marcelbinz/Llama-3.1-Centaur-70B
Llama-3.1-Centaur-70B by Marcel Binz is a 70 billion parameter foundation model designed to predict and simulate human behavior in behavioral experiments described in natural language. This model is built upon the Llama 3.1 architecture and specializes in cognitive modeling. It requires specific prompt formatting using "<<" and ">>" tokens for optimal performance and is intended for research in human cognition.
1.0 based on 1 review
Hugging Face It is incredibly difficult to work with. It accepts no more than 5 sentences, otherwise it starts to stupidly copy the prompt. The answer is also limited to 5 sentences, then endless slashes. It only works in English. It is impossible to load the context. The quality of the answers is worse than that of Gemini 2.5. I tested only the behavior model of 1 individual based on his psychological profile.