AB
Adel Bibi
cs.LGcs.AIcs.CVcs.CLcs.CRstat.MLcs.CYmath.OCcs.HCcs.MA
On Valency
published · living versionsW_t9xernyt·v1 · currentpublished
How Much Do Activation Monitors Benefit from Original Model Computations?
with James Oldfield, Shawn Im, Jay J Culligan, Philip Torr +1
1 version
Preprints & journals
75 papers in the corpus · 2018–2026How Much Do Activation Monitors Benefit from Original Model Computations?2610.24613v1 · James Oldfield, Shawn Im, Jay J Culligan et al.2026 · 0 citationsarXivon Valency
APEX-Harm: Evaluating Agent Safety in Shared Professional Workspaces2610.13623v1 · Ziyi Yang, Adel Bibi, Guohao Li et al.2026 · 0 citationsarXiv
BazaarBench: Delegation Safety in Decentralized C2C Marketplaces Run by LLM Agents2610.06748v1 · Ziyan Wang, Shuqing Shi, James Oldfield et al.2026 · 0 citationsarXiv
Towards a Unified Misuse Monitoring Benchmark2610.07089v1 · Aniruddh Pramod, James Oldfield, Adel Bibi2026 · 0 citationsarXiv
Towards Better Exploration in Sequential Test-Time Scaling2609.39632v2 · Joseph Rance, Fabio Pizzati, Juil Sock et al.2026 · 0 citationsarXiv
Stealth Apart, Harm Together: Skill Cascading Attacks on Skill-Based Agent Systems2609.30383v2 · Zihao Zhu, Siwei Lyu, Adel Bibi et al.2026 · 0 citationsarXiv
Running the Gauntlet: Hard Agentic Tasks2606.14397v5 · Mykola Vysotskyi, Runqi Lin, Grzegorz Biziel et al.2026 · 0 citationsarXiv
-Monitor: Dynamic Safety Monitoring for Diffusion LLMs via Hesitation Signals2605.25893v2 · Aoxi Liu, Yupeng Chen, James Oldfield et al.2026 · 0 citationsarXiv
The Authorization-Execution Gap Is a Major Safety and Security Problem in Open-World Agents2605.11003v1 · Baoyuan Wu, Qingshan Liu, Adel Bibi et al.2026 · 0 citationsarXiv
OMNI-LEAK: Orchestrator Multi-Agent Network Induced Data Leakage2602.13477v2 · Akshat Naik, Jay Culligan, Yarin Gal et al.2026 · 0 citationsarXiv
Reliable and Responsible Foundation Models: A Comprehensive Survey2602.08145v1 · Xinyu Yang, Junlin Han, Rishi Bommasani et al.2026 · 0 citationsarXiv
Safer by Diffusion, Broken by Context: Diffusion LLM's Safety Blessing and Its Failure Mode2602.00388v2 · Zeyuan He, Yupeng Chen, Lang Lin et al.2026 · 0 citationsarXiv
The Alignment Curse: Modality Alignment Supercharges Audio Attacks via Text Transfer2602.02557v2 · Yupeng Chen, Junchi Yu, Aoxi Liu et al.2026 · 0 citationsarXiv
It's a TRAP! Task-Redirecting Agent Persuasion Benchmark for Web Agents2512.23128v3 · Karolina Korgul, Yushi Yang, Arkadiusz Drohomirecki et al.2025 · 1 citationarXiv
Unforgotten Safety: Preserving Safety Alignment of Large Language Models with Continual Learning2512.10150v1 · Lama Alssum, Hani Itani, Hasan Abed Al Kader Hammoud et al.2025 · 1 citationarXiv
Measuring what Matters: Construct Validity in Large Language Model Benchmarks2511.04703v1 · Andrew M. Bean, Ryan Othniel Kearns, Angelika Romanou et al.2025 · 7 citationsarXiv
Tahakom LLM Guidelines and Recipes: From Pre-training Data to an Arabic LLM2510.13481v2 · Areej AlOtaibi, Lina Alyahya, Raghad Alshabanah et al.2025 · 0 citationsarXiv
ToolTweak: An Attack on Tool Selection in LLM-based Agents2510.02554v1 · Jonathan Sneh, Ruomei Yan, Jialin Yu et al.2025 · 0 citationsarXiv
Beyond Linear Probes: Dynamic Safety Monitoring for Language Models2509.26238v4 · James Oldfield, Philip Torr, Ioannis Patras et al.2025 · 0 citationsarXiv
BiasBusters: Uncovering and Mitigating Tool Selection Bias in Large Language Models2510.00307v2 · Thierry Blankenstein, Jialin Yu, Zixuan Li et al.2025 · 0 citationsarXiv
FORCE: Transferable Visual Jailbreaking Attacks via Feature Over-Reliance CorrEction2509.21029v3 · Runqi Lin, Alasdair Paren, Suqin Yuan et al.2025 · 0 citationsarXiv
Rethinking Safety in LLM Fine-tuning: An Optimization Perspective2508.12531v1 · Minseon Kim, Jin Myung Kwak, Lama Alssum et al.2025 · 0 citationsarXiv
MIP against Agent: Malicious Image Patches Hijacking Multimodal OS Agents2503.10809v3 · Lukas Aichberger, Alasdair Paren, Guohao Li et al.2025 · 0 citationsarXiv
Mixture of Experts Made Intrinsically Interpretable2503.07639v1 · Xingyi Yang, Constantin Venhoff, Ashkan Khakzar et al.2025 · 0 citationsarXiv
Shh, don't say that! Domain Certification in LLMs2502.19320v2 · Cornelius Emde, Alasdair Paren, Preetham Arvind et al.2025 · 0 citationsInternational Conference on Learning Representations (ICLR) 2025
On the Coexistence and Ensembling of Watermarks2501.17356v1 · Aleksandar Petrov, Shruti Agarwal, Philip H.S. Torr et al.2025 · 1 citationarXiv
Open Problems in Machine Unlearning for AI Safety2501.04952v1 · Fazl Barez, Tingchen Fu, Ameya Prabhu et al.2025 · 0 citationsarXiv
Detecting LLM Hallucination Through Layer-wise Information Deficiency: Analysis of Ambiguous Prompts and Unanswerable Questions2412.10246v2 · Hazel Kim, Tom A. Lamb, Adel Bibi et al.2024 · 6 citationsarXiv
Bi-Factorial Preference Optimization: Balancing Safety-Helpfulness in Language Models2408.15313v2 · Wenxuan Zhang, Philip H.S. Torr, Mohamed Elhoseiny et al.2024 · 1 citationarXiv
FedMedICL: Towards Holistic Evaluation of Distribution Shifts in Federated Medical Imaging2407.08822v1 · Kumail Alhamoud, Yasir Ghunaim, Motasem Alfarra et al.2024 · 9 citationsarXiv
Model Merging and Safety Alignment: One Bad Model Spoils the Bunch2406.14563v1 · Hasan Abed Al Kader Hammoud, Umberto Michieli, Fabio Pizzati et al.2024 · 8 citationsarXiv
Do as I do (Safely): Mitigating Task-Specific Fine-tuning Risks in Large Language Models2406.10288v3 · Francisco Eiras, Aleksandar Petrov, Philip H.S. Torr et al.2024 · 0 citationsarXiv
Towards Interpretable Deep Local Learning with Successive Gradient Reconciliation2406.05222v1 · Yibo Yang, Xiaojie Li, Motasem Alfarra et al.2024 · 1 citationarXiv
Universal In-Context Approximation By Prompting Fully Recurrent Models2406.01424v2 · Aleksandar Petrov, Tom A. Lamb, Alasdair Paren et al.2024 · 0 citationsarXiv
Towards Certification of Uncertainty Calibration under Adversarial Attacks2405.13922v3 · Cornelius Emde, Francesco Pinto, Thomas Lukasiewicz et al.2024 · 1 citationInternational Conference on Learning Representations (ICLR) 2025
Risks and Opportunities of Open-Source Generative AI2405.08597v3 · Francisco Eiras, Aleksandar Petrov, Bertie Vidgen et al.2024 · 7 citationsarXiv
Near to Mid-term Risks and Opportunities of Open-Source Generative AI2404.17047v2 · Francisco Eiras, Aleksandar Petrov, Bertie Vidgen et al.2024 · 2 citationsarXiv
Continual Learning on a Diet: Learning from Sparsely Labeled Streams Under Constrained Computation2404.12766v2 · Wenxuan Zhang, Youssef Mohamed, Bernard Ghanem et al.2024 · 1 citationarXiv
No "Zero-Shot" Without Exponential Data: Pretraining Concept Frequency Determines Multimodal Model Performance2404.04125v3 · Vishaal Udandarao, Ameya Prabhu, Adhiraj Ghosh et al.2024 · 15 citationsarXiv
On Pretraining Data Diversity for Self-Supervised Learning2403.13808v3 · Hasan Abed Al Kader Hammoud, Tuhin Das, Fabio Pizzati et al.2024 · 8 citationsarXiv
Efficient Lifelong Model Evaluation in an Era of Rapid Progress2402.19472v2 · Ameya Prabhu, Vishaal Udandarao, Philip Torr et al.2024 · 0 citationsarXiv
Prompting a Pretrained Transformer Can Be a Universal Approximator2402.14753v1 · Aleksandar Petrov, Philip H.S. Torr, Adel Bibi2024 · 1 citationarXiv
Can Large Language Model Agents Simulate Human Trust Behavior?2402.04559v4 · Chengxing Xie, Canyu Chen, Feiran Jia et al.2024 · 30 citationsarXiv
SynthCLIP: Are We Ready for a Fully Synthetic CLIP Training?2402.01832v2 · Hasan Abed Al Kader Hammoud, Hani Itani, Fabio Pizzati et al.2024 · 4 citationsarXiv
Label Delay in Online Continual Learning2312.00923v2 · Botos Csaba, Wenxuan Zhang, Matthias Muller et al.2023 · 5 citationsarXiv
From Categories to Classifiers: Name-Only Continual Learning by Exploring the Web2311.11293v2 · Ameya Prabhu, Hasan Abed Al Kader Hammoud, Ser-Nam Lim et al.2023 · 0 citationsarXiv
When Do Prompting and Prefix-Tuning Work? A Theory of Capabilities and Limitations2310.19698v2 · Aleksandar Petrov, Philip H.S. Torr, Adel Bibi2023 · 6 citationsarXiv
Segment, Select, Correct: A Framework for Weakly-Supervised Referring Segmentation2310.13479v3 · Francisco Eiras, Kemal Oksuz, Adel Bibi et al.2023 · 2 citationsarXiv
Efficient Error Certification for Physics-Informed Neural Networks2305.10157v2 · Francisco Eiras, Adel Bibi, Rudy Bunel et al.2023 · 0 citationsarXiv
Language Model Tokenizers Introduce Unfairness Between Languages2305.15425v2 · Aleksandar Petrov, Emanuele La Malfa, Philip H.S. Torr et al.2023 · 55 citationsarXiv
Career total: 106 works. 75 are in this corpus.Showing the 50 most recent.
Profile built from the corpus for this byline.
Author records are still filling in while the Hub is in alpha. If this is your page, you'll be able to claim it soon. Spot a mistake? Tell us.