This work presents Skill-Based Agentic DocMiner, a unified agentic workflow with three recurring components: multimodal context enrichment from captions and paper text, iterative skill refinement using development-set feedback, and collaborative verification by specialist reviewers.
Abstract
We present Skill-Based Agentic DocMiner, our submission to Sci-ImageMiner 2026 for Task 1 (figure classification), Task 2 (data table extraction), and Task 4 (visual question answering). Our system follows a unified agentic workflow with three recurring components: multimodal context enrichment from captions and paper text, iterative skill refinement using development-set feedback, and collaborative verification by specialist reviewers. For classification, we use class prototype documents to distinguish visually similar figure types. For table extraction, we combine chart parsing with Markdown structure validation and content checking. For VQA, we route each question to a specialist agent aligned with the official reasoning categories. On the official leaderboard, our submission ranked fourth on classification, fifth on table extraction, and third on VQA. The framework is simple, modular, and well matched to structured scientific reasoning. Our code is publicly available at https://github.com/1gst/Sci-ImageMiner.
A cross-task context-injection pipeline that prepends the structured outputs of earlier tasks to downstream prompts at inference time, grounding summarizations and VQA answers in explicitly extracted values rather than visual estimation, without additional training is proposed.
Tirtha Vinchurkar, Rithika Sai Konakalla, Kareem Abdelmaqsoud· Open Conference Proceedings· 0 citations
Scientific figures encode critical experimental findings that are often not fully captured in textual descriptions. Despite recent advances in MLLMs, their performance on domain-specific scientific figures remains limited due to insufficient grounding in relevant context. In this work, we focus on in Task 3 (summarizat...
Xiaole Lv, Wei Wu, Yang Tan et al.· Open Conference Proceedings· 0 citations
In this work, we address two core challenges in the ALD-E domain: auxiliary evidence construction via summarization and data table extraction, and visual question answering (VQA). Our approach first generates auxiliary summaries and structured tables to support test-time VQA prompting, using Qwen3.5-9B vision-language...
Vassilis Sioros· Open Conference Proceedings· 0 citations
This paper presents a solution to the Sci-ImageMiner Competition: Task 1 (classification) at ICDAR 2026. The objective is to accurately classify individual scientific figure panels from Atomic Layer Deposition and Etching (ALD/E) literature into a domain-specific taxonomy. We formulate this challenge as a conditional t...
Soumyajyoti Mohanta, Shresth Kasyap, Sasmit Shashwat et al.· Open Conference Proceedings· 0 citations
This report presents Team TeleOCR-VL’s system for Tasks 2 and 3 of the ICDAR 2026 Sci-ImageMiner Competition. Sci-ImageMiner focuses on scientific chart understanding in real ALD/ALE research papers, requiring systems to extract structured data from complex figures and generate summaries consistent with scientific cont...
Yikun Wang, Peng Cai, Zhaofan Zou et al.· Open Conference Proceedings· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.