Ancient Chinese documents are vital for historical research, necessitating high-precision character recognition and layout analysis for digitization. This paper introduces EvaHan2026, the inaugural international shared task for simultaneous optical character recognition and layout parsing of ancient texts. The evaluation framework comprehensively assesses model performance across diverse calligraphic styles and complex structures, including main body text, interlinear annotations, and illustrations. Among thirteen participating teams, four successfully completed all tasks within the closed track. Experimental results reveal that character recognition accuracy reached 97.36% on engraved texts (Test Set A) and 95.71% on handwritten texts (Test Set C) when accounting for character variants. For layout recognition in complex layouts (Test Set B), the best team achieved a peak mean Average Precision (mAP) of 59.41% and an Intersection over Union (loU) of 76.38%. Our analysis indicates that calligraphic variability, layout density, and character variants significantly modulate system performance. Consequently, enhancing robustness within complex layouts and developing synergistic models that integrate textual and structural information remain primary challenges for intelligent interpretation of ancient writings .

While Vision-Language Models (VLMs) have demonstrated strong capabilities in general multimodal tasks, they face significant challenges in low-resource, cross-temporal settings such as ancient Chinese documents, where models must overcome complex visual noise and leverage deep linguistic priors. The functional organization of task-relevant components in existing VLMs remains understudied from a computational interpretability perspective. To address this, we introduce AncientVision-3T, an intervention-oriented diagnostic benchmark designed for mechanistic interpretability experiments on VLMs, organized across OCR, image classification, and image understanding tasks of increasing abstraction. Based on this benchmark, we propose the Ancient Chinese Target-Conditioned Gradient-based Neuron analysis framework (AC-TCGN), which identifies and validates task-relevant neurons for dissecting functional organization in general and domain-specific VLMs. Our experiments demonstrate that AC-TCGN effectively identifies neurons that are critical for model predictions. Moreover, our experimental results reveal a transition from separation to entanglement in the functional organization of decoder feed-forward networks: general VLMs show stronger task-wise separability, whereas the ancient Chinese domain-specific VLM shows greater cross-task neuron sharing and reduced task-wise separability. This suggests enhanced coupling between visual form and historical-semantic processing in ancient Chinese document image-text tasks.