Jev vs AnyJev
AnyJev, from Nokia Applied Research, turns an LLM you already have into a decision model by reading its hidden states.
What AnyJev is
AnyJev is Apache-2.0. It does not train a task model. It reads an existing LLM's hidden states, applies debiasing and calibration, and serves its own decision endpoint. The LLM can be a Hugging Face model or a vLLM server.
Training-free still means you operate it
You still choose the base model, serve it, and check calibration on your own cases. Accuracy depends on the model you point it at. AnyJev does not include a hosted Jev account.
When it fits
Use AnyJev when that LLM is already in your stack and you want choice, score and yes/no on top of it. Use JEV Gateway when the decision should be a single hosted API call.