AI Research & Safety
AI Research
NAVER conducts fundamental research that enables us to deliver new user experiences and address real-world challenges. From HyperCLOVA X, NAVER’s generative AI, to AI safety that’s gained traction in recent years to AI’s advanced capabilities that are sure to open the door to new possibilities, our research teams focus their work across many areas.
Research areas
- 01
Machine Learning
Our research in machine learning involves the development of ML models that support different modalities, algorithms (including optimization), and AI safety.
- Multimodal learning algorithms
- ML reliability
- Generic ML algorithms
- 02
Backbone Models
We take a broad approach to research in deep learning and fundamental AI.
- New foundation models and elements
- New training methods
- Cost-efficiency improvements
- Data-related topics
- 03
Language Models
The research goal for language models is to create safe and trustworthy models aligned with the values of humanity and society while driving AI innovation. Our research team works in academia and industry to tackle foundational and real-world challenges.
- Building safe and trustworthy large language models (LLMs)
- Datasets, benchmarks, and evaluation metrics for LLMs
- LLM security involving attacks, defenses, and detection
- Safety alignment, learning, and inference algorithms
- LLM agents, single or multi-agent interaction, decision-making, and autonomous agents
- 04
Content Generation
We aim to push the boundaries of visual generation and develop models that produce high-quality content. We hope to spark creativity in our users and make AI available in every part of the world.
- Multimodal generation models
- Neural 3D representation learning and its applications
- Generative applications tailored for LLMs
- 05
Human-Computer Interaction
Our HCI research focuses on understanding ways to design AI technologies that will benefit end users when embedded into computing systems.
- AI-infused interactive systems
- Digital health and well-being applications
- Accessibility and safety of AI
- LLM-driven computing systems and empathetic agents
Our research
AI Safety
NAVER runs the Future AI Center, which is dedicated to research on AI safety. We collaborate with researchers and institutions worldwide to pursue responsible AI and shape trustworthy policies around the technology.
We work to enhance the capabilities of AI models with quality datasets grounded on safety research. We also take various technological and policy actions to advance safe and responsible AI development.
AI ethics principles and practices
NAVER ASF(AI Safety Framework) Beta
Here, we introduce NAVER’s AI Safety Framework, where we identify, assess, and manage the potential risks of AI that teams across NAVER might encounter when developing and deploying AI systems.
01
Recognize the potential risks to advanced AI capabilities and implement risk assessment and management processes.
02
Foster collaboration between cross-functional teams across the lifecycle of AI systems.
03
Establish management decision-making processes and work with external stakeholders to address AI safety challenges.
NAVER’s AI Safety Framework (ASF)
Ensuring AI safety through research
AI Safety Initiatives
We build datasets and benchmarks and make them available to everyone for safe AI research and development. Our efforts are focused on creating a healthy AI ecosystem.
AI safety datasets and
benchmarks
Build AI datasets and benchmarks for safe and trustworthy AI R&D.
AI red-teaming
Trigger LLMs to generate harmful content to expose their vulnerabilities.
AI security (attacks, defenses, and detection)
To enhance the security of AI systems, take a comprehensive approach to attack and defense strategies and detection technologies.
Learning and inference toward AI alignment
Ensure AI models are aligned with human values during learning and inference.
