Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- James BaileySuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works38 from public data
- 121
- Safety at Scale: A Comprehensive Survey of Large Model and Agent Safety76
This survey provides a systematic review of current safety research on large models, covering Vision Foundation Models, Large Language Models, Vision-Language Pre-training models, Vision-Language Models, Diffusion Models, and large-model-powered Agents, and presents a comprehensive taxonomy of safety threats to these models.
- BackdoorLLM: A Comprehensive Benchmark for Backdoor Attacks and Defenses on Large Language Models49
BackdoorLLM is the first comprehensive benchmark for systematically evaluating backdoor threats in text-generation LLMs and provides a diverse suite of attack modalities, including data poisoning, weight poisoning, hidden-state manipulation, and chain-of-thought hijacking.
- 44
- Distilling Cognitive Backdoor Patterns within an Image43
Despite the various forms and sizes of trigger patterns used by different attacks, the CPs of backdoor samples are all surprisingly and suspiciously small, so one can leverage the learned mask to detect and remove backdoor examples from poisoned training datasets.
- X-Transfer Attacks: Towards Super Transferable Adversarial Attacks on CLIP32
X-Transfer is introduced, a novel attack method that exposes a universal adversarial vulnerability in CLIP and generates a Universal Adversarial Perturbation (UAP) capable of deceiving various CLIP encoders and downstream VLMs across different samples, tasks, and domains.
- 31
- Detecting Backdoor Samples in Contrastive Language Image Pretraining26
This work analyzes the representations of backdoor-poisoned samples learned by CLIP models and finds that they exhibit unique characteristics in their local subspace, i.e., their local neighborhoods are far more sparse than that of clean samples.
- Shortcuts Everywhere and Nowhere: Exploring Multi-Trigger Backdoor Attacks25
This study explores the concept of Multi-Trigger Backdoor Attacks (MTBAs), where multiple adversaries leverage different types of triggers to poison the same dataset, and demonstrates that multiple triggers can coexist, overwrite, or cross-activate one another.
- Imbalanced gradients: a subtle cause of overestimated adversarial robustness16
A margin decomposition (MD) attack that decomposes a margin loss into individual terms and then explores the attackability of these terms separately via a two-stage process to exploit imbalanced gradients.
- LDReg: Local Dimensionality Regularized Self-Supervised Learning15
By increasing the local intrinsic dimensionality, it is demonstrated through a range of experiments that LDReg improves the representation quality of SSL and the results show that LDReg can regularize dimensionality at both local and global levels.
- 9
- AutoBackdoor: Automating Backdoor Attacks via LLM Agents7
This work introduces AutoBackdoor, a general framework for automating backdoor injection, encompassing trigger generation, poisoned data construction, and model fine-tuning via an autonomous agent-driven pipeline, and demonstrates that the method achieves over 90% attack success with only a small number of poisoned samples.
- Neural Architecture Search via Combinatorial Multi-Armed Bandit7
A tree-based method called Nested Monte-Carlo Search is used to tackle the CMAB-NAS problem and discovers a cell structure that achieves a low error rate that is comparable to the state of the art, using only 0.58 GPU days, which is 20 times faster than current tree-search methods.
- Safety in Embodied AI: A Survey of Risks, Attacks, and Defenses6
This survey provides a comprehensive and structured review of safety research in embodied AI, examining attacks and defenses across the full embodied pipeline, from perception and cognition to planning, action and interaction, and agentic system and introduces a multi-level taxonomy that unifies fragmented lines of work.
- 5
- Toward Universal and Transferable Jailbreak Attacks on Vision-Language Models4
This work proposes Universal and transferable jailbreak (UltraBreak), a framework that constrains adversarial patterns through transformations and regularisation in the vision space, while relaxing textual targets through semantic-based objectives to enable universal and transferable jailbreaks.
- T2UE: Generating Unlearnable Examples from Text Descriptions4
Text-to-Unlearnable Example (T2UE), a novel framework that enables users to generate UEs using only text descriptions, demonstrates the feasibility of ''zero-contact data protection', where personal data can be safeguarded based solely on their textual descriptions, eliminating the need for direct data exposure.
- AUDETER: A Large-scale Dataset for Deepfake Audio Detection in Open Worlds3
Extensive experiments show that existing detection models struggle to generalise to novel deepfakes and human speech in AUDETER, whereas XLR-based detectors trained on AUDETER achieve strong cross-domain performance across multiple benchmarks, achieving an EER of 1.87% on In-the-Wild.
- Towards Million-Scale Adversarial Robustness Evaluation With Stronger Individual Attacks3
This work proposes a novel individual attack method, Probability Margin Attack (PMA), which defines the adversarial margin in the probability space rather than the logits space, and analyzes the relationship between PMA and existing cross-entropy or logits-margin-based attacks, showing that PMA can outperform the current state-of-the-art individual methods.
- Semantic-aware Adversarial Fine-tuning for CLIP2
Semantic-aware Adversarial Fine-Tuning (SAFT), which fine-tunes CLIP's image encoder with semantic-aware AEs, and outperforms current methods, achieving substantial improvements in zero-shot adversarial robustness across 16 datasets.
- AudioMosaic: Contrastive Masked Audio Representation Learning2
Extensive experiments show that AudioMosaic achieves state-of-the-art performance on several standard audio benchmarks under both linear probing and fine-tuning and that integrating the pretrained AudioMosaic encoder into audio-language models improves performance on audio-language tasks.
- Just Ask: Curious Code Agents Reveal System Prompts in Frontier LLMs2
This work identifies system prompt extraction as an emergent vulnerability intrinsic to code agents and presents a self-evolving framework that autonomously discovers effective extraction strategies through interaction alone, exposing recurring design- and architecture-level vulnerabilities.
- 2
- Expose Before You Defend: Unifying and Enhancing Backdoor Defenses via Exposed Models2
The results demonstrate the importance of backdoor exposure for backdoor defense, showing that the exposed models can significantly benefit a range of downstream defense tasks, including backdoor label detection, backdoor trigger recovery, backdoor model detection, and backdoor removal.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.