RefusalScope
Zero-dependency detector and classifier for LLM refusals, deflections, and capability disclaimers — CLI + library with scoring, categories, JSONL batch, and a CI gate.
reference Python implementation of the Cognometric Fingerprint Specification v1.0 — cognitive observability for LLM agents (hallucination/refusal/drift/sycophancy detection). MIT.
PAN@CLEF 2026 Reasoning Trajectory Detection: 1st place source detection (0.85 F1) and 3rd safety classification (0.66 F1) — feature robustness under domain shift.
Add a description, image, and links to the refusal-detection topic page so that developers can more easily learn about it.
To associate your repository with the refusal-detection topic, visit your repo's landing page and select "manage topics."