Omni-Embed-Audio：用多模态 LLM 实现鲁棒的音频-文本检索

View PDF HTML (experimental)

Abstract:Audio-text retrieval systems based on Contrastive Language-Audio Pretraining (CLAP) achieve strong performance on traditional benchmarks; however, these benchmarks rely on caption-style queries that differ substantially from real-world search behavior, limiting their assessment of practical retrieval robustness. We present Omni-Embed-Audio (OEA), a retrieval-oriented encoder leveraging multimodal LLMs with native audio understanding. To systematically evaluate robustness beyond caption-style queries, we introduce User-Intent Queries (UIQs) - five formulations reflecting natural search behaviors: questions, commands, keyword tags, paraphrases, and exclusion-based negative queries. For negative queries, we develop a hard negative mining pipeline and propose discrimination metrics (HNSR, TFR) assessing models' ability to suppress acoustically similar distractors. Experiments on AudioCaps, Clotho, and MECAT show that OEA achieves comparable text-to-audio retrieval performance to state-of-the-art M2D-CLAP, while demonstrating clear advantages in two critical areas: (1) dominant text-to-text retrieval (+22% relative improvement), and (2) substantially superior hard negative discrimination (+4.3%p HNSR@10, +34.7% relative TFR@10), revealing that LLM backbones provide superior semantic understanding of complex queries.

Comments:	Accepted at ACL 2026 Main Conference. Final version
Subjects:	Sound (cs.SD); Computation and Language (cs.CL)
Cite as:	arXiv:2604.18360 [cs.SD]
	(or arXiv:2604.18360v4 [cs.SD] for this version)
	https://doi.org/10.48550/arXiv.2604.18360 arXiv-issued DOI via DataCite

Submission history

From: HaeJun Yoo [view email]
[v1] Mon, 20 Apr 2026 14:50:33 UTC (771 KB)
[v2] Mon, 1 Jun 2026 04:50:48 UTC (770 KB)
[v3] Tue, 7 Jul 2026 23:45:36 UTC (773 KB)
[v4] Wed, 14 Oct 2026 06:38:52 UTC (773 KB)