====================================================================================
Call for Papers — CVPR 2026 Workshop
GRAIL-V: Grounded Retrieval & Agentic Intelligence for Vision-Language
https://grailworkshops.github.io/ | https://openreview.net/group?id=thecvf.com/CVPR/2026/Workshop/GRAIL-V

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026)
June 3–7, 2026 • Denver Convention Center, Denver, USA
Workshops/Tutorials: June 3–4, 2026 | Main Conference: June 5–7, 2026
====================================================================================


WORKSHOP DESCRIPTION
-------------------------
Computer vision is moving beyond passive perception. The next frontier is agentic intelligence:
systems that don’t just “see” data, but actively plan, retrieve, reason, verify, and act to solve
complex, real-world workflows.

GRAIL-V is a dedicated venue for bridging research capability and deployment reliability. We
bring together the CV, IR, NLP, HCI, and systems communities to tackle the hard parts of
multimodal agents: grounded retrieval, evidence-backed reasoning, calibrated verification,
and real-world execution under latency/cost constraints.

(Workshop loop: plan → retrieve → reason/rerank → verify)


KEYNOTE SPEAKERS
-------------------------
- Kristen Grauman (UT Austin)
- Mohit Bansal (UNC Chapel Hill)
- Dan Roth (University of Pennsylvania / Oracle AI)
- Scott Wen-Tau Yih (Meta)
- Sujith Ravi (Oracle)
+ Panel: “From Retrieval to Action: What Should Agentic Vision Systems Verify?”


TOPICS
----------
We welcome archival papers and demos on grounded, efficient, and safe agentic vision-language
systems, including (but not limited to):

- Multimodal Retrieval:
* Unified search across images, video, charts, documents, and UI
* Hybrid dense+sparse retrieval, multilingual retrieval, long-context retrieval
* Fine-grained region/page/moment retrieval with evidence

- Reranking & Grounded Selection:
* Cross-encoders / late-interaction rerankers for grounded ranking
* Evidence-aware reranking with citations/provenance
* Calibration and abstention behavior at top-k for agent handoffs

- Agentic Planning & Tool Use:
* Tool routing, query reformulation, multi-step workflows, agent memory/state
* Document/layout/chart tools, structured querying, safe tool-use and recovery

- Grounding, Evidence, and Reliability:
* Citation provenance, evidence overlays, audit-ready faithfulness
* Robustness to UI drift and format changes; verification protocols

- Benchmarks & Evaluation:
* Reproducible evaluation harnesses, traces, leaderboards, and reporting practices
* Efficiency metrics: latency, memory, energy, and cost
* Production constraints and deployment learnings


IMPORTANT DATES (Anywhere on Earth)
------------------------
- Paper Submission Deadline: March 5, 2026 (23:59 AoE)
- Notification to Authors: March 18, 2026
- Camera-Ready Deadline: April 5, 2026
- Workshop Days: June 3–4, 2026 (Denver, USA)


SUBMISSION AND DECISIONS
------------------------
- Format: CVPR template, double-blind, up to 8 pages (references excluded)
- Submission Portal: OpenReview (link above)
- Review criteria: technical merit, grounded evaluation, efficiency, reproducibility, and
broader safety considerations
- Presentation: accepted papers will be presented as posters and/or short orals; at least one
author must present in person (per CVPR policy)

CONFERENCE PROCEEDINGS
-----------------------------------
GRAIL-V is archival. Accepted papers will appear in the CVPR 2026 Workshop Proceedings.


ORGANIZERS
-------------
Amit Agarwal (Oracle AI)
Vivek Gupta (Arizona State University)
Vivek Srikumar (University of Utah)
Tao Sheng (Oracle AI)
Alice Oh (KAIST)
Sara Hooker (Adaption Labs)
Jyotika Singh (Oracle AI)
Hitesh Laxmichand Patel (Oracle AI)

CONTACT
-------------
General inquiries / sponsorship / program: grailworkshops@googlegroups.com
====================================================================================