Quantum
When Normalization Selects the Sign: Auditing Robustness Ablations in Quantum Attention
Publication: quant-ph updates on arXiv.org
Author: Owen Friedewald, Srikar Alla, Ali Shiri Sichani, Chi-Ren Shyu
Source published: 05/10/2026, 04:00:00 UTC
Briefing created: 05/10/2026, 04:22:21 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 05/10/2026, 04:13:25 UTC.
Briefings
A four-qubit quantum-attention detector on generated power-grid trajectories showed that removing an input-scaling module altered detection outcomes. While a learned scaling module appeared beneficial at a fixed physical attack budget, matching an upper bound on perturbations at the encoder reversed the ordering. Neither comparison alone established a robustness benefit caused by the module, as initial tests also perturbed clean examples into attacked examples while retaining their original labels. Retraining the linear classification layer restored detection rates but changed individual predictions, leaving the comparison descriptive rather than causal. The evidence is limited to ten seeds, exact simulation, synthetic data, and a restricted set of attacks; classical baselines achieved better clean prediction.
Key points
- Removing an input-scaling module in a four-qubit quantum-attention detector altered detection outcomes on generated power-grid trajectories.
- A learned scaling module appeared beneficial at a fixed physical attack budget, but matching an upper bound on perturbations at the encoder reversed the ordering.
Context and limitations
Evidence covers only ten seeds, exact simulation, synthetic data, and a restricted set of attacks. Classical baselines achieved better clean prediction.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 05/10/2026, 04:22:21 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeif7vfauxflgv6jcmg3rsgas3mna4tyejlnpucuonoaklfkvfoldge