i

iliashum's activity

• All
• Models
• Datasets
• Spaces
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

authored 19 papers 7 days ago

UnUnlearning: Unlearning is not sufficient for content regulation in advanced generative AI

Paper • 2407.00106 • Published Jun 27, 2024 • 6

Locking Machine Learning Models into Hardware

Paper • 2405.20990 • Published May 31, 2024

ImpNet: Imperceptible and blackbox-undetectable backdoors in compiled neural networks

Paper • 2210.00108 • Published Sep 30, 2022

Wide Attention Is The Way Forward For Transformers?

Paper • 2210.00640 • Published Oct 2, 2022 • 1

Buffer Overflow in Mixture of Experts

Paper • 2402.05526 • Published Feb 8, 2024 • 8

Revisiting Block-based Quantisation: What is Important for Sub-8-bit LLM Inference?

Paper • 2310.05079 • Published Oct 8, 2023

Measuring memorization in RLHF for code completion

Paper • 2406.11715 • Published Jun 17, 2024 • 7

A False Sense of Safety: Unsafe Information Leakage in 'Safe' AI Responses

Paper • 2407.02551 • Published Jul 2, 2024 • 9

Operationalizing Contextual Integrity in Privacy-Conscious Assistants

Paper • 2408.02373 • Published Aug 5, 2024 • 5

Measuring memorization through probabilistic discoverable extraction

Paper • 2410.19482 • Published Oct 25, 2024 • 4

Stealing User Prompts from Mixture of Experts

Paper • 2410.22884 • Published Oct 30, 2024 • 15

Trusted Machine Learning Models Unlock Private Inference for Problems Currently Infeasible with Cryptography

Paper • 2501.08970 • Published Jan 15 • 6

Cascading Adversarial Bias from Injection to Distillation in Language Models

Paper • 2505.24842 • Published May 30 • 6

Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities

Paper • 2507.06261 • Published Jul 7 • 63

Reasoning Introduces New Poisoning Attacks Yet Makes Them More Complicated

Paper • 2509.05739 • Published Sep 6 • 2

The Attacker Moves Second: Stronger Adaptive Attacks Bypass Defenses Against Llm Jailbreaks and Prompt Injections

Paper • 2510.09023 • Published 24 days ago • 8

SynthID-Image: Image watermarking at internet scale

Paper • 2510.09263 • Published 24 days ago • 1

Extracting alignment data in open models

Paper • 2510.18554 • Published 13 days ago • 7

Soft Instruction De-escalation Defense

Paper • 2510.21057 • Published 10 days ago • 3

upvoted a paper 7 days ago

Soft Instruction De-escalation Defense

Paper • 2510.21057 • Published 10 days ago • 3