Unverified50% confidenceOpinionExact time
A model can achieve 99% accuracy on an in-distribution test set yet drop to random-guessing level on text from new models.
1
Sources
50%
Confidence
Long-term
Relevance
11/18/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe experiments designed control groups for persistence and dominance results, keeping the accuracy of random guessing at around 0.27.77% similarUnverifiedRandom accuracy is about 0.33 for three-class classification and about 0.25 for four-class classification; only the magnitude of improvement relative to the random level can reflect whether the model has truly learned anything.77% similarUnverifiedWhen evaluating a binary classification model, the majority-class baseline when positive and negative samples are evenly split means a model that always predicts 'normal' can also achieve 50% accuracy.77% similarUnverifiedIn stochastic sampling decoding, a high-confidence predictive distribution can still sample low-probability tokens, so testing the distribution alone cannot guarantee the reliability of the tokens actually written.76% similarUnverifiedDuring testing, cases emerged where a model claimed 95% confidence in a move but scored 0 points according to the evaluation criteria.76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1183183API
curl https://kongchang.com/api/v1/knowledge/claims/1183183MCP
get_claim(id=1183183)