AI Safety Research | Award-Winning Papers | Virtue AI

Research Insights

From award-winning academic research to production-grade AI safety.

Recognized by

Outstanding Paper | NeurIPS

DecodingTrust: A Comprehensive Assessment of Trustworthiness in GPT Models

Best Paper | NeurIPS

When Do Universal Image Jailbreaks Transfer Between Vision-Language Models?

Best Research Paper Finalist | PVLDB

LLM-PBE: Assessing Data Privacy in Large Language Models

Research Papers

arXiv

SmallToLarge (S2L): Scalable Data Selection for Fine-tuning Large Language Models by Summarizing Training Trajectories of Small Models

IEEE SatML

Shake to Leak: Amplifying the Generative Privacy Risk through Fine-tuning

ICML

SHINE: Shielding Backdoors in Deep Reinforcement Learning

ICML

Rob-FCP: Certifiably Byzantine-Robust Federated Conformal Prediction.

ICLR

Ring-A-Bell! How Reliable are Concept Removal Methods For Diffusion Models?

arXiv

RigorLLM: Resilient Guardrails for Large Language Models against Undesired Content.

CVPR

PerAda: Parameter-Efficient Federated Learning Personalization with Generalization Guarantees

CVPR

MMSum: A Dataset for Multimodal Summarization and Thumbnail Generation of Videos

ICML

InstructRetro: Instruction Tuning post Retrieval-Augmented Pretraining

IEEE SatML

Improving Privacy-Preserving Vertical Federated Learning by Efficient Communication with ADMM.

Blogs

See all Blogs ->

Announcement

We’re Not Just Hosting a Conference, We’re Closing the Gap Between AI Research and AI Risk

There's a conversation that needs to happen in AI security. Not a panel discussion. Not a keynote.

Announcement

Introducing PolicyGuard: Define, Edit, and Enforce Custom Guardrails for Every Enterprise Model, Agent, and Application

PolicyGuard gives you complete control over how AI policy is defined and enforced across your environment at runtime, without retraining your models.

Announcement

Virtue Presents: CTRL+AI A Conference on Trust, Risk, Leadership and AI

CTRL+AI is an exclusive one-day conference bringing together the best and brightest minds from leading enterprises and AI labs: the researchers, executives, and operators who are building the reality of secure, scalable AI.

Secure your autonomous enterprise with the team that defines the standard.

Don't let safety be the friction that grounds your innovation. Virtue AI is the launchpad for agentic systems, clearing the path from research to production-ready performance.

Thank you! Your submission has been received!

Oops! Something went wrong while submitting the form.