Tag
EarlyEval introduces a lightweight framework that predicts LLM agent outcomes from intermediate behavior to reduce evaluation costs by halting runs early with high accuracy. It eliminates 13%-26% of agent steps and up to 44% input tokens across benchmarks like SWE-bench Verified, TerminalBench, and Toolathlon.
This paper presents an explainable machine learning framework using SHAP analysis to predict and profile broadband adoption disparities at the census-tract level in the United States, identifying key factors like income and education for policy targeting.
This paper presents a controlled comparison of seven categorical encoding methods on the IEEE-CIS fraud dataset, evaluating accuracy, interpretability, and computational trade-offs with a fixed LightGBM learner, finding that entity embeddings achieve the highest AUC-ROC but no encoder dominates both metrics.
This paper proposes EMA-FS, an algorithm-level optimization that uses exponential moving averages of per-feature split gains to selectively construct histograms only for top-K features during GBDT training, achieving up to 2.61x speedups while maintaining or improving accuracy through implicit regularization.
A blog post from Flyback demonstrates how a LightGBM feature that ranked #1 in importance actually worsened predictions due to target encoding leakage, highlighting the danger of relying solely on feature importance metrics.