PersonalQuery-Benchmark: 6,516 queries test retrieval robustness to personalized expression differences
_reachsumit · x · 2026-10-26
The arXiv paper introduces PersonalQuery-Benchmark (PQB), which evaluates how personalized expression differences—syntax and writing errors—shift rankings for a fixed information need. Built from users' historical texts across three Amazon product domains with 6,516 query instances, it benchmarks five retrieval architectures (lexical, learned sparse, dense bi-encoder, late-interaction, hybrid) plus two LLM rankers. Results show measurable ranking variation even with a fixed need, with sensitivity differing clearly across models.
More from Research
- HKU/ZJU/Fudan with Lightspeed launch HiSTEVE: hierarchical VLA agent playing Minecraft via native mouse and keyboard — tencent · 2026-10-26
- Sony researchers apply flow matching to four-stem stereo music source separation in FLOMSS — Shuyang Cui · 2026-10-26
- Zhejiang University and Tencent propose Easy-3Dfy: decoupling object generation and layout for single-image 3D scene construction — Ziang Zhang · 2026-10-26
- One week: 80-year-old cells reset to under 20, AI designing working proteins — rand_longevity · 2026-10-26
- Anthropic interpretability demo shows neurons combining into tiny pirate/wizard personas — Sauers_ · 2026-10-26
- 17-page HMM chapter PDF from Jurafsky & Martin's NLP textbook shared — burny_tech · 2026-10-26