[Submitted on 27 Jun 2026 (v1), last revised 22 Oct 2026 (this version, v3)]
Abstract:We introduce GRAB, a constructor-encoder-bridge pipeline for table question answering. Our method lifts relational data into an heterogeneous graph, encodes it via message passing, and transfers the signals to an LLM through a small set of query-conditioned latent tokens. This provides the LLM with a compact, task-relevant structural representation together with the flattened text. Crucially, the LLM remains strictly frozen to preserve its general reasoning capabilities; we train only the lightweight graph encoder and latent bridge (91M parameters), allowing the entire pipeline to be trained efficiently. Our pipeline significantly improves performance on relational Question Answering, with the largest gains in demanding multi-table settings, offering an efficient, principled way to connect relational deep learning with LLMs.
Submission history
From: Simone Varriale [view email]
[v1] Sat, 27 Jun 2026 13:48:20 UTC (2,228 KB)
[v2] Wed, 21 Oct 2026 15:06:47 UTC (410 KB)
[v3] Thu, 22 Oct 2026 15:52:50 UTC (410 KB)
