Unlike objective type evaluation, descriptive answer evaluation is challenging due to unpredictable answers and free writing style of answers. Because of these, descriptive answer evaluation has received special attention from many researchers. Automatic answer evaluation is useful for the following situations. It can avoid human intervention for marking, eliminates bias marking and most important is that it can save huge manpower. To develop an efficient and accurate system, there are several open challenges. One such open challenge is cleaning the document, which includes struck-out words removal and restoring the struck-out words. In this paper, we have proposed a system for struck-out handwritten word detection and restoration for automatic descriptive answer evaluation. The work has two stages. In the first stage, we explore the combination of ResNet50 and the diagonal line (principal and secondary diagonal lines) segmentation module for detecting words and then classifying struck-out words using a classification network. In the second stage, we explore the combination of U-Net as a backbone and Bi-LSTM for predicting pixels that represent actual text information of the struck-out words based on the relationship between sequences of pixels for restoration. Experimental results on our dataset and standard datasets show that the proposed model is impressive for struck-out word detection and restoration. A comparative study with the state-of-the-art methods shows that the proposed approach outperforms the existing models in terms of struck-out word detection and restoration.

Scene text recognition (STR) has benefited from vision–language modeling and multi-modal in-context learning, which enables models to exploit annotated demonstrations during inference without parameter updating. However, existing in-context STR methods mainly select demonstrations according to global visual similarity. Such image-level representations jointly encode textual content, font appearance, background, degradation, and geometric deformation, making it difficult to distinguish appearance compatibility from character-level relevance. As a result, globally similar samples may fail to provide effective evidence for ambiguous characters. To address this limitation, we propose Scene-to-Glyph, a hierarchical coarse-to-fine retrieval framework for multi-modal in-context scene text recognition. At the coarse-grained stage, an appearance-oriented scene representation is extracted from visual features to characterize shared conditions such as font style, degradation, deformation, and background appearance while reducing the influence of textual content. Scene-level similarity is then used to retrieve a compact set of appearance-compatible contextual candidates. At the fine-grained stage, uncertain character positions are identified from the recognition output, and decoder–vision correspondence is exploited to derive localized glyph representations. These representations are matched with character-level features from the candidate demonstrations, enabling the model to select demonstrations that are both scene-compatible and relevant to ambiguous characters. Experiments on conventional and challenging STR benchmarks validate the effectiveness of Scene-to-Glyph and the complementary roles of scene-level and glyph-level retrieval.

Retrieval-Augmented Generation (RAG) has been widely utilized to integrate external knowledge into Large Language Models (LLMs) for enhancement onmultiple tasks. Recently, graph-guided RAG approaches have demonstrated stronger support for more accurate, context-aware responses. However, traditional RAG methods rely on static pipelines, failing to adaptively balance heterogeneous modalities on complex tasks. To address this issue, we propose GMS-RAG, a novel agentic RAG framework. GMS-RAG operates via a bi-stage autonomous agent system: a Graph-Memory Construction Agent and a Modality-Stratified Retrieval Agent. The construction stage first integrates visual contents along with textual ones into a unified graph memory, then a balanced retrieval mechanism is employed to aggregate evidence for answer generation. Experiments on four benchmarks indicate that GMS-RAG effectively enhances the multi-modal RAG capability of backbone models. Performance measured by F1 score shows that our framework outperforms all baselines with improved quality of answer generation and generalizability over modalities. The code will be available soon.