Skip to main content
Search

# Safety Detection over Reasoning Traces

1 related articles

Subscribe to updates
  1. openbiox Bizard

    ReasoningShield moves safety checks onto the reasoning trace: its open 1B and 3B moderators reach 91.8% average F1 on open-source reasoning-model subsets

    Large reasoning models can write unsafe content into intermediate reasoning steps even when the final answer looks safe. ReasoningShield moves safety detection onto chain-of-thought traces, with a taxonomy of 10 risk categories and 3 safety levels plus 7K training samples and 2.2K human-annotated test samples. Its 1B and 3B moderators are trained in two stages, and the 3B model reaches 91.8% average F1 on open-source reasoning-model subsets and 81.4% on commercial reasoning-model subsets.

You’re all caught up