XGBoost

Stub. Created as a placeholder — expand with vault-sourced content.

What it is

Gradient-boosted decision tree library with learning-to-rank objectives (rank:ndcg, rank:map, rank:pairwise). Frequently deployed as a feature-based LTR reranker, including via ONNX in serving engines.

Vault references (existing coverage)

• Dropbox — search ranking model trained with XGBoost on human + LLM-labeled data
• E-commerce Search and Recommendation with Vespa — LTR via XGBoost (ONNX)
• Mirror Mirror - All About Search Suggestions — XGBoost rescoring via Vespa / ES LTR plugin
• Hybrid Search and Learning-to-Rank with Metarank — XGBoost as a LambdaMART implementation; missing-value handling
• Metarank — ships an xgboost model for its LambdaMART ranker

TODO

• Document ranking objectives → link to Ranking Objectives.

Related

• LambdaMART · LightGBM · CatBoost · Learning to Rank · Ranking Objectives · Metarank