# CORA: A Protocol for Diagnosing Boundary Robustness in Text-to-Audio Retrieval under Query Reformulations

arXiv: https://arxiv.org/abs/2609.33433 · HTML: https://mlsanity.com/paper/2609.33433/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 13 of its 18 references that are in the map.
- First posted on arXiv 27 Sep 2026.

## Builds on (most cited first, 13 of 13)

- [Large-scale Contrastive Language-Audio Pretraining with Feature Fusion and Keyword-to-Caption Augmentation](https://mlsanity.com/paper/2211.06687/) (2022)
- [Contrastive Learning with Hard Negative Samples](https://mlsanity.com/paper/2010.04592/) (2020)
- [Clotho: An Audio Captioning Dataset](https://mlsanity.com/paper/1910.09387/) (2019)
- [WavCaps: A ChatGPT-Assisted Weakly-Labelled Audio Captioning Dataset for Audio-Language Multimodal Research](https://mlsanity.com/paper/2303.17395/) (2023)
- [Verbalized Sampling: How to Mitigate Mode Collapse and Unlock LLM Diversity](https://mlsanity.com/paper/2510.01171/) (2025)
- [Evaluating the Robustness of Retrieval Pipelines with Query Variation Generators](https://mlsanity.com/paper/2111.13057/) (2021)
- [Judge's Verdict: A Comprehensive Analysis of LLM Judge Capability Through Human Agreement](https://mlsanity.com/paper/2510.09738/) (2025)
- [T-CLAP: Temporal-Enhanced Contrastive Language-Audio Pretraining](https://mlsanity.com/paper/2404.17806/) (2024)
- [Advancing Multi-grained Alignment for Contrastive Language-Audio Pre-training](https://mlsanity.com/paper/2408.07919/) (2024)
- [Hard-Negative Sampling for Contrastive Learning: Optimal Representation Geometry and Neural- vs Dimensional-Collapse](https://mlsanity.com/paper/2311.05139/) (2023)
- [Do Audio-Language Models Understand Linguistic Variations?](https://mlsanity.com/paper/2410.16505/) (2024)
- [The language of sound search: Examining User Queries in Audio Search Engines](https://mlsanity.com/paper/2410.08324/) (2024)
- [Omni-Embed-Audio: Leveraging Multimodal LLMs for Robust Audio-Text Retrieval](https://mlsanity.com/paper/2604.18360/) (2026)

## Abstract (arXiv)

Text-to-Audio (T2A) retrievers are typically evaluated with caption style queries, but the same user intent can be expressed in many forms. We introduce CORA (Caption-Offset Retrieval for Audio), a caption anchored diagnostic protocol that rewrites each source caption into five intent preserving forms (Command, Question, Indirect, Key phrase, and Statement) while fixing the target audio. By tracking the same target across query forms, CORA defines RankDrop, a metric revealing failures hidden by Recall@k. Using Pearson's correlation coefficient r, we find that RankDrop is weakly associated with raw text space movement (r=0.084), but strongly associated with Target Alignment Loss and Target Boundary Margin Degradation (r=0.508 and r=0.615). The same pattern appears in OEA retrievers, where RankDrop is better explained by boundary degradation (r=0.472/0.478) than by query movement (r=0.084/0.046). Overall, these results suggest that robust T2A retrieval requires preserving the target's boundary advantage over competing audio under reformulation.