@tobi: LCM for Large Choice Models, or what are we going to call the Jevs?
Summary
The post discusses the naming of Large Choice Models and questions what to call 'Jevs' in the AI context.
Similar Articles
Jev / TypesafeAI is revolutionary as LLM’s
Jev is a novel AI model that outputs scores, choices, or binary decisions, praised for its speed, affordability, and accuracy when queried creatively, unlike traditional frontier models.
Jev introduces a new shape of LLM - System One, aka Decision Models
Jev is a new AI model introduced by TypeSafe AI, categorized as a System One or Decision Model, that outputs floating-point numbers for classification tasks like spam detection and ranking with competitive pricing.
Typesafe's JEV model work as an LLM [P]
Describes a conversational AI system that uses Typesafe's Jev non-generative model to select from pre-written responses through parallel classification and scoring, providing a cost-effective and transparent alternative to generative LLMs.
@neural_avb: https://x.com/neural_avb/status/2107173836273938662
The author documents training a choice-order-invariant Jev-style System One decision model, releasing a 276K-example dataset and a 0.4B-parameter lightweight model, with attention-mask and positional-id tricks applied to a converted Qwen3 model. The post also critiques Jev's sensitivity to choice ordering and its non-generative probabilistic nature.
@NFT_Chen: Awesome! Instantly transform any LLM into a Jev model with a localized decision engine! LLM2Jev brings Jev's unified Ch…
LLM2Jev is an open-source tool that adapts local HuggingFace models to perform structured decisions with Choice, Score, and Noul frameworks, offering prefill-only inference and integration with Transformers and SGLang.