Saved in:
Bibliographic Details
Main Author: Zamir, Or
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.15756
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918343292747776
author Zamir, Or
author_facet Zamir, Or
contents A natural and informal approach to verifiable (or zero-knowledge) ML inference over floating-point data is: ``prove that each layer was computed correctly up to tolerance $δ$; therefore the final output is a reasonable inference result''. This short note gives a simple counterexample showing that this inference is false in general: for any neural network, we can construct a functionally equivalent network for which adversarially chosen approximation-magnitude errors in individual layer computations suffice to steer the final output arbitrarily (within a prescribed bounded range).
format Preprint
id arxiv_https___arxiv_org_abs_2602_15756
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle A Note on Non-Composability of Layerwise Approximate Verification for Neural Inference
Zamir, Or
Cryptography and Security
Machine Learning
A natural and informal approach to verifiable (or zero-knowledge) ML inference over floating-point data is: ``prove that each layer was computed correctly up to tolerance $δ$; therefore the final output is a reasonable inference result''. This short note gives a simple counterexample showing that this inference is false in general: for any neural network, we can construct a functionally equivalent network for which adversarially chosen approximation-magnitude errors in individual layer computations suffice to steer the final output arbitrarily (within a prescribed bounded range).
title A Note on Non-Composability of Layerwise Approximate Verification for Neural Inference
topic Cryptography and Security
Machine Learning
url https://arxiv.org/abs/2602.15756