DEO: Training-Free Direct Embedding Optimization for Negation-Aware Retrieval
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866914380898107392 |
|---|---|
| author | Lee, Taegyeong Park, Jiwon Hwang, Seunghyun Jang, JooYoung |
| author_facet | Lee, Taegyeong Park, Jiwon Hwang, Seunghyun Jang, JooYoung |
| contents | Recent advances in Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) have enabled diverse retrieval methods. However, existing retrieval methods often fail to accurately retrieve results for negation and exclusion queries. To address this limitation, prior approaches rely on embedding adaptation or fine-tuning, which introduce additional computational cost and deployment complexity. We propose Direct Embedding Optimization (DEO), a training-free method for negation-aware text and multimodal retrieval. DEO decomposes queries into positive and negative components and optimizes the query embedding with a contrastive objective. Without additional training data or model updates, DEO outperforms baselines on NegConstraint, with gains of +0.0738 nDCG@10 and +0.1028 MAP@100, while improving Recall@5 by +6\% over OpenAI CLIP in multimodal retrieval. These results demonstrate the practicality of DEO for negation- and exclusion-aware retrieval in real-world settings. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_09185 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | DEO: Training-Free Direct Embedding Optimization for Negation-Aware Retrieval Lee, Taegyeong Park, Jiwon Hwang, Seunghyun Jang, JooYoung Computation and Language Recent advances in Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) have enabled diverse retrieval methods. However, existing retrieval methods often fail to accurately retrieve results for negation and exclusion queries. To address this limitation, prior approaches rely on embedding adaptation or fine-tuning, which introduce additional computational cost and deployment complexity. We propose Direct Embedding Optimization (DEO), a training-free method for negation-aware text and multimodal retrieval. DEO decomposes queries into positive and negative components and optimizes the query embedding with a contrastive objective. Without additional training data or model updates, DEO outperforms baselines on NegConstraint, with gains of +0.0738 nDCG@10 and +0.1028 MAP@100, while improving Recall@5 by +6\% over OpenAI CLIP in multimodal retrieval. These results demonstrate the practicality of DEO for negation- and exclusion-aware retrieval in real-world settings. |
| title | DEO: Training-Free Direct Embedding Optimization for Negation-Aware Retrieval |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2603.09185 |