Measuring LLM lexical tics: excess vocabulary, baselines, drift
Parent: Writing and Documentation · Topic entry · 16 branches · skill writing-expert/references/tic-measurement.md
A published reference is not available for this topic yet.
Children
- Baseline contamination by AI text (frontier)
- Content-word vs style-word annotation (frontier)
- Counterfactual frequency projection (frontier)
- Era-tagging tic entries (frontier)
- Excess-vocabulary frequency method (frontier)
- Frequency ratio vs frequency gap (frontier)
- Human baselines and control words for tic detection (frontier)
- Humanizer tools and paraphrase evasion (frontier)
- Maintaining a living tic list (frontier)
- Non-native writer false positives (frontier)
- Per-model idiolect profiles (frontier)
- Perplexity and burstiness detectors (frontier)
- RLHF-induced lexical overrepresentation (frontier)
- Retiring decayed tics (frontier)
- Stylometry and AI-text detector limits (frontier)
- Tic drift across model versions (frontier)
Frontier under this node: Baseline contamination by AI text, Content-word vs style-word annotation, Counterfactual frequency projection, Era-tagging tic entries, Excess-vocabulary frequency method, Frequency ratio vs frequency gap, Human baselines and control words for tic detection, Humanizer tools and paraphrase evasion, Maintaining a living tic list, Non-native writer false positives, Per-model idiolect profiles, Perplexity and burstiness detectors, RLHF-induced lexical overrepresentation, Retiring decayed tics, Stylometry and AI-text detector limits, Tic drift across model versions