Issue in mind is not so much human bias but the fact that the full distribution of correct or desirable answers to your prompts is almost certainly not present in your dataset, only a few samples.
PROMPT ENGINEERING
-
Document Upload Transparency and AI Processing Visibility
By
–
I want to know exactly what happens to documents I upload to it – how they are split and chunked, how text is extracted from PDFs, what gets embedded and stored, and what gets retrieved and included in prompts (Ideally I'd like a debug interface that lets me see it for myself)
-
The Ambiguity Problem in LLM Labeling and Training Data
By
–
Consider being a labeler for an LLM. The prompt is “give me a random number between 1 and 10”. What SFT & RM labels do you contribute? What does this do the network when trained on? In subtle way this problem is present in every prompt that does not have a single unique answer.
-
OpenAI Chunking Strategy for RAG Implementation Details
By
–
I'm begging you @OpenAIDevs
, please tell us how your chunking works! It's a small detail, but it makes a huge difference in helping me make decisions about how to effectively use your RAG implement -
AI Agents Make Coding Accessible to Non-Technical Users
By
–
Agents are a big deal not just due to autonomy but also because they make sense to non-coders GPT-4 isn't quite there yet, but you get quite far with: "Hey Devin the AI agent make a more engaging version of my website… add links" The results were cute: http://
chic-tulumba-3fbd63.netlify.app -
Prompt Injection Threats Against GPT Wrapper Applications
By
–
GPT wrapper apps are exactly the things that need to worry about prompt injection – it's not an attack against the models, it's an attack against applications built on top of the models
-
Optimize Multiple LLM Calls by Mixing Models for Lower Latency
By
–
Most LLM apps and AI Agents need multiple calls to an LLM, especially a moderately complex LLM app/AI agent. Calling GPT-4 or Claude is impractical and you will soon be in high-latency hell. The optimal way to do this is to mix and match LLMs depending on the latency,
-
Claude 3 Opus prompt result on Poe impresses user
By
–
Just ran a prompt on Claude 3 Opus via Poe . com and this popped up. Love to see it.
-
Understanding LLM Rules and Patterns for Better Usage
By
–
absolutely right. I think the facade of communicating like humans makes it hard for people to understand this. But the LLM has its own rules that, if you learn, you can use to your advantage
-
Fine-tuning risks and diminishing returns in AI model optimization
By
–
Fine tuning can help if you have a specific use-case for the model that you can optimize for, but even then it's a whole lot of expensive and risky work for something that may not help much and will likely be obsoleted by the next model release