Unified Evidence Fusion Framework Tackles Chaos-Conflict and Historical Reliability

A recent submission on arXiv (2608.13108) introduces a comprehensive reasoning framework based on Dempster-Shafer theory to tackle two ongoing issues in the fusion of multi-source evidence. The first issue arises from current conflict measures that evaluate inter-evidence inconsistency and intra-evidence uncertainty separately, resulting in partial assessments. The second issue is that existing fusion techniques rely solely on immediate comparisons of evidence sources, neglecting their long-term reliability in varied decision-making contexts. The new framework presents a chaos-conflict measurement that simultaneously assesses cross-evidence conflict and intra-evidence non-specificity, supported by five rigorously validated properties. Furthermore, it employs a historical experience-based weighting approach that utilizes spectral clustering and regret theory to derive context-specific reliability profiles from previous fusion results. These elements contribute to a hybrid combination rule that merges the chaos-conflict measurement with historical weighting, enhancing the robustness of evidence fusion in intricate decision-making environments.