87.1%
19Nova Prov1· Stanford HELM specialized suites data docs87.1%
20gemma2-27b· Stanford HELM specialized suites data docs86.5%
21GPT 4 1106 Preview· Stanford HELM specialized suites data docs86.5%
22Meta Llama 3 70b· Stanford HELM specialized suites data docs86.5%
23gemini-1.5-flash-002· Stanford HELM specialized suites data docs85.9%
24Jamba 1.5 Large· Stanford HELM specialized suites data docs85.9%
25Claude 3 Sonnet· Stanford HELM specialized suites data docs85.3%
26gemini-1.5-flash-001· Stanford HELM specialized suites data docs85.3%
27Qwen1.5-72B· Stanford HELM specialized suites data docs85.3%
28DBRX-instruct· Stanford HELM specialized suites data docs84.7%
29DeepSeek LLM 67B Chat· Stanford HELM specialized suites data docs84.7%
30Mistral Large2407· Stanford HELM specialized suites data docs84.7%
31Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs84%
32Claude 2.1· Stanford HELM specialized suites data docs83.4%
33Llama-3.2-90B-Vision-Instruct-Turbo· Stanford HELM specialized suites data docs83.4%
34Phi-3 Medium 4K Instruct· Stanford HELM specialized suites data docs82.8%
35Qwen1.5-110B-Chat· Stanford HELM specialized suites data docs82.8%
36Claude Haiku 3.5· Stanford HELM specialized suites data docs82.2%
37gemma-2-9b· Stanford HELM specialized suites data docs81.6%
38Llama 3.3 70B Instruct Turbo· Stanford HELM specialized suites data docs81.6%
39Claude Instant 1.2· Stanford HELM specialized suites data docs81%
40Jamba 1.5 Mini· Stanford HELM specialized suites data docs81%
41Mistral Largesetting not named by the source· Stanford HELM specialized suites data docs81%
42Nova Litev1· Stanford HELM specialized suites data docs81%
43Phi-3 Small 8K Instruct· Stanford HELM specialized suites data docs81%
44Mistral Small· Stanford HELM specialized suites data docs80.4%
45Command R+· Stanford HELM specialized suites data docs79.8%
46Nova Microv1· Stanford HELM specialized suites data docs79.8%
47Claude 3 Haiku· Stanford HELM specialized suites data docs79.1%
48Command R Plus· Stanford HELM specialized suites data docs79.1%
49GPT-3.5 Turbosetting not named by the source· Stanford HELM specialized suites data docs79.1%
50Llama 2 70b· Stanford HELM specialized suites data docs79.1%

Showing 1–50 of 68 results.

Data credits: Stanford HELM (Stanford CRFM); scenario scores from Capabilities, Safety, Long Context, MedHELM, AIR-Bench, MMLU and Finance