memes.cjtrowbridge.com

llama-3.2-vision

The meme is a screenshot of a table comparing the performance of various AI models on the benchmark dataset. The table has two columns: "Pre-Trained" and "Instructed", and lists several AI models, including Meta Llama 3, MMLU, and others. The table shows that Meta Llama 3 400B+ outperforms other models in both pre-trained and instructed settings, with the highest score in both categories.

first-seen

2024-05-04T09:02:37+00:00

tesseract

Meta Llama 3 400B+ (still training) Checkpoint as of Apr 15, 2024 PRE-TRAINED INSTRUCT Meta Llama 3 Meta Llama 3 400B+ 400B+ MMLU 84.8 MMLU 86.1 5-shot . 5-shot . AGIEval English GPQA 3-5-shot 69.9 O-shot 48.0 BIG-Bench Hard HumanEval 3-shot, CoT 85.3 O-shot 84.1 ARC-Challenge GSM-8K 25-shot 96.0 8-shot, CoT 94.1 DROP MATH 3-shot, F1 83.5 4-shot, CoT 57.8