Robertha: Eigenspectrum Regularized Attention for Robust Natural Language Understanding

Andreia Podasca, Anup Das


Abstract
We study asymmetric vulnerability to embedding corruption in encoder-based language models, where uniform perturbations dis-proportionately degrade low-magnitude embeddings compared to high-magnitude ones. Since critical words concentrate in low-norm space, this asymmetry causes catastrophic degradation of grammatical and semantic structure even under moderate corruption. Existing robustness approaches either sacrifice clean performance or fail to generalize to higher corruption levels. To address this problem, we propose Robertha, an attention mechanism built on Modern Hopfield Networks, in which semantic patterns act as stable states (attractors) that pull corrupted embeddings toward correct representations. We introduce iterative refinement for differential recovery: heavily corrupted embeddings require multiple convergence steps, while lightly corrupted embeddings converge quickly. To strengthen this mechanism, we introduce Eigenspectrum Regularization (ESR), which enforces low-rank key structures by controlling eigenvalue entropy, creating strong, well separated attractors with wide recovery basins. Across 13 GLUE and SuperGLUE tasks, Robertha significantly outperforms existing robustness methods while maintaining competitive clean performance.
Anthology ID:
2026.acl-long.696
Volume:
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
July
Year:
2026
Address:
San Diego, California, United States
Editors:
Maria Liakata, Viviane P. Moreira, Jiajun Zhang, David Jurgens
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
15229–15264
Language:
URL:
https://aclanthology.org/2026.acl-long.696/
DOI:
10.18653/v1/2026.acl-long.696
Bibkey:
Cite (ACL):
Andreia Podasca and Anup Das. 2026. Robertha: Eigenspectrum Regularized Attention for Robust Natural Language Understanding. In Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 15229–15264, San Diego, California, United States. Association for Computational Linguistics.
Cite (Informal):
Robertha: Eigenspectrum Regularized Attention for Robust Natural Language Understanding (Podasca & Das, ACL 2026)
Copy Citation:
PDF:
https://aclanthology.org/2026.acl-long.696.pdf
Checklist:
 2026.acl-long.696.checklist.pdf