Thread

1. hbarka · 1789982542 · focus · HN ↗

   If Jev is fundamentally trained using RLCD while you’re building on a Qwen model that was trained using RLHF, how can the resulting model be considered Jev-like?

   1. llm_nerd · 1789990471 · focus · HN ↗

      > If Jev is fundamentally trained using RLCD

      Big if. More likely, it seems, is they started with an open LLM model and fine-tuned and repurposed it via their "RLCD" process.

      1. jrmg · 1790001649 · focus · HN ↗

         The FAQ on the Jet announcement (<a href="https:&#x2F;&#x2F;typesafe.ai&#x2F;blog&#x2F;introducing-system-one-models-and-jev" rel="nofollow">https:&#x2F;&#x2F;typesafe.ai&#x2F;blog&#x2F;introducing-system-one-models-and-j...) claims it was trained with their own data:

         Where Does Our Training Data Come From?

         TypeSafe is primarily a data research lab, which is how the biggest results in AI get made. We make all the data ourselves. We wouldn’t train on your data even if you asked us to (no offense). We do some pretty sophisticated stuff, but if you want to find out more, we’d have to hire you.

         1. llm_nerd · 1790004393 · focus · HN ↗

            &gt; We make all the data ourselves.

            I mean, this claim is simply preposterous, and is discountable as ridiculous nonsense on its face.

            1. brookst · 1790021446 · focus · HN ↗

               Really? I read it as saying they use 100% synthetic data, which is made in house. Why is that preposterous?

               1. alfalfasprout · 1790025629 · focus · HN ↗

                  Yeah, and synthetic data is pretty darn good nowadays especially with access to frontier models.

                  1. Zambyte · 1790037634 · focus · HN ↗

                     &gt; especially with access to frontier models

                     ... so not made in house.

Unofficial Hacker News client; not affiliated with Y Combinator.