PulseAugur
EN
LIVE 08:25:10

New VELD-QA benchmark targets visual evidence in long documents

By PulseAugur Editorial · [1 sources] ·

Researchers have introduced VELD-QA, a new large-scale benchmark designed to evaluate visual evidence grounding and question answering capabilities within lengthy documents. This benchmark comprises 15,000 human-authored question-answer pairs derived from 5,859 real-world PDFs, totaling over 370,000 pages. VELD-QA includes detailed annotations for supporting pages and visual elements, allowing for separate assessment of retrieval and interpretation skills. Evaluations of nine vision-language models and seven retrieval-augmented systems revealed significant challenges in evidence localization and interpretation, indicating that even with full document access, performance remains limited. AI

IMPACT This benchmark aims to improve multimodal document QA systems by identifying bottlenecks in visual interpretation and evidence localization.

RANK_REASON The item describes a new academic dataset and benchmark for research purposes. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

New VELD-QA benchmark targets visual evidence in long documents

How we ranked this

Signal score
17 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
The item describes a new academic dataset and benchmark for research purposes. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
paper, other
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. arXiv cs.CL TIER_1 English(EN) · Zhivar Sourati, Marianne Menglin Liu, Rongguang Wang, Zheng Wang, Yazhe Hu, Lan Guo, Yuying Wang, Mengqing Guo, Bhavesh Dnyaneshwar Gore, Shubham Pathak, Ramapriya Ranganath, Sandeep Sharma, Tao Sheng, Sujith Ravi, Dan Roth ·

    VELD-QA: A Large-Scale Dataset for Visual Evidence Grounding and Question Answering in Long Documents

    arXiv:2610.20964v1 Announce Type: cross Abstract: Existing long-document QA benchmarks rarely combine large-scale human-authored questions with detailed visual evidence annotations, limiting diagnosis of evidence localization and interpretation failures. We present VELD-QA, a lar…

RELATED ENTITIES

RELATED TOPICS