Rohan Paul@rohanpaul_ai

2 years ago

Smaller LLMs struggle with complex mathematical reasoning due to inability to detect and fix errors. Teacher-student framework enhances mathematical reasoning in smaller language models. With Hierarchical thought templates and cross-model DPO **Solution in this Paper** 🧠: •… https://t.co/jVUfNDFaZs https://t.co/3WFIRMzi4N

Ethan Mollick@emollick

2 years ago

I wrote about how understanding some of how LLMs work can help you get an intuition for how they "think" & to get better results out of them (though even knowing the underlying principles of how they work doesn't easily explain everything that AIs can do) https://t.co/1k5b9DWWhG

Can Large Language Models (LLMs) truly reason? https://t.co/JruUQ0fnt6 TLDR: no