arXiv subjects

Qiliang Lin

Publications and source records attributed to Qiliang Lin.

3 recordsLinked to original sources

arXiv2026

Breadth Beats Depth: Improving GCG-Based Jailbreak Optimization with Breadth-Oriented Suffix Search

Optimization-based jailbreak attacks such as Greedy Coordinate Gradient (GCG) achieve strong effectiveness and transferability by optimizing adversarial suffixes on white-box source models. However, existing GCG-based methods rely on averaged adversarial loss and deep greedy search, which can over-emphasize easy-to-jailbreak behaviors and overlook promising regions of the suffix space. We propose BOSS, a plug-and-play framework that improves GCG-based jailbreak optimization through breadth-oriented suffix search. BOSS uses Tail-Focused Adversarial Loss (TFAL), standard source loss, and behavior coverage to select terminal suffixes, then explores multiple short trajectories and selectively continues promising suffixes. Experiments on public benchmarks show that BOSS improves attack success rates across multiple GCG-based methods while reducing optimization time.

cs.CL↗

arXiv2026

Global Existence for 3D Anisotropic MHD system with Horizontal Dissipation and Small Horizontal Variations

This paper establishes the global well-posedness for the 3D anisotropic MHD system with partial dissipation: $\Delta_\mathrm{h}u$ for velocity and $\partial_1^2b$ for magnetic field, near background field $(0,1,0)$. Crucially, only horizontal components $(u^\mathrm{h}_0,b^\mathrm{h}_0)$ need to be small in $H^2(\R^3)$, while $(u^3_0,b^3_0)$ can be arbitrarily large. Our analysis develops novel techniques including component-decoupled energies and iterative control of dangerous nonlinearities using the background field structure. This establishes the global result for anisotropic MHD equations allowing large vertical data, breaking the full-smallness requirement of previous works.

math.AP↗

arXiv2026

PivotAttack: Rethinking the Search Trajectory in Hard-Label Text Attacks via Pivot Words

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.

cs.CL↗