Lattice AI Research

Research focus

Red-Teaming & Adversarial Robustness (2)Constitutional AI & AI Ethics (1)Eval Frameworks & Benchmarks (1)Multimodal Models (1)

Frequent co-authors

Or Biton (1)Tomer Krichli (1)Joseph Keshet (1)

Papers (2)

Aug 24, 2026

4d ago·also Affiliation:

Hidden in the Request: Explaining Unethical LLM Compliance through Token Relevance

LLMs are more likely to comply with unethical requests when benign framing tokens overshadow cue-tokens, revealing a hidden vulnerability in their alignment.

Or Biton, Tomer Krichli, Itai Allouche +1

Constitutional AI & AI Ethics Red-Teaming & Adversarial Robustness

May 3, 2026

Mitigating Multimodal LLMs Hallucinations via Relevance Propagation at Inference Time

MLLMs hallucinate less when you nudge them to pay more attention to non-text inputs during inference, without any training.

Itai Allouche

Eval Frameworks & Benchmarks Multimodal Models Red-Teaming & Adversarial Robustness

Search

Itai Allouche

Publication activitypapers/week, last 8 weeks

Research focus

Frequent co-authors

Papers (2)