aiqbench analysis
This study challenges the common assumption in spectral foundation models that preprocessing invariance results from learned robustness, demonstrating instead that standard normalization steps often mathematically enforce this property before any training occurs. The analysis reveals that claims of "learned" invariance across various Raman and other modalities are frequently artifacts of these pre-processing pipelines rather than genuine model capabilities. Consequently, researchers should evaluate encoder performance against the baseline provided by input normalization alone to accurately assess what is truly learned during training.
Attributing Preprocessing Invariance in Spectral Foundation Models
arXiv:2608.14227v1 Announce Type: cross Abstract: Preprocessing invariance is an appealing goal for spectral foundation models: a frozen model should remain useful when laboratories preprocess spectra differently. It is usually measured by training a classifier under one preprocessing pipeline and testing it under another, with preserved accuracy read as evidence of learning. We revisit that reading, using a Raman foundation model as a case study. Such models normalize their inputs before any learned parameter is applied. If that normalization maps two differently preprocessed spectra to the same vector, the encoder receives identical inputs, so the invariance cannot be attributed to learning. For a normalization that uses each spectrum's own statistics, this happens exactly when one spectrum is a positive multiple of the other plus a constant. Several standard preprocessing operations take that form. The encoder should therefore be measured against the normalization alone, which has no learned parameters. On six Raman evaluation datasets, the model does not measurably outperform its own normalization. It improves on raw spectra, but so does the normalization alone. Training does improve the encoder over random initialization, and a controlled experiment shows that it learns to ignore a transformation only when that transformation reaches it. A numerical test settles which transformations a given normalization removes. Across released systems in five modalities, most normalizations already remove transformations of that form, and several of those systems claim that invariance as learned. Replicating the comparison on two of them shows no gain either.
Abstract:Preprocessing invariance is an appealing goal for spectral foundation models: a frozen model should remain useful when laboratories preprocess spectra differently. It is usually measured by training a classifier under one preprocessing pipeline and testing it under another, with preserved accuracy read as evidence of learning. We revisit that reading, using a Raman foundation model as a case study. Such models normalize their inputs before any learned parameter is applied. If that normalization maps two differently preprocessed spectra to the same vector, the encoder receives identical inputs, so the invariance cannot be attributed to learning. For a normalization that uses each spectrum's own statistics, this happens exactly when one spectrum is a positive multiple of the other plus a constant. Several standard preprocessing operations take that form. The encoder should therefore be measured against the normalization alone, which has no learned parameters. On six Raman evaluation datasets, the model does not measurably outperform its own normalization. It improves on raw spectra, but so does the normalization alone. Training does improve the encoder over random initialization, and a controlled experiment shows that it learns to ignore a transformation only when that transformation reaches it. A numerical test settles which transformations a given normalization removes. Across released systems in five modalities, most normalizations already remove transformations of that form, and several of those systems claim that invariance as learned. Replicating the comparison on two of them shows no gain either.
Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG)
arXiv:2608.14227 [cs.AI]
(or arXiv:2608.14227v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2608.14227 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 341 words
Related coverage
Reflections on unlearning and inoculation
LessWrong · Sep 20, 2026
EEG-PRISM: Physiologically-Grounded Interpretability of Predictions by EEG Foundation Models
arXiv · Aug 17, 2026
EEG-Xplain: Decoding Neural Black-Boxes of EEG Foundation Models
arXiv · Sep 16, 2026
Pretrained, Frozen, Still Leaking: Auditing Cross-Encoder Attribute Transfer in EEG Foundation Models
arXiv · Jun 9, 2026