# Prototype-Rectified Iterative Self-supervised Manifold Denoising under Severe Acoustic Shift

arXiv: https://arxiv.org/abs/2608.15037 · HTML: https://mlsanity.com/paper/2608.15037/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 12 of its 34 references that are in the map.
- First posted on arXiv 15 Aug 2026.

## Builds on (most cited first, 12 of 12)

- [Large-scale Contrastive Language-Audio Pretraining with Feature Fusion and Keyword-to-Caption Augmentation](https://mlsanity.com/paper/2211.06687/) (2022)
- [A multi-device dataset for urban acoustic scene classification](https://mlsanity.com/paper/1807.09840/) (2018)
- [ReCLAP: Improving Zero Shot Audio Classification by Describing Sounds](https://mlsanity.com/paper/2409.09213/) (2024)
- [Adapting Language-Audio Models as Few-Shot Audio Learners](https://mlsanity.com/paper/2305.17719/) (2023)
- [E-BATS: Efficient Backpropagation-Free Test-Time Adaptation for Speech Foundation Models](https://mlsanity.com/paper/2506.07078/) (2025)
- [Uncertainty-Aware Vision-Language Segmentation for Medical Imaging](https://mlsanity.com/paper/2602.14498/) (2026)
- [PCA++: How Uniformity Induces Robustness to Background Noise in Contrastive Learning](https://mlsanity.com/paper/2511.12278/) (2025)
- [Efficient Text-Guided Convolutional Adapter for the Diffusion Model](https://mlsanity.com/paper/2602.14514/) (2026)
- [UNITY: Attention Flow Networks for Adaptive Conditioning in Diffusion](https://mlsanity.com/paper/2606.20971/) (2026)
- [tinyCLAP: Distilling Constrastive Language-Audio Pretrained Models](https://mlsanity.com/paper/2311.14517/) (2023)
- [Domain Adaptation Method and Modality Gap Impact in Audio-Text Models for Prototypical Sound Classification](https://mlsanity.com/paper/2506.04376/) (2025)
- [Grad-CL: Source Free Domain Adaptation with Gradient Guided Feature Disalignment](https://mlsanity.com/paper/2509.10134/) (2025)

## Abstract (arXiv)

Audio-Text Foundation Models (ATMs) fail catastrophically under severe acoustic noise, yet existing adaptation strategies either rely on gradient-based Test-Time Adaptation (TTA), which reinforces noise rather than signal, or on prompt tuning that requires privileged noise annotations unavailable at inference. We address these failures with PRISM (Prototype-Rectified Iterative Self-supervised Manifold Denoising), a training-free, source-free TTA framework grounded in the Affine Noise Hypothesis: severe acoustic noise induces a low-rank affine shift in the multimodal latent space, with more than 90% of distortion energy confined to the leading 60 principal components. PRISM estimates and reverses this distortion from an unlabeled target batch using frozen text prototypes as geometric anchors via three closed-form geometric corrections compiled into a single static projection matrix by Affine Bias Regression. At inference, adaptation reduces to one matrix-vector multiplication in 0.0009 ms, making it substantially faster than gradient-based TTA while requiring no additional training. On UrbanSound8K, PRISM improves over the zero-shot baseline by 12.94 percentage points and surpasses an oracle-assisted TTA baseline by 9.41 percentage points, despite never observing its privileged augmented noise prompts. We further identify the Polyphonic Trap, a principled failure mode of subspace deflation for broadband classes, and resolve it via Confidence-Aware Regression (CAR), recovering up to 8.16 percentage points for the worst-affected class.