PixelVerse t1 – CoT prompting outperforms flagship LLMs ai.pixelverse.tech 9 points by hayden_k a day ago
hayden_k a day ago OpenAI o1-like CoT and logical thinking prompting strategy significantly enhances llm responses.PixelVerse t1 is powered by Llama 3.1 70b and 3.2 90b. However, with a detailed CoT prompt, it answers complex questions correctly, much better than it's base model and even sometimes beating flagship models like GPT 4o, Claude 3 and Gemini.Try it at: https://ai.pixelverse.tech/app/cortexchat
red2awn 16 hours ago prompt: how many "r"s in the word "raspberry"?response: There are 2 "r"s in the word "raspberry".
OpenAI o1-like CoT and logical thinking prompting strategy significantly enhances llm responses.
PixelVerse t1 is powered by Llama 3.1 70b and 3.2 90b. However, with a detailed CoT prompt, it answers complex questions correctly, much better than it's base model and even sometimes beating flagship models like GPT 4o, Claude 3 and Gemini.
Try it at: https://ai.pixelverse.tech/app/cortexchat
First try with llama 70b found two R's in strawberry :) Gemma did better
I asked the 9.8 vs. 9.11 question from the examples and it got it wrong :/
prompt: how many "r"s in the word "raspberry"?
response: There are 2 "r"s in the word "raspberry".
Can we see the detailed CoT prompt?
I guess there really are two r’s in strawbery.
lol, this feels like Reflection 70b all over again.