Skip to main content
Search

# Thinking Jev

1 related articles

Subscribe to updates
  1. arXiv

    No training needed to decide: frozen Qwen3.5-4B reaches 81.4% on public JevBench items, matching training-free readouts on the same backbone

    A Microsoft team presents LLM-as-Jev, which turns off-the-shelf causal language models into decision models that return a probability distribution over candidate options without changing architecture or adding classification heads. Frozen Qwen3.5-4B scores 81.4% on the 231 public JevBench items, matching or slightly exceeding the training-free SemIf readout (80.5%) on the same backbone and beating the fine-tuned reflex 4B (79.7%). Fine-tuning brings clear gains mainly for the 0.6B model and high-cardinality routing, and KL anchors keep generation behavior from degrading.

You’re all caught up