Benjamin Han · @BenjaminHan
334 followers · 720 posts · Server sigmoid.social
3/
2. To test how much LLMs remember raw content, the authors prompted the models (#GPT3, #BLOOM, #AnthropicLM, #Cohere etc) with samples of original texts/titles, and found memorization is more frequent with popular material/larger LLMs (screenshot 2).
#gpt3 #bloom #anthropiclm #cohere #generativeAI #law #ethics #fairuse #copyright #paper
Last updated 2 years ago