10/ Speed up inference with Prompt Prefix Caching
PROMPT ENGINEERING
-
8 Prompts to Level Up with ChatGPT-4o
By
–
I’ve learned so much using ChatGPT-4o, and I didn’t need to pay for courses. Here are 8 prompts you can use to level up fast!
-
Predibase and Comet Partner for High-Performance GenAI Applications
By
–
Predibase + @Cometml
: everything you need to build high performance #GenAI applications We're excited to be a launch partner for Comet's #Opik! Together we make it easy to customize best-of-breed small models (#SLMs), seamlessly track and #evaluate your experiments, and serve -
Co-LLM Collaboration Improves Mathematical Problem Solving Accuracy
By
–
The base LLM tried to solve math problems like “a^3 · a^2 if a=5,” where it incorrectly calculated the answer to be 125. Co-LLM trained the model to collaborate w/the large math LLM Llemma, and together they determined that the correct solution was 3,125.
-
Co-LLM: Collaborative Expert Switch for Enhanced Token Generation
By
–
If you asked Co-LLM to name some examples of extinct bear species, two models would draft answers together. The general-purpose LLM begins to put together a reply, w/the switch variable intervening where it can slot in a better token from the expert mode (i.e. adding the year
-
Custom Retrievers Format Context Metadata for LLMs
By
–
3/5
— FlowiseAI (@FlowiseAI) 17 septembre 2024
Custom Retriever let you specify the format of retrieved context.
Sometimes you might want to tie the context to relevant metadata:
{{ context }}
Souce: {{ metadata.source }}
Page: {{ metadata.pageNo }}
This way you can prompt the LLM to return in specific format. pic.twitter.com/Im7K8F6Hz83/5 Custom Retriever let you specify the format of retrieved context. Sometimes you might want to tie the context to relevant metadata:
{{ context }}
Souce: {{ metadata.source }}
Page: {{ metadata.pageNo }} This way you can prompt the LLM to return in specific format. -
LLMs struggle with word length and final letters
By
–
It’s not always right of course — the other three answers I saw in five samples were “facetiously” (twice) and “abstentious.” But 2/5 here is pretty good — LLMs are famously bad at identifying the length or final letters of words
-
Prompting ChatGPT-4o with chain-of-thought reduces correctness
By
–
For comparison, when I give the same prompt to ChatGPT 4o, but replacing the last sentence with “Think step-by-step,” it returns 0 correct answers in 10 samples: “salubrious,” “abstemious” (x4), “illegible,” “repetitive,” “luminously”, “abstinential,” and “inexpensive.”
-
LLM Token Capacity vs Instruction Comprehension Challenge
By
–
Trillion tokens sounds cool but it would be even more cool if the LLM can understand/retrieve/reason about all of them, and not miss any important instruction from the prompt.