Reproduce “Context Aware Query Rewriting for Text Rankers using LLM” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Abhijit Anand, Venktesh V, Vinay Setty, Avishek Anand
The Context Aware Rewriter (CAR) framework enhances text ranking performance by using Large Language Models (LLMs) to generate context-aware query rewrites exclusively during the training phase. This approach improves ranking accuracy for ambiguous queries, showing gains of up to 33.1% in nDCG@10 for passage ranking without incurring LLM inference costs.
Key insights
- Grounding LLM-generated rewrites with specific relevant document context during training effectively prevents topic drift and ensures semantic alignment.
- Knowledge from powerful LLMs can be effectively transferred to a smaller, more efficient ranker through fine-tuning on context-aware rewritten queries.
- Using supervised passage selection for long documents further refines the context provided to LLMs, enhancing rewrite quality and downstream ranking performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.