Improving Information Extraction with Learned Queries
Ranking
Overall
81
Content
100
Popularity
37
Observed public metrics from 1 member.
Merged summary
TL;DR - This paper shows that learning better, document-specific queries can improve clinical information extraction more than simply using larger LLMs. Its LoQ and FeedQ methods raise performance by 18.6 F1 points across four clinical benchmarks and five models.
- LoQ generates tailored lists of questions for each source document.
- FeedQ iteratively optimizes questions using feedback from extraction outcomes.
- Fine-tuned 4B-parameter question generators match or beat expert-derived baselines and outperform much larger untuned models.
- The authors release a dataset containing 12,820 optimized questions.
Sources (1)
Improving Information Extraction with Learned Queries
Public signals
Semantic Scholar citations 0 · Semantic Scholar influential citations 0
TL;DR - This paper shows that learning better, document-specific queries can improve clinical information extraction more than simply using larger LLMs. Its LoQ and FeedQ methods raise performance by 18.6 F1 points across four clinical benchmarks and five models.
- LoQ generates tailored lists of questions for each source document.
- FeedQ iteratively optimizes questions using feedback from extraction outcomes.
- Fine-tuned 4B-parameter question generators match or beat expert-derived baselines and outperform much larger untuned models.
- The authors release a dataset containing 12,820 optimized questions.