Search papers, labs, and topics across Lattice.
This study investigates pixel-level table compression for question answering in documents that contain multiple tables, focusing on the impact of image representation on model performance. By evaluating five vision-language models (VLMs) across various benchmarks and visual-token budgets, the authors find that while native resolution tables yield optimal performance, downscaled tables lead to less effective reasoning despite preserving relevance signals. The proposed two-step method, which first identifies relevant tables from a pixel-compressed context and then reasons over them at native resolution, achieves a 41% token reduction and a 7-point accuracy improvement compared to traditional single-step approaches.
Pixel-level table compression can dramatically reduce token usage while enhancing accuracy in document question answering, challenging conventional methods.
Answering questions over real-world documents requires processing long inputs that interleave text with tables. Optical context compression, which represents context as images, promises to reduce token cost, but its effect on table understanding remains unclear. We study pixel-level table compression for question answering over documents with multiple tables, evaluating five VLMs across two benchmarks and five visual-token budgets. Representing tables as images at native resolution matches text in both performance and efficiency, but downscaling them makes models compensate the loss in readability with longer, less effective reasoning traces that cancel the expected savings. Highly downscaled tables, however, preserve enough signal to identify whether they are relevant to a question. We exploit this asymmetry with a training-free, two-step method: the model first identifies the tables needed to answer a question from a pixel-compressed context, and then reasons over those at native resolution. On long documents, our method saves 41% of total tokens and gains 7 accuracy points over single-step QA with native resolution tables. It also uses 15% fewer tokens than the most efficient single-step compressed configuration, with no accuracy loss.
