Tagged articles
Document Understanding
24 articles · Page 1 of 1
Tencent has open-sourced WeKnora, a RAG-based framework for understanding and semantically searching complex documents like PDFs and Word files, offering Agent-mode reasoning, hybrid retrieval, MCP tool integration, and private deployment, with over 36k GitHub stars.
AgentDocument UnderstandingLLM
0 likes · 8 min read

Tencent open-sources WeKnora, a modular RAG framework that combines multimodal document parsing, hybrid retrieval (keyword, vector, knowledge graph), and agent-based reasoning to enable semantic search and Q&A over complex documents, with support for local deployment and flexible model integration.
AgentDocument UnderstandingLLM
0 likes · 8 min read

This survey analyzes 150+ papers on long-document visual question answering, decomposing systems into parsing, evidence retrieval, generation, and evaluation components, revealing how research focus shifted from content understanding to sufficient evidence acquisition and cross-turn evidence management, with the key insight that local relevance doesn't guarantee sufficient evidence.
Document UnderstandingEvidence RetrievalLong-Document VQA
0 likes · 23 min read

Doc‑V* introduces a dynamic, thumbnail‑driven approach that lets a model decide which pages to read, achieving a 49.7% improvement over RAG variants on multi‑page document QA benchmarks without larger models or longer context windows, and demonstrates how strategic evidence acquisition outperforms naïve full‑document reading.
AIDocument UnderstandingRAG
0 likes · 10 min read

Xiaomi’s AI team announced seven ACL 2026 papers that span low‑bit KV‑cache quantization for 8.3× faster LLM inference, OCR‑free multi‑page document VQA, a new attention‑basin analysis, non‑autoregressive spoken dialogue generation, a comprehensive mobile‑agent benchmark, a success‑rate‑aware training policy, and a progressive universal information‑extraction framework.
Document UnderstandingInference Optimizationbenchmark
0 likes · 12 min read

UniRec‑0.1B, a lightweight OCR model with only 0.1 B parameters, achieves accuracy comparable to or better than multi‑billion‑parameter visual‑language models across text, formula, and mixed‑content tasks, thanks to hierarchical supervision training, a semantic‑decoupled tokenizer, and a large 40 M‑sample dataset, while delivering 2‑9× faster inference and full open‑source availability.
Document UnderstandingHierarchical SupervisionOCR
0 likes · 12 min read

This weekly AI paper roundup reviews five cutting‑edge OCR studies—DeepSeek‑OCR 2, LightOnOCR‑2‑1B, HunyuanOCR, PaddleOCR‑VL, and GOT—detailing their novel visual‑language architectures, training data, benchmark evaluations, and performance gains over previous models.
DeepSeekDocument UnderstandingGoT
0 likes · 9 min read

DeepSeek-OCR 2 introduces a novel dual‑stream (bidirectional + causal) attention architecture that replaces fixed raster scanning, leverages a Qwen2‑0.5B encoder, and achieves state‑of‑the‑art accuracy on OmniDocBench while reducing token budget and improving reading‑order consistency.
DeepEncoderDeepSeekDocument Understanding
0 likes · 8 min read

This article presents a training‑free, architecture‑agnostic method that leverages LaTeX‑style structured inputs to preserve document hierarchy and spatial relationships, thereby improving multimodal large language model performance on document question answering tasks across multiple benchmarks.
AIDocQADocument Understanding
0 likes · 8 min read


AntTechApr 10, 2025 · Artificial Intelligence
At the ICLR 2025 live session in Singapore, Ant Group showcased four cutting‑edge papers—CodePlan, Animate‑X, Group Position Embedding, and OmniKV—demonstrating advances in large‑language‑model reasoning, universal character animation, layout‑aware document understanding, and efficient long‑context inference.
AI researchDocument Understandinglarge language models
0 likes · 6 min read

Mistral AI’s newly launched OCR API claims to deliver world‑class document understanding with multilingual support, high speed, and self‑hosting options, and benchmark tests show it outperforms Azure OCR and Google Doc AI, yet independent evaluations reveal limitations on complex tables and legal forms, prompting a balanced assessment of its readiness for enterprise use.
AI ModelDocument UnderstandingMistral AI
0 likes · 7 min read

This article explores multimodal Retrieval‑Augmented Generation (RAG), detailing five core topics—including semantic extraction, visual‑language models, scaling strategies, technical roadmap choices, and a Q&A—while presenting three implementation pathways, performance evaluations, and future directions for AI‑driven document understanding.
Document UnderstandingRAGTensor Retrieval
0 likes · 11 min read

The article presents a MarkupLM‑based approach that enriches BERT with XPath embeddings to jointly model webpage text and structure, enabling site‑level detection of malicious content‑scraping pages that bypass traditional rule‑based filters and demonstrating the critical role of structural cues in improving spam classification accuracy.
Document UnderstandingMarkupLMXPath embedding
0 likes · 16 min read


NewBeeNLPJan 2, 2025 · Artificial Intelligence
This article examines the implementation paths and future prospects of multimodal Retrieval‑Augmented Generation, covering semantic extraction, transformer‑based OCR, visual language models, scaling challenges, tensor indexing, and practical evaluations with tools like Infinity and ColPali.
AI RetrievalDocument UnderstandingInfinity Database
0 likes · 12 min read

At the 2024 Global Machine Learning Conference in Beijing, 360 AI Research Institute showcased cutting‑edge multimodal large‑model research, fine‑grained open‑world object detection, and document understanding technologies, highlighting open‑source releases, real‑world deployments, and competitive achievements in AI competitions.
AI researchDocument Understandingknowledge graph
0 likes · 7 min read

The talk outlines RAG2.0’s design challenges—low vector recall, complex documents, semantic gaps—and presents a two‑stage architecture using deep multimodal understanding and knowledge‑graph‑enhanced retrieval, detailing advanced chunking, multi‑index and multi‑path retrieval, efficient sorting models like ColBERT, and future multi‑modal and memory‑augmented agent directions.
ColBERTDelayed InteractionDocument Understanding
0 likes · 23 min read

The 360LayoutAnalysis project from 360 AI Lab releases lightweight, yolov8‑based layout analysis models covering Chinese and English papers, Chinese research reports, and a general document scenario, providing fast inference, paragraph‑level detection, and open‑source code and weights for flexible document‑understanding pipelines.
AI ModelDocument UnderstandingYOLOv8
0 likes · 9 min read

This article presents a comprehensive overview of document intelligence, covering its background, the evolution of related technologies, large‑model approaches such as multimodal pre‑training and domain‑specific models, and concrete enterprise use cases across various business functions.
Document Understandingdocument intelligenceenterprise AI
0 likes · 14 min read


AntTechAug 25, 2023 · Artificial Intelligence
LayoutGCN is a lightweight, graph‑based framework that jointly encodes text, layout, and image features of visually rich documents, achieving competitive performance on multiple downstream tasks while drastically reducing model size and computational cost, making it suitable for edge deployment.
Document UnderstandingGraph Neural NetworkLayoutGCN
0 likes · 24 min read


AntTechJul 31, 2023 · Artificial Intelligence
LayoutMask introduces a novel multi-modal pre‑training model that replaces global 1D position with local 1D position and adds Whole Word Masking, Layout‑Aware Masking, and Masked Position Modeling, achieving state‑of‑the‑art results on various visually‑rich document understanding tasks.
AIDocument UnderstandingMultimodal Pretraining
0 likes · 15 min read

This article provides a detailed introduction to OCR technology, covering its fundamental concepts, major categories (document, scene, and handwritten OCR), typical processing pipelines, a suite of open‑source models on ModelScope—including detection, recognition, and table OCR—and recent multimodal pre‑training methods such as VLDoc and VLPT.
Document UnderstandingModelScopeOCR
0 likes · 15 min read


AntTechJun 15, 2022 · Artificial Intelligence
XYLayoutLM introduces a layout‑aware multimodal network that improves visually‑rich document understanding by augmenting XY‑Cut for robust reading order generation and employing a Dilated Conditional Position Encoding to handle variable‑length inputs, achieving state‑of‑the‑art performance on XFUN and FUNSD datasets.
Document UnderstandingXYCutlayout-aware
0 likes · 10 min read

The article outlines five major trends—neural-network-enhanced search, semantic search, document understanding, image and voice search, and knowledge graphs—that are transforming enterprise use of unstructured data by leveraging AI to deliver precise, context-aware answers and insights.
AIDocument UnderstandingSemantic Search
0 likes · 15 min read

The article outlines how advances in neural networks, semantic search, document understanding, image and voice search, and knowledge graphs are transforming enterprise search of unstructured data, enabling more accurate, context‑aware answers and new business use cases across organizations.
AIDocument UnderstandingSemantic Search
0 likes · 13 min read
