data leakage
The unintended exposure of information from the training set to the testing set, which can lead to overly optimistic model evaluations and poor generalization to new data.
- ClinicalLab: Aligning Agents for Multi-Departmental Clinical Diagnostics in the Real World
- Factor Decorrelation Enhanced Data Removal from Deep Predictive Models
- Image Super-Resolution with Guarantees via Conformalized Generative Models
- MLIP Arena: Advancing Fairness and Transparency in Machine Learning Interatomic Potentials via an Open, Accessible Benchmark Platform
- MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models
- Rethinking the Role of Verbatim Memorization in LLM Privacy
- ThinkBench: Dynamic Out-of-Distribution Evaluation for Robust LLM Reasoning