benchmarkPerformance

8 triples
GPTKB property

Random triples
Subject Object
gptkb:Llama_3 competitive with GPT-3.5
gptkb:Llama_2_13B_Chat competitive with GPT-3.5 on some tasks
gptkb:Llama_3 approaches GPT-4 (in some tasks)
gptkb:pre-trained_wav2vec2_models state-of-the-art on LibriSpeech
gptkb:Falcon_LLM high on Open LLM Leaderboard
gptkb:ViT-L state-of-the-art on ImageNet (2020)
gptkb:Llama_3 outperforms Llama 2
gptkb:ConvNeXt-Large Top-1 accuracy 83.5% on ImageNet-1K