Researchers have introduced Direct Embedding Optimization (DEO), a novel training-free method designed to improve text and multimodal retrieval for negation and exclusion queries. DEO works by decomposing queries into positive and negative components and then optimizing the query embedding using a contrastive objective. This approach avoids the computational costs and deployment complexities associated with traditional embedding adaptation or fine-tuning methods. In evaluations, DEO demonstrated significant improvements over existing baselines, including a +0.0738 nDCG@10 gain on the NegConstraint benchmark and a +6% increase in Recall@5 for multimodal retrieval when compared to OpenAI CLIP. AI
IMPACT Enhances retrieval systems for complex queries, potentially improving accuracy in applications relying on negation and exclusion.
RANK_REASON The cluster contains an academic paper detailing a new method for retrieval. [lever_c_demoted from research: ic=1 ai=1.0]
- Deo
- Direct Embedding Optimization
- large-language models
- NegConstraint
- OpenAI CLIP
- retrieval-augmented generation
- Taegyeong Lee
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →