Sources
Loading...
Additional media
Loading...
DeepNewz, mobile.
People-sourced. AI-powered. Unbiased News.
OpenAI's o1 model, which integrates reasoning directly into its architecture, has been a significant topic of discussion in the AI community. The model avoids Monte Carlo Tree Search (MCTS) and focuses on embedding reasoning within the model itself. Noam Brown has been a key figure in this development. However, there have been reports of AnthropicAI's Claude 3.5 Sonnet outperforming OpenAI's o1 in reasoning tasks when using specific prompting techniques. These techniques involve using Chain-of-Thought (CoT) prompting, where thoughts are enclosed within tags and solutions are broken down into clear steps. Dynamic Chain of Thoughts and verbal reinforcement are also part of these techniques. Despite these advancements, some experts believe that with better prompting, OpenAI's o1 could still outperform Claude 3.5 Sonnet.
"claude 3.5 sonnet to outperform openai o1 in terms of reasoning" with prompting 🤔 ---- Prompt from the article: Begin by enclosing all thoughts within <thinking> tags, exploring multiple angles and approaches. Break down the solution into clear steps within <step> tags.… https://t.co/nX23azzPWh
1/n Enhancing Spatial Reasoning in Large Language Models with Symbolic Prompting Large Language Models (LLMs) have demonstrated remarkable capabilities in various domains, including complex reasoning tasks. Chain-of-Thought (CoT) prompting has emerged as a powerful technique for… https://t.co/XMLdIMESsJ
Reading a lot about making Claude 3.5 Sonnet beat o1 in reasoning by tricking it to follow chain of thought using prompts However @ikristoph tested it and finds that though it does improve Claude output but still bit below o1. This means OpenAI is still ahead of Anthropic. Am… https://t.co/S6VOBdyWjY https://t.co/i3M5zucSLA
