Archival paper
Up to 8 pages (references excluded), CVPR template, double-blind review, appears in workshop proceedings.
Computer Vision/CORE Unranked
CVPR 2026
CVPR 2026 Workshop on Grounded Retrieval and Agentic Intelligence for Vision-Language
GRAIL-V at CVPR 2026 is a workshop focused on grounded retrieval and agentic intelligence for vision-language systems, emphasizing evidence-centric approaches that integrate planning, retrieval, reasoning, and verification. It brings together researchers from computer vision, information retrieval, NLP, and HCI to advance deployable multimodal agents with calibrated, efficient, and auditable decision-making.
Key deadlines
Verified September 9, 2026Full paper
March 9, 2026
AoE
Workshop timeline
Submission and decisions
Full paperKey deadline
March 9, 2026 · AoE
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Archival short paper
Early or in-progress ideas that will appear in the CVPR workshop proceedings.
Demo / system paper
Working systems with grounded evidence, evaluation, and reproducible artifacts.
Benchmark / dataset
New tasks, datasets, or evaluation harnesses for grounded retrieval and verification.
Research areas in scope
01
Primary Track
Vision-language modelsGrounded reasoningMultimodal agentsAgentic memory
02
Secondary Track
Multimodal learningRecognition (categorization, detection, retrieval)Vision applications and systemsDocument analysis and understandingFoundation models
03
Visual Grounding & Evidence
Region groundingTemporal groundingEvidence overlaysCitation provenance
04
Retrieval & Ranking
Multimodal indexingSearchRerankingHybrid searchDense/sparse retrievalStructured & unstructured sourcesLong-context retrieval
05
Agentic Tools & Planning
Tool routingStructured queriesLayout parsingGuardrails
06
Evaluation & Efficiency
BenchmarksReproducibilityLatency & costEnergy & memory
Policies worth checking twice
- Submissions must follow CVPR 2026 author guidelines.
- Double-blind review process with 3 reviews and AC oversight.
- Maximum of 8 pages (references excluded) for archival papers.
- Accepted papers must have at least one author present in person.
- Conflict of interest includes same-employer, advisor-advisee, recent co-author, or close personal relationships.
- Code, data, and models are strongly encouraged with licenses and provenance.
- Non-archival and demo-only submissions are accepted.
- Evaluation criteria include technical merit, grounded evaluation, efficiency, reproducibility, and broader safety considerations.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026