Toshiaki Koike-Akino

Research focus

Red-Teaming & Adversarial Robustness (2)Multimodal Models (1)Reasoning & Chain-of-Thought (1)RLHF & Preference Learning (1)

Frequent co-authors

Ye Wang (2)Jing Liu (2)Vanshaj Khattar (1)Md Rafi ur Rashid (1)

Papers (2)

Mar 16, 2026

Ye Wang +21d ago

Directional Embedding Smoothing for Robust Vision Language Models

Aligning noise with token embeddings makes vision-language models more resilient to jailbreaking attacks, slashing success rates on the JailBreakV-28K benchmark.

Ye Wang, Jing Liu, Toshiaki Koike-Akino

Multimodal Models Red-Teaming & Adversarial Robustness

Mitsubishi Electric Research1d ago

Amplification Effects in Test-Time Reinforcement Learning: Safety and Reasoning Vulnerabilities

Test-time RL, intended to boost LLM reasoning, can backfire by amplifying harmful tendencies and degrading reasoning when exposed to adversarial prompts.

Vanshaj Khattar, Md Rafi ur Rashid, Moumita Choudhury +4

Reasoning & Chain-of-Thought Red-Teaming & Adversarial Robustness RLHF & Preference Learning

Search

Toshiaki Koike-Akino

Publication activitypapers/week, last 8 weeks

Research focus

Frequent co-authors

Papers (2)