DetectGPT – an approach for zero-shot machine-generated text detection. Uses raw log probabilities from the LLM to determine if the passage was sampled from it. 8 of 11
LLMS
-
Watermarking Framework for Proprietary Language Models
By
–
A Watermark for LLMs – a watermarking framework for proprietary language models. 4 of 11
-
H3: Bridging Performance Gap Between State Space Models and Attention
By
–
H3 – an approach to reduce the gap, in terms of performance and hardware utilization, between state space models and attention for language modeling. 3 of 11
-
Generic Prompt Solves Mathematical Problem Correctly
By
–
With Generic Prompt: Input: You are a brilliant mathematician, try to solve the following problem:
What is 100*100/200*56?
Output: 2800 (Correct answer) -
Math Problem Solving Fails With Generic AI Prompts
By
–
Let's look at a simple example to solve a math problem With Generic Prompt: Input: What is 100*100/200*56?
Output: 1120 (Wrong answer) -
Role-based prompting for industry-specific language models
By
–
Introducing Role-based prompting With role-based prompts, you can tailor the language model to understand the specific language and terminology used in your industry which leads to more accurate and relevant results.
-
10X Better AI Output: Essential Prompt Engineering Tricks
By
–
With products like ChatGPT becoming more mainstream, Prompt engineering has now become more important than ever!! Here is a simple prompt trick to make get 10X better output from these AI models. (A thread)
-
spaCy Floret Vectors: New Approach for Transformer Models
By
–
Here's how we do it in @spacy_io : https://
explosion.ai/blog/floret-ve
ctors
… We're interested to try this trick in transformer models. We haven't really tried that yet. -
Critique of journalists’ ChatGPT interviews
By
–
Please dear journalists, stop with the 'we interviewed ChatGPT' articles. #ChatGPT is in no way capable of 'intelligence' to answer interviews; it just spits out 'obvious' answers without interest. OK, storytelling, but this is borderline dishonest as a procedure.
-
Building a Transformer with simplified layers in nanoGPT
By
–
Yep, that's the one! (as @Thom_Wolf linked earlier too). I'd expect it's possible to build a Transformer with that kind of layer alone, would look much more pleasing. Will see if I can prototype in nanoGPT.