Input Text
Refined Output
Refined text will appear here after you click Refine →
Edit Log
Analysis
Predictability ?
—
High = more AI-formulaic tokens flagged
Similarity ?
—
SBERT cosine (threshold 0.95)
NLI ?
—
Semantic entailment check
Token Δ
—
Fraction of tokens changed (max 35%)
Binoculars Detection Score ?
—
Detection simulation disabled.
Set
ASR_ENABLE_BINOCULARS=true in .env and
restart the server to activate the Binoculars scorer.
Academic Prose Mode ?
Calibration ?
Loading calibration data…
Configured threshold
—
Domain-calibrated threshold
—
Academic reduction
—
Contributors
Label this refinement
Run a refinement first, then label the source text below.
✓ Label submitted. Thank you — every sample improves the threshold estimate.