Large language models (LLMs) can infer user preferences from interaction histories and reviews, yet the rationales they produce often do not reflect the information actually used for recommendation. When the link between evidence and rationale is weak or the rationale has little impact on the final ranking, a "grounding‑influence gap" arises.
We introduce the PROVE‑REC framework to enable verifiable preference reasoning. Pass A compresses the full pre‑target history into a compact preference proof consisting of positive and avoidance claims, each linked to selected evidence entries. Pass B predicts the next item using only this proof and its evidence, preventing the recommender from bypassing the reasoning path.
To verify evidence‑to‑proof grounding, we compare the effect of masking the selected evidence with masking a comparable control entry. To verify proof‑to‑recommendation influence, we remove a single claim and measure the resulting drop in the target item's ranking margin. A ranking‑preservation objective further retains useful information from the complete history.
Comprehensive experiments on diverse real‑world datasets show that PROVE‑REC consistently outperforms strong sequential, generative, and LLM‑enhanced baselines, achieving up to a 7.45% improvement. Controlled ablations confirm the effectiveness of the two‑pass architecture and verification objectives. Moreover, PROVE‑REC produces claims that are more strongly grounded in historical evidence and more influential to recommendation while preserving ranking quality.
Review