New jailbreaking technique: pure repetition. AIs are getting big context windows, it turns out if you fill a lot of it with examples of bad behavior, the AI becomes much more willing to breach its own guardrails. Security people are used to rules-based systems. This is weirder.
PROMPT ENGINEERING
-
Building Powerful RAG Systems with Large Language Models
By
–
🚀 Blog-tastic Tuesdays 🚀https://t.co/qhmO6lROB2
— Satya Mallick (@LearnOpenCV) 2 avril 2024
Our latest blog explores the exciting realm of RAG systems.
By the end of this article, you’ll be equipped to build a powerful and dynamic LLM solution that leverages the strengths of both pre-trained models and up-to-date… pic.twitter.com/Zs6J4ZIazLBlog-tastic Tuesdays https://
learnopencv.com/rag-with-llms/ Our latest blog explores the exciting realm of RAG systems.
By the end of this article, you’ll be equipped to build a powerful and dynamic LLM solution that leverages the strengths of both pre-trained models and up-to-date -
Asking Opus to Double-Check Its Own Answers
By
–
have you tried simply asking opus to doublecheck its own answers
-
Two Advanced Webinars on LLM Optimization and Flow Engineering
By
–
We've got two AMAZING webinars coming up on two advanced topics: "Optimization of LLM Systems" with Omar K. (author of DSPy) https://
us06web.zoom.us/webinar/regist
er/WN_n4HUNqhFQFa4ax3kYrZW8g#/registration
… "Flow Engineering" with Itamar Friedman (author of Alpha Codium) https://
us06web.zoom.us/webinar/regist
er/WN_fVikSl9eQv68b3ZUdQgwzA#/registration
… -
Claude Model Optimization Through In-Context Examples
By
–
Have you considered adding a few examples to the Claude code, maybe even having "Claude with examples" as a separate entry on the board? My expertise with Opus and Haiku and Sonnet is that they are extremely sensitive to examples, it helps them out a lot
-
GPT-4 Provides More Opinionated Directions to Haiku
By
–
Here an example of GPT-4 directing Haiku, GPT-4 is much more opinionated with directions.
-
Multi-Model AI Architecture: GPT-4 and Claude Haiku Workflow
By
–
The reason I am saying this is that I have been experimenting with this flow: GPT-4 directs and refines instances of Claude Haiku to accomplish a goal. Opus refines and provide the entire output end to end. It's super powerful.
-
Claude Function Prompts Using Gorilla Code Implementation
By
–
That's using this code, right? https://
github.com/ShishirPatil/g
orilla/blob/6971033188838d975bfdf78f9decb244851586fa/berkeley-function-call-leaderboard/model_handler/claude_handler.py
… Using this code to construct the Claude function prompts? -
Debug with GPT-4, Code with Claude 3 Strategy
By
–
The biggest alpha in coding with AI right now is: Debug with GPT-4, Code with Claude 3. GPT-4 is still king when it comes to logic, but it's extremely lazy.
Meanwhile, Claude would do anything you ask. The duo together is unbeatable. -
Tried a ‘Roast My Website’ Custom GPT — it burns
By
–
Just added the instructions of "Roast My Website" GPT to my Custom GPTs Toolkit. And tried it as a test again and damn, it burns. Haha. It's fun (not).