Cycle-of-Thought: Self-Evaluative Iterative Reasoning for Table Visual Question Answering in Document Images

A paper posted to arXiv (2610.32423v1) proposes Self-RefineVQA, a training-free, inference-time self-correction framework that lets a single vision-language model iteratively act as generator, critic and refiner to improve table visual question answering in document images without labeled data or model updates. The authors report consistent accuracy gains on tabular VQA benchmarks in monolingual and cross-lingual settings, with the largest boost for small models, enabling 2B-7B parameter models to rival significantly larger counterparts. A comparative complexity analysis finds inference cost rises linearly while data efficiency improves, lowering cumulative cost by removing the human-annotation bottleneck required for supervised fine-tuning.

arXiv:2610.32423v1 Announce Type: cross Abstract: Recent advances in Vision-Language Models VLMs have significantly improved Visual Question Answering VQA performance, yet their reasoning over structured content such as tables remains inconsistent, often plagued by spatial misalignment and logical hallucinations. Existing approaches for improving table VQA typically rely on architectural modifications or expensive supervised fine-tuning SFT , which impose high barriers to entry due to the need for large-scale, expert-annotated datasets and significant computational resources. In this work, we propose Self-RefineVQA, a training-free, inference-time self-correction framework that improves table VQA performance without additional labeled data or model updates. Inspired by human revision behavior, our method enables a single VLM to iteratively act as a generator, critic, and refiner, forming a closed self-feedback loop that corrects common reasoning errors through prompt-guided re-evaluation. Experiments on tabular VQA benchmarks demonstrate that Self-RefineVQA yields consistent accuracy gains in monolingual and cross-lingual understanding, with a particularly significant boost for small models, enabling lightweight models e.g., 2B-7B parameters to rival the performance of significantly larger counterparts. Furthermore, we provide a comparative complexity analysis demonstrating that while inference cost increases linearly, our framework offers superior data efficiency and a lower cumulative cost by eliminating the prohibitive annotation bottleneck of human labeling required for fine-tuning. Our results highlight the potential of prompt-based self-improvement as a scalable, accessible paradigm for reliable table reasoning in VLMs.