We present the multimodal Optical Character Recognition (OCR) and layout analysis methods developed for the EvaHan 2026 competition. Our approach is built upon the Qwen2.5-VL-7B-Instruct architecture and integrates two core strategies: (1) a reinforcement learning alignment pipeline utilizing Direct Preference Optimization (DPO) and Group Relative Policy Optimization (GRPO) to explicitly mitigate hallucination and coordinate instability; and (2) a four-stage curriculum learning framework that synthesizes domain-specific historical artifacts to enhance open-modality generalization. Using this approach, we achieve competitive results, notably reaching a Character Error Rate (CER) of 0.0303 on printed texts (Task A) and 0.0552 on handwritten manuscripts (Task C), as well as an Average Intersection over Union (IoU) of 0.7638 on layout element analysis (Task B).

Early Long-context Document Visual Question Answering (DocVQA) methods struggle with preserving visual semantics or handling finite context windows. Conversely, recent RAG-based approaches suffer from “semantic gaps” and “structural disconnections” due to passive retrieval mechanisms that ignore logical dependencies. To address these challenges, we introduce TRACE (Traversal Retrieval-Augmented Chain of Evidence). By navigating a Bi-Layered Graph that encodes both physical adjacency and semantic relevance, TRACE transforms retrieval from static matching into adaptive evidence chain construction. Furthermore, we propose M5BookVQA, a benchmark designed to assess deep, multi-hop reasoning in books, addressing the limitations of existing datasets. Extensive experiments show that TRACE achieves an average accuracy improvement of 14.07% on M5BookVQA and exhibits robust generalization with a 13.38% gain across four established benchmarks. Our source code is available at https://github.com/shimurenhlq/TRACE.

Word segmentation stands as a cornerstone of Natural Language Processing (NLP). Based on the concept of “comprehend first, segment later”, we propose a new framework to explore the limit of unsupervised word segmentation with Large Language Models (LLMs) and evaluate the semantic understanding capabilities of LLMs based on word segmentation. We employ current mainstream LLMs to perform word segmentation across multiple languages to assess LLMs’ “comprehension”. Our findings reveal that LLMs are capable of following simple prompts to segment raw text into words. There is a trend suggesting that models with more parameters tend to perform better on multiple languages. Additionally, we introduce a novel unsupervised method, termed LLACA (Large Language Model-Inspired Aho-Corasick Automaton). Leveraging the advanced pattern recognition capabilities of Aho-Corasick automata, LLACA innovatively combines these with the deep insights of well-pretrained LLMs. This approach not only enables the construction of a dynamic n-gram model that adjusts based on contextual information but also integrates the nuanced understanding of LLMs, offering significant improvements over traditional methods. Our source code is available at https://github.com/hkr04/LLACA