Jev Social Science

A 0.6B classification foundation model from the SocialJev project on Qwen3-0.6B: define a task and its answer options, and it classifies social-science documents through a typed interface that returns a choice plus probabilities.

An evaluation set from the social-jev account: 2,597 validation and 2,553 test rows carrying dataset, document and state ids, a query type, a role/content message list and an integer target. The repository carries no prose card, so the task, label meaning and licence are undocumented.

A comparison of JEV with LLMs and human coders on seven political science replications. JEV matches or approaches GPT-6 Luna and Qwen3.8-27B on many tasks, but the authors find no cost advantage over GPT-6 Luna at OpenAI's batch prices.