Implicit-Knowledge Visual Question Answering with Structured Reasoning Traces
This addresses the issue of weak and inconsistent reasoning in IK-KVQA models, offering a more transparent and accurate approach for multimodal AI applications, though it is an incremental improvement over existing methods.
The paper tackles the problem of implicit-knowledge visual question answering (IK-KVQA), where models answer questions using only multimodal large language models without external knowledge retrieval, by proposing a framework that adds structured reasoning traces to improve reasoning transparency and accuracy. The result is up to 11.3% higher answer accuracy on the OK-VQA benchmark compared to the strongest baseline.
Knowledge-based Visual Question Answering (KVQA) requires models to ground entities in images and reason over factual knowledge. Recent work has introduced its implicit-knowledge variant, IK-KVQA, where a multimodal large language model (MLLM) is the sole knowledge source and answers are produced without external retrieval. Existing IK-KVQA approaches, however, are typically trained with answer-only supervision: reasoning remains implicit, justifications are often weak or inconsistent, and generalization after standard supervised fine-tuning (SFT) can be brittle. We propose MODELNAME, a framework that equips IK-KVQA with dual-path structured reasoning traces (symbolic relation paths over text and vision together with path-grounded natural-language explanations) to provide a stronger inductive bias than generic answer-only supervision. These traces act as modality-aware scaffolds that guide the model toward relevant entities and attributes, offering more structure than generic chain-of-thought supervision while not constraining reasoning to any single fixed path. Using a single open-source MLLM, MODELNAME constructs and selects traces to build an offline trace-enriched dataset and then performs structure-aware self-distillation; no external retrievers, verifiers, or curated knowledge bases are used, and inference is a single autoregressive pass. Across benchmarks, MODELNAME consistently improves both answer accuracy and the transparency of intermediate reasoning, achieving up to 11.3% higher answer accuracy on OK-VQA over the strongest baseline.