Applicant qualification

• Applicant must be enrolled in a MS or PhD full-time course student for the entire 2026 – 2027 academic year in their university.
• Applicant must be the 1st author of applied paper to the program.
  • If applied paper has co-1st authors, all co-1st authors must satisfy full-time student condition.
• Applicant can submit only one paper.
  • The submit number counting is based on the leftmost name at author list of paper.
• Current and former employees and consultants of Sponsor (Qualcomm), Sponsor’s respective parent companies, affiliates and subsidiaries will not be accepted as Applicant and therefore will not be permitted to participate the Program.

Paper qualification

• Paper language : English only
• Published paper only
  • Publish date must be from 16th Aug, 2025.
  • Definition of ‘published’ :
    1. The paper is opened to web so anyone can search the paper via web-search such as arXiv.
    2. Applicant agrees to open the contents and idea of the paper to anyone.

• When paper applied to the program, author list must be hidden for blind review.
• Sponsor (Qualcomm)’s owned paper or papers supported by Sponsor, or its affiliates won’t be accepted for the Program.

Research Area

• Basic domain: On-Device AI / Autonomous Driving Physical AI / Agentic AI
• Applicant should select an area of interest for the submitted paper.
• Area of Interest:
  1. On-Device AI
     1. On-device Training
        1. Resource-efficient fine-tuning
        2. Parameter-efficient fine-tuning (PEFT)
        3. Low precision training
        4. Fast training technology
        5. Adapter design
        6. On-device adaptation
        7. On-device personalization
        8. Personalization use case & technology for LLM, LVM, and LMM
        9. Test-time adaptation
        10. Knowledge distillation
        11. Federated learning
     2. Model compression
        1. Quantization
        2. Pruning and Sparsity
     3. Efficient Inference
        1. Acceleration of inference
        2. Efficient architecture
        3. Memory optimization
     4. Efficient Hardware
        1. Heterogeneous computing
        2. Edge computing
        3. SW/HW co-design
  2. Autonomous Driving & Physical AI
     1. End-to-End Solution
        1. Integration of Imitation Learning (IL) and Reinforcement Learning (RL)
        2. E2E Planning: Beyond IL and RL
        3. Handling corner cases and long-tail scenarios
        4. World Models and Foundation Models for E2E
        5. Data Efficient E2E Systems
        6. Evaluation of E2E Systems
     2. LLMs, VLMs, VLAs, Foundation Models
        1. Vision-Language-Action models
        2. Reasoning Models
        3. Agentic Driving Systems for Autonomous Driving
        4. Instruction-driven Multimodal Agents for Autonomous Driving
        5. The Alignment Problem
        6. VLAs at the Edge: Designing for Edge Deployment
     3. Simulation
        1. Generative AI and World Models for Autonomy
        2. High Throughput and High-Speed Sensor Simulation
        3. Real2Sim and Sim2Real at scale
        4. Large-Scale Simulation for Autonomous Driving
        5. Large-scale Natural Behavior Modeling
        6. GenAI and Gaussian Splatting
        7. Sim Based Evaluation of Autonomous Driving
     4. Perception, Planning, Multimodal Fusion for Autonomy
        1. Advanced Sensors and Sensor Fusion
        2. Imaging Radar
        3. Computer Vision for Autonomy
        4. Trajectory Planning/Prediction with Deep Learning
        5. Multimodal Sensor Fusion with Deep Learning
  3. Agentic AI
     1. Voice interface
        1. ASR, TTS, Voice conversion, Emotion recognition
        2. Full-duplex (conversation)
        3. Spoken Language Model (SLM)
        4. Spoken Language Understanding (SLU)
     2. Interaction
        1. Natural and Multimodal User Interfaces for Agent Interaction
        2. Multi-Agent Systems: Collaboration and Coordination
        3. Human-Agent Interaction and Collaboration
     3. Perception
        1. Multimodal Perception for Agent Awareness
     4. Memory & Personalization
        1. Memory Architectures for Long-Context Agents
        2. Efficient and Private On-Device Personalization for Agents
        3. User Modeling for Personalized Agent Interaction
        4. Personalized Agent Learning: Adapting to User Preferences and Styles
        5. Context-Aware and Proactive Personalization in Agentic Systems
        6. Privacy-Preserving Techniques for Personalized Agent Learning
        7. Continual and Lifelong Learning for Agents
        8. Data-Centric AI for Agent Training
     5. Reasoning and Action
        1. Goal-Oriented Reasoning and Planning
        2. Hierarchical Reinforcement Learning for Complex Tasks
        3. Safe and Reliable Agent Decision-Making
        4. Explainable AI (XAI) for Agent Behavior
        5. Agent Tool Use and Function Calling
        6. Self-Supervised Learning for Agent Skill Discovery
        7. World Models for Agent Planning and Imagination
        8. Foundation Models for Agent Control
     6. Embodied AI: Perception, Reasoning, and Action Integration
     7. Efficient Agent Architectures for On-Device Deployment

Please refer below ‘Program details & submission matters’ file in the ‘Submission matters’ section for detail.
 

Application submission

• Each applicant should submit an application by the specified deadline that must include :
  • Paper – hidden author list for blind review
  • Signed copy of QIFK rules
  • Complete and sign Foreign Corrupt Practices Act (FCPA) Questionnaire
    • If an application has more than one co-1st authors, each co-1st authors must complete and sign QIFK rule and FCPA questionnaire.
• Self evaluation section
  • Applicant assign self evaluation score and comment about your paper by two criteria – Innovation, Research potential.
• Applicant information by application page request (Name, e-mail address and the others)
• After submission, Qualcomm sends application number for managing notification

Paper review

• Qualcomm AI Research members review submitted papers with review criteria.
• After paper review phase for submitted paper, max 30 papers will be selected next phase, presentation and poster session.
• Review criteria
  • Innovative research idea
  • Clear understanding research area and realistic implementation
  • Potential of research for further research
  • Applicant strength

Final session

• Finalists will have presentation and poster session at QIFK 2026 final session day.
• Max 15 papers will be chosen final scholarship selection.

Scholarship reward

• Cash awards of 4,000,000 KRW per selected paper will be awarded to his or her affiliated university.
• Applicant should follow his or her university scholarship policy when receiving award from the university.

Submission matters

• Submit page
• Program details & submission matters

Need further information?

Please direct your questions to : [email protected]

Need more Qualcomm Korea event?

Please join the Snapdragon Insiders community in Korea for the latest news, unique experiences, and giveaways.

Follow us on : Instagram, YouTube, Website