> What it does was published in arxiv back in 2025

This is false. You made a statement which was false.

If you want to say that the author of the paper made a really bad knockoff of Jev in 2026 after Jev was published, then say that.

> Jev is basically the embeddings side of an LLM

This is how the paper you are referencing is describing the "key contribution" as it relates to embeddings:

> Novel state representation techniques using Azure OpenAI embeddings (3072 dimensions) with sales-specific features

And now you are claiming that the author of the paper created:

> "generalized" version of same

It's a bit hard to guess what you are trying to say, but if I were to steelman your argument, I would guess that you mean: training a model with a vocabulary that has some tokens representing things like "OPTION_A", "OPTION_B" is in your mind the same thing as "generalized version of sales-specific features in embeddings"? Is this what you were trying to say?

I don't know how good Jev's "confidence scores" are, but I would be surprised if they were in any sense better than logits from some good LLM. One advantage of Jev here is that the confidence scores are easy to access. Most LLM API providers don't provide an easy/convenient way to access the logits. But that's a minor point, you could of course build something like this with LLMs (and many people have).

Would Jev be more accurate in a specific task if it had been developed only for that task, as opposed to general purpose? Of course it would. So, sure, Jev is trading accuracy for generality. According to you "nobody thought it's a good tradeoff", which again is false, there was huge demand for a cheap and accurate general purpose classifier.

Jev is:

- accurate

- general purpose

- fast and cheap

Models we had before Jev had at most 2/3 of above qualities, but none of them were 3/3.

Copypasta:

• A reinforcement learning architecture specifically designed for sales conversation analysis and conversion prediction

• A synthetic data generation pipeline leveraging GPT-4O to create diverse and realistic sales conversations

• Novel state representation techniques using Azure OpenAI embeddings (3072 dimensions) with sales-specific features

• A meta-learning approach enabling the system to express confidence in its predictions based on conversation similarity to training data

• Integration mechanisms providing real-time guidance within existing sales platforms

• Extensive comparative evaluation demonstrating significant performance improvements over LLM-based approaches