6th IAPR TC10/TC11 Summer School

Next-Gen Document Understanding: RAG, VLMs, and Structured Knowledge Extraction

May 25-29, 2026
Vall de Núria, Catalonia
50 Participants

About the Summer School

Overview

The rapid development of Document Intelligence (DI) has transformed traditional Document Analysis and Recognition (DAR) into a sophisticated, AI-driven field. This summer school provides cutting-edge tools for information extraction using Retrieval-Augmented Generation (RAG), Vision-Language Models (VLMs), and structured knowledge representation.

Through a Challenge-Based Learning framework, participants will progress from foundational lectures to hands-on practice sessions and culminate in solving real-world industry challenges in collaborative teams.

Learning Outcomes

Design and implement DocVQA systems with RAG techniques
Master multimodal data generation for efficient VLM training
Build knowledge graphs from complex document collections
Apply solutions to low-resource and historical document analysis

Scientific Program

Lectures

Vision-Language Models

Designing and training VLMs

Lessons learned from training large-scale vision-language models.

Document Processing with VLMs

How to use VLMs to extract structured information from documents.

Graphs as a Document Representation

Learning on Graphs: GNNs in Document Analysis.

Use graph neural networks to analyze and interpret document structures.

Knowledge Graph Embeddings and Knowledge Representation

Convert documents into structured knowledge graphs for enhanced understanding.

GraphRAG and Agentic Architectures: From data to insight

How to augment your models with graph retrieval-augmented generation techniques.

Trends on Document Analysis

Trends on Trustworthy Document Analysis

Panel with experts working on trustworthiness and explainability in document analysis.

Trends on Historical Document Analysis

Panels with experts discussing the latest research and challenges in historical document analysis.


Hands-On Lab Sessions

Fine-Tuning Vision Language Models

Fine-tune a small VLM on documents of a specific domain.

Graph Neural Networks

Explore the use of graph neural networks for document structure analysis.

Agentic GraphRAG

Build an agentic system using GraphRAG for advanced document analysis.

Document Understanding 2026 Challenge

Teams of 5 participants will tackle real-world industry problems.

Call for Participants

Target Audience

PhD Students (Priority)

Students in their 1st-3rd year working on document analysis, information retrieval, or related fields

Master's/Post-Master's Students

Advanced students preparing for PhD studies in document analysis

Early-Career Practitioners

Industry professionals seeking to deepen expertise in document understanding

Desired Skills

Deep Learning Frameworks

Experience with PyTorch or TensorFlow

Research Skills

Analytical approach to research and industry challenges

Team Collaboration

Ability to work effectively in team-based activities

Motivation

Clear interest in document analysis and AI

5 Full Grants Available!

We offer 5 full registration fee grants. Apply during registration!

Evaluation criteria:

  • Geographic distance
  • Financial constraints
  • Academic background and CV
  • Research relevance

Location & Venue

Vall de Núria

A stunning glacial valley nestled in the heart of the Pyrenees mountains of Catalonia, at approximately 2,000 meters elevation. Accessible only by scenic cogwheel train, offering a peaceful and inspiring environment away from everyday distractions.

Location

Eastern Pyrenees, Queralbs, Catalonia, Spain

Starting Point

Day 1: Computer Vision Center (CVC), UAB Campus

Transport provided to Vall de Núria

Accommodation

All participants stay at the hotel in the center of the valley. Shared rooms (double/triple) with private bathrooms. All meals included: buffet-style breakfast, lunch, and dinner with vegetarian/vegan options.

Vall de Núria, Catalonia

2,000m elevation • Car-free valley

Social Activities

  • Guided hiking and nature walks with mountain views
  • Board games and social activities in communal spaces
  • Gala dinner on the final evening
  • Informal networking throughout the week

Registration

850€

Includes accommodation, all meals, coffee breaks, transport, and materials

Important Information

Capacity: Limited to 50 participants • Format: In-person only • Language: English


What's Included

  • ✓ 4 nights accommodation at Vall de Núria
  • ✓ All meals (breakfast, lunch, dinner)
  • ✓ Daily coffee breaks
  • ✓ Transport from/to Barcelona
  • ✓ Conference materials and goodie bag
  • ✓ Access to all lectures and lab sessions
  • ✓ Recorded lectures (available post-event)

Grant Opportunities

5 full registration grants available thanks to IAPR support. Apply during registration!

Evaluation criteria:

  • Geographic distance
  • Financial constraints
  • Academic background and CV
  • Research relevance

Questions? Contact: {amolina, allabres}@cvc.uab.es

Sponsors & Partners

Platinum Sponsor

Organizer

Supporter

Co-Organizer

Co-Organizer

Become a Sponsor

Support the next generation of document understanding researchers and gain visibility in the academic community.

Organizing Team

Prof. Dimosthenis Karatzas

Prof. Dimosthenis Karatzas

Associate Director, CVC

Head of Vision, Language and Reading Group

Prof. Josep Lladós

Prof. Josep Lladós

Director, CVC

Head of Document Analysis Group

Dr. Ernest Valveny

Dr. Ernest Valveny

Researcher, CVC

And Coordinator of the AI Degree at UAB

Adri Molina

Adri Molina

PhD Student, CVC

Document Analysis • Retrieval Systems

Artemis Llabrés

Artemis Llabrés

PhD Student, CVC

Vision & Language • Document Understanding

Nuria Martinez
Aurora Garcia
Xavier Galvez

CVC Communications Team

Nuria Martinez • Aurora Garcia • Xavier Galvez