Science ❯ Computer Science ❯ Artificial Intelligence

Evaluation Methods

Benchmarking Counterfactual Probing Performance Metrics Self-Prediction in AI Parity-Controlled Evaluation Rubric-based Assessment KG-based Evaluation Performance Comparison Datasets Human Evaluation Performance Assessment Robustness in AI Model Reliability Interpretable Models Statistical Analysis Empirical Findings Reproducibility Concealed Information Test

ArXiv Preprint Says Prompts Drive Many LVLM Hallucinations

The authors present tools to test and reduce instruction-induced mistakes in image-text models.

Study Says Google’s AI Overviews Miss About 1 in 10 Answers; Google Disputes Findings