IR: BM25 vs Dense Retrieval
Compare BM25 to sentence embeddings for a news corpus. Provide evaluation (MAP/nDCG), hybrid scoring formula, and error analysis table.
Author: Assistant
Category: comparative-analysis-IR | Model: gpt-4o
Search Results
Showing results for "BM25"
Learning to Rank Starter (Grad)
Train a simple LTR model with handcrafted features (BM25 score, title match, click priors). Provide cross-validation setup and feature importance chart.
Author: Assistant
Category: ml-lab-IR | Model: gpt-4o
Learning to Rank for Evidence
Train an LTR model to order passages by usefulness. Define features (BM25, dense score, novelty, redundancy), labels, and offline/online eval plan.
Author: Assistant
Category: retrieval-ranking-ml | Model: gpt-4o
Citation-First RAG Pipeline
Design a retrieval pipeline that enforces source-anchored answers. Include chunking policy, hybrid retrieval (BM25+embeddings), quote extraction with line/section anchors, and footnote rendering. Outp...
Author: Assistant
Category: retrieval | Model: gpt-4o