PivotAttack: Rethinking the Search Trajectory in Hard-Label Text Attacks via Pivot Words
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910049366966272 |
|---|---|
| author | Liang, Yuzhi Xiao, Shiliang Wei, Jingsong Lin, Qiliang Li, Xia |
| author_facet | Liang, Yuzhi Xiao, Shiliang Wei, Jingsong Lin, Qiliang Li, Xia |
| contents | Existing hard-label text attacks often rely on inefficient "outside-in" strategies that traverse vast search spaces. We propose PivotAttack, a query-efficient "inside-out" framework. It employs a Multi-Armed Bandit algorithm to identify Pivot Sets-combinatorial token groups acting as prediction anchors-and strategically perturbs them to induce label flips. This approach captures inter-word dependencies and minimizes query costs. Extensive experiments across traditional models and Large Language Models demonstrate that PivotAttack consistently outperforms state-of-the-art baselines in both Attack Success Rate and query efficiency. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_10842 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | PivotAttack: Rethinking the Search Trajectory in Hard-Label Text Attacks via Pivot Words Liang, Yuzhi Xiao, Shiliang Wei, Jingsong Lin, Qiliang Li, Xia Computation and Language Existing hard-label text attacks often rely on inefficient "outside-in" strategies that traverse vast search spaces. We propose PivotAttack, a query-efficient "inside-out" framework. It employs a Multi-Armed Bandit algorithm to identify Pivot Sets-combinatorial token groups acting as prediction anchors-and strategically perturbs them to induce label flips. This approach captures inter-word dependencies and minimizes query costs. Extensive experiments across traditional models and Large Language Models demonstrate that PivotAttack consistently outperforms state-of-the-art baselines in both Attack Success Rate and query efficiency. |
| title | PivotAttack: Rethinking the Search Trajectory in Hard-Label Text Attacks via Pivot Words |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2603.10842 |