Read on Twitter
With all eyes on #WWDC today and the incredible unveiling of Apple's Vision Pro glasses, another small release may have slipped under your radar!
@OpenAI just published a comprehensive guide on GPT best practices in their documentation
Here are some helpful takeaways from it 👇
@OpenAI just published a comprehensive guide on GPT best practices in their documentation
Here are some helpful takeaways from it 👇
1. Task breakdown
For better performance and higher accuracy, it is very important to break complex tasks into simpler ones.
An interesting tactic is to use intent classification to identify the most relevant instructions for a user's query:
For better performance and higher accuracy, it is very important to break complex tasks into simpler ones.
An interesting tactic is to use intent classification to identify the most relevant instructions for a user's query:
2. Giving GPT time to "think"
Requesting immediate answers from the model might yield inaccuracies as it doesn't allow for 'reasoning' time. Remember how LLMs operate is heavily reliant on thoughtful processing.
A tactic is asking the model if it missed anything previously:
Requesting immediate answers from the model might yield inaccuracies as it doesn't allow for 'reasoning' time. Remember how LLMs operate is heavily reliant on thoughtful processing.
A tactic is asking the model if it missed anything previously:
3. Offload specific workload to tools
Whether it's a text retrieval system enhancing document relevance or @Wolfram_Alpha assisting in math & science - strategically offload tasks to maximize efficiency
@LangChainAI and @llama_index provides multiple ways for you to do this.
Whether it's a text retrieval system enhancing document relevance or @Wolfram_Alpha assisting in math & science - strategically offload tasks to maximize efficiency
@LangChainAI and @llama_index provides multiple ways for you to do this.
4. Evaluating changes
OpenAI highlights how you can evaluate model outputs w/ reference to gold answer.
It's helpful in certain cases, but not without its nuances. Dive deeper into our evaluation post to unlock the intricacies:
OpenAI highlights how you can evaluate model outputs w/ reference to gold answer.
It's helpful in certain cases, but not without its nuances. Dive deeper into our evaluation post to unlock the intricacies:
https://twitter.com/agiguide_/status/1664769496282968064
You can find @OpenAI's new guide here:
platform.openai.com/docs/guides/gp…
And, if you value practical production-ready LLM tips, like and retweet! It helps us help you 🚀
platform.openai.com/docs/guides/gp…
And, if you value practical production-ready LLM tips, like and retweet! It helps us help you 🚀
• • •
Missing some Tweet in this thread? You can try to
force a refresh
