justin @jxtngx

Developer Advocate @LightningAI | @drexeluniv @floridaatlantic | @USMC S. FL & NYC Joined March 2017

Tweets

1K
Followers

294
Following

398
Likes

2K

Hugh Zhang @hughbzhang

2 days ago

Data contamination is a huge problem for LLM evals right now. At Scale, we created a new test set for GSM8k *from scratch* to measure overfitting and found evidence that some models (most notably Mistral and Phi) do substantially worse on this new test set compared to GSM8k.

32 189 955 210K 319

Download Image

justin @jxtngx

a day ago

Who’s building AI in Miami

0 0 0 101 0

justin @jxtngx

3 days ago

¿uh, didn’t google create flutter and dart?

TechCrunch @TechCrunch

3 days ago

¿uh, didn’t google create flutter and dart?

6 18 58 37K 5

1 0 0 415 0

justin @jxtngx

4 days ago

tools used in this quick study: @GroqInc @MistralAI @ollama @llama_index @LightningAI @cohere @OpenAI

justin @jxtngx

4 days ago

tools used in this quick study: @GroqInc @MistralAI @ollama @llama_index @LightningAI @cohere @OpenAI

1 0 2 204 0

Download Image

0 0 0 131 0

justin @jxtngx

5 days ago

eliza a chat agent from mit, made in the 60s en.wikipedia.org/wiki/ELIZA

0 0 0 112 1

LlamaIndex 🦙 @llama_index

6 days ago

Multi-Stage Advanced RAG with @llama_index and @cohere reranking Making retrieval a multi-hop process 💡 instead of single-shot can ensure better LLM context and reduce hallucinations. This article by Michael R. (@kxsystems) highlights an example of this multi-stage…