Regenerating model responses on de-identified production conversation prefixes yields pre-deployment misbehavior rate forecasts that track realized production rates within 2-5x and outperform adversarial-prompt baselines.
OpenAI GPT-5 System Card.https://openai.com/index/gpt-5-system-card/,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Predicting LLM Safety Before Release by Simulating Deployment
Regenerating model responses on de-identified production conversation prefixes yields pre-deployment misbehavior rate forecasts that track realized production rates within 2-5x and outperform adversarial-prompt baselines.