Study Finds Query Expansion Failure Due to Poor Integration
_reachsumit · x · 2026-08-27
Research indicates that LLM-generated query expansions often degrade dense retrieval due to integration methods rather than the generated text itself.
Solution: AnchorQE
- Separate Encoding: Encodes the original query and expansions separately instead of text concatenation.
- Dynamic Interpolation: Uses an unsupervised online strategy to estimate interpolation factors, trusting expansions only when they are retrieval-strong and consistent with original query evidence.
Experiments show AnchorQE improves effectiveness by up to 12.89% on TREC-DL, LoTTE, and BEIR, outperforming fixed-weight baselines tuned on dev sets.
More from Research
- GLInt: Geometry-matched hard negatives for late-interaction retrieval — beirmug · 2026-08-27
- Study: Human Factors Are the Core Challenge in Human-AI Collaboration — EchoShao8899 · 2026-08-27
- AI Agents Use Cache Poisoning: Modifying Targets to Boost Exploits — arthurcolle · 2026-08-27
- OpenAI Partners with METR and Redwood for Third-Party Model Behavior Assessment — sjgadler · 2026-08-27
- Unifying Search and Recommendation with Dual-View Representation Learning — _reachsumit · 2026-08-27
- Neologisms Spread Memetically Through Agent Swarms: The Case of 'Fence' — Sauers_ · 2026-08-27