TraceBack: Multi-Agent Decomposition for Fine-Grained Table Attribution

TL;DR

TraceBack introduces a multi-agent framework for fine-grained, cell-level attribution in table question answering, improving transparency and interpretability with a new benchmark and evaluation metric.

Contribution

It presents a scalable, modular approach for detailed cell attribution in table QA, along with a new benchmark and a reference-less evaluation metric.

Findings

TraceBack outperforms baselines across datasets and granularities.

FairScore effectively estimates attribution quality without human labels.

The framework enhances transparency in table QA systems.

Abstract

Question answering (QA) over structured tables requires not only accurate answers but also transparency about which cells support them. Existing table QA systems rarely provide fine-grained attribution, so even correct answers often lack verifiable grounding, limiting trust in high-stakes settings. We address this with TraceBack, a modular multi-agent framework for scalable, cell-level attribution in single-table QA. TraceBack prunes tables to relevant rows and columns, decomposes questions into semantically coherent sub-questions, and aligns each answer span with its supporting cells, capturing both explicit and implicit evidence used in intermediate reasoning steps. To enable systematic evaluation, we release CITEBench, a benchmark with phrase-to-cell annotations drawn from ToTTo, FetaQA, and AITQA. We further propose FairScore, a reference-less metric that compares atomic facts derived from predicted cells and answers to estimate attribution precision and recall without human cell labels. Experiments show that TraceBack substantially outperforms strong baselines across datasets and granularities, while FairScore closely tracks human judgments and preserves relative method rankings, supporting interpretable and scalable evaluation of table-based QA.