Notas do episódio
This research paper evaluates the performance of DeepSeek, a new large language model (LLM), against other popular models like Claude, Gemini, GPT, and Llama. The comparison focuses on two classification tasks: determining the authorship of text (human or AI-generated) and classifying academic citations based on their function. The study introduces new datasets, MadStatAI and CitaStat, for benchmarking LLMs in these tasks. Results indicate that DeepSeek is competitive, outperforming some models but generally falling short of Claude in accuracy. However, DeepSeek offers a balance of…
Site: podcasters.spotify.com

