Kristen Grauman
University of Texas at Austin
Leads work on egocentric video understanding and retrieval at scale, with a focus on long-horizon grounding and efficient perception.Grounded Retrieval and Agentic Intelligence for Vision-Language plan > retrieve > reason > verify
A CVPR 2026 workshop for researchers and practitioners building grounded multimodal retrieval, reranking, and verification systems that can be deployed with confidence.
Jun 3
7:30 AM-12:30 PM · Room 506 · Denver
4 invited talks
Dan Roth · Kristen Grauman · Scott Wen-tau Yih · Mohit Bansal
10:00 AM
Coffee Break & Poster Session
1 industry panel
11:15 AM-12:15 PM · Moderated by Sujith Ravi
Workshop starts
Jun 3, 2026 · 7:30 AM
P-R-R-V Loop
Plan tasks and tool use, retrieve candidates, rerank with grounded reasoning, then verify with calibrated evidence.
GRAIL-V
CVPR 2026
Confirmed program
June 3 Agenda
07:30 AM - 07:45 AM
Welcome & Opening Remarks
Organizing Committee
07:45 AM - 08:15 AM
08:15 AM - 08:45 AM
Kristen Grauman
University of Texas at Austin
Invited Talk 2
Kristen Grauman
University of Texas at Austin
Grounding Temporal Reasoning in Video Evidence
08:45 AM - 09:00 AM
Coffee Break
Coffee available during this short break.
09:00 AM - 09:30 AM
Scott Wen-Tau Yih
Meta
Invited Talk 3
Scott Wen-Tau Yih
Meta
MetaCLIP: Open, Scalable Data Curation for Vision-Language Models
09:30 AM - 10:00 AM
Mohit Bansal
University of North Carolina at Chapel Hill
Invited Talk 4
Mohit Bansal
University of North Carolina at Chapel Hill
Long-Horizon Video Reasoning and Generation
10:00 AM - 11:00 AM
Coffee Break & Poster Session
Coffee available during this window.
Posters will be in Exhibit Hall A, boards 188-207.
11:00 AM - 11:15 AM
Paper Presentation
11:15 AM - 12:15 PM
Sujith Ravi
GVP, Oracle AI
Vijay Krishnan
Founder and CTO, Turing
Scott Wen-Tau Yih
Meta
Kenneth Marino
University of Utah / Ex-DeepMind
Ming-Hsuan Yang
UC Merced / DeepMind
Industry Panel
Sujith Ravi
GVP, Oracle AI
Vijay Krishnan
Founder and CTO, Turing
Scott Wen-Tau Yih
Meta
Kenneth Marino
University of Utah / Ex-DeepMind
Ming-Hsuan Yang
UC Merced / DeepMind
From Understanding to Action: Building the next AI frontier with Multimodal Agents, World Models, and Real-world Intelligence
Moderated by Sujith Ravi with panelists Vijay Krishnan, Scott Wen-Tau Yih, Kenneth Marino, and Ming-Hsuan Yang.
12:15 PM - 12:30 PM
Thanks & Closing Remarks
Organizing Committee
Speakers
Mohit Bansal
University of North Carolina at Chapel Hill
Researches multimodal language/vision agents, grounded reasoning, and controllable generation for real-world tasks.
Dan Roth
University of Pennsylvania / Oracle AI
Pioneer in structured, grounded reasoning and robust inference for language/vision systems deployed in real settings.
Scott Wen-Tau Yih
Meta
Research Scientist at Meta FAIR and affiliate professor at the University of Washington. His work spans NLP, ML, and information retrieval, including DPR and RAG, and he was named an ACL Fellow in 2024.Focus areas at a glance
We’re looking for work that ties mutlimodal agentic tools,evidence to decisions, scales retrieval and reranking, and evaluates real deployment constraints.
Primary Track
Vision-language models
Grounded reasoning
Multimodal agents
Agentic memory
Secondary Track
Detection & retrieval
Video understanding
Document & chart analysis
Foundation models
Visual Grounding & Evidence
Region grounding
Temporal grounding
Evidence overlays
Citation provenance
Retrieval & Ranking
Hybrid search
Dense/sparse retrieval
Structured & unstructured sources
Rerankers
Long-context retrieval
Agentic Tools & Planning
Tool routing
Structured queries
Layout parsing
Guardrails
Evaluation & Efficiency
Benchmarks
Reproducibility
Latency & cost
Energy & memory
Why now
Grounded evidence is the missing piece for agentic vision
Vision-language agents increasingly rely on a loop of plan, retrieve, rerank, and verify before acting. But how we measure evidence grounding, calibration, and end-to-end efficiency is still fragmented across communities.
GRAIL-V brings together CV, IR, NLP, HCI, and systems researchers and practitioners working on evidence-centric retrieval and verification for deployable agentic systems.
Why submit or attend
- Get feedback on grounded retrieval, reranking, and verification.
- Share real-world evaluations, demos, and deployment lessons.
- Connect with researchers building agentic VLM pipelines.
- Help shape community benchmarks and best practices.
Important dates
Timeline (Anywhere on Earth)
CVPR 2026 workshop submission deadline
Extended deadline. OpenReview submission closes at 23:59 AoE.
Mar 18, 2026
Notification to authors
Decisions released via OpenReview.
Apr 5, 2026
Camera-ready due
Final versions for CVPR workshop proceedings.
Jun 3, 2026 · 7:30 AM
Workshop day in Room 506, Denver
Confirmed half-day program with invited talks, paper presentations, posters, and an industry panel.
Updates
Latest announcements
Jun 3, 2026
GRAIL-V Best Paper and Outstanding Paper announced
Congratulations to SHOE - Semantic HOI Open-vocabulary Evaluation metric, winner of the $2,000 Best Paper award, and HTEF: Holistic Brand-Theme Alignment Scoring as a Catalog Gate for Grounded Conversational Recommendation, winner of the $1,500 Outstanding Paper award.
Jun 1, 2026
Poster session location and boards posted
The poster session will run on June 3, 2026 from 10:00 AM to 11:00 AM local time in Exhibit Hall A, boards 188-207.
In-person with hybrid support
Location
Room 506, Denver, USA. Confirmed workshop start: Jun 3, 2026 at 7:30 AM local time.
Hybrid plan
Hybrid participation follows CVPR guidance; details will be posted when available.
Registration
Registration is handled via CVPR. We will link the official registration page when it opens.
Accessibility
Reserved front-row seating, mic runners, captioning support.
Stay connected
General inquiries
Reach out to the organizers with questions about submissions, sponsorship, or program.