Graded feedback for TrustPropRAG

Study TrustPropRAG under graded, rather than binary, human feedback that represents documents as varying in reliability instead of categorizing them only as reliable or unreliable.

Background

TrustPropRAG models human feedback as binary labels: documents are marked either reliable or unreliable, and the optimization procedure propagates these labels through the document relation graph. In practical deployments, however, users may provide feedback with different degrees of confidence or quality rather than an all-or-nothing judgment.

The paper explicitly leaves the extension to graded feedback unresolved. Such an extension would require adapting the feedback representation and potentially the trust-score optimization objective to accommodate continuous or ordinal reliability assessments.

References

Studying TrustPropRAG under graded feedback is left to future work.

— Feedback-Assisted Trust Propagation over Document Relation Graphs for Retrieval-Augmented Generation  (2609.00543 - Li et al., 1 Sep 2026) in Section 'Limitations'

Applying trust propagation to long-form or multi-hop generation, where reliability interacts with compositional reasoning, is left to future work.

— Feedback-Assisted Trust Propagation over Document Relation Graphs for Retrieval-Augmented Generation  (2609.00543 - Li et al., 1 Sep 2026) in Section 'Limitations'