Revisiting Structural Dependency in Autoregressive Multi-Task Table Recognition via Order-Independent Cell-Level Representations
For table recognition tasks, this work addresses a key bottleneck in autoregressive multi-task frameworks, offering a practical improvement in accuracy and efficiency.
Multi-task table recognition suffers from order-dependent cell representations due to autoregressive decoders. The proposed structural refinement module with non-causal attention achieves order-independent cell features, improving cell localization and end-to-end recognition while reducing inference time by ~3x.
Multi-task table recognition jointly addresses table structure prediction, cell localization, and cell content recognition within a unified framework. Existing approaches often rely on autoregressive decoders to generate table structures and reuse their hidden states for cell localization and content recognition. This autoregressive generation process can make cell representations order-dependent, degrading global consistency across cells. This paper proposes a structural refinement module that produces order-independent cell features through non-causal attention. This design enables parallel inference of cell contents while conditioning each cell on global context encoded in the refined features. Experiments on two large datasets demonstrate consistent gains in cell localization and end-to-end recognition, while reducing overall inference time by around threefold.