ArbGraph Credibility Arbitration Framework Overview

Paper context

Paper title: ArbGraph: Conflict-Aware Evidence Arbitration for Reliable Long-Form Retrieval-Augmented Generation Abstract: Retrieval-augmented generation (RAG) remains unreliable in long-form settings, where retrieved evidence is noisy or contradictory, making it difficult for RAG pipelines to maintain factual consistency. Existing approaches focus on retrieval expansion or verification during generation, leaving conflict resolution entangled with generation. To address this limitation, we propose ArbGraph, a framework for pre-generation evidence arbitration in long-form RAG that explicitly resolves factual conflicts. ArbGraph decomposes retrieved documents into atomic claims and organizes them into a conflict-aware evidence graph with explicit support and contradiction relations. On top of this graph, we introduce an intensity-driven iterative arbitration mechanism that propagates credibility signals through evidence interactions, enabling the system to suppress unreliable and inconsistent claims before final generation. In this way, ArbGraph separates evidence validation from text generation and provides a coherent evidence foundation for downstream long-form generation. We evaluate ArbGraph on two widely used long-form RAG benchmarks, LongFact and RAGChecker, using multiple large language model backbones. Experimental results show that ArbGraph consistently improves factual recall and information density while reducing hallucinations and sensitivity to retrieval noise. Additional analyses show that these gains are evident under conflicting or ambiguous evidence, highlighting the effectiveness Passages referencing this figure: the conventional linear RAG pipeline into an arbitrate-first, generate-later paradigm. Instead of implicitly integrating potentially conflicting evidence during generation, ArbGraph explicitly performs evidence-level arbitration prior to decoding. This design separates credibility reasoning from text generation , enabling more robust handling of noisy and contradictory evidence. As illustrated in Figure 1 , ArbGraph consists of three stages: (1) query-aware claim filtering, (2) evidence graph construction, and (3) iterative credibility arbitration. These stages progressively refine the evidence pool and produce a validated claim set 𝒱 val \mathcal{V}_{\text{val}} , which serves as the trusted context for long-form generation. We briefly outline each stage below and defer implementation det of three stages: (1) query-aware claim filtering, (2) evidence graph construction, and (3) iterative credibility arbitration. These stages progressively refine the evidence pool and produce a validated claim set 𝒱 val \mathcal{V}_{\text{val}} , which serves as the trusted context for long-form generation. We briefly outline each stage below and defer implementation details to subsequent sections. Figure 1 . Overview of the ArbGraph framework. The pipeline comprises three stages: query-aware claim filtering, evidence graph construction, and iterative credibility arbitration, which together realize an arbitrate-first, generate-later paradigm. Stage I: Query-Aware Claim Filtering. ArbGraph first converts retrieved documents into atomic claims and filters them with respect to the query. This s