提出新评估基准与双证据融合方法，提升分割归因的可信度
Toward Faithful Segmentation Attribution via Benchmarking and Dual-Evidence Fusion
- 构建干预式评估基准，测试归因的因果性与稳定性
- 双证据融合使归因在梯度不稳时仍保持因果支持，删除法性能更优
- 适用于需要高可信解释的医学、自动驾驶等关键场景
现有语义分割归因方法多依赖视觉合理性判断，但看似合理未必真实反映模型决策依据，也易出现跨区域泄漏。本文提出可复现的评估基准，基于Pascal VOC和SBD数据集，针对三种预训练骨干网络，在干预忠实性、非目标泄漏、扰动鲁棒性和运行时间四方面进行评测。为验证基准，提出轻量级双证据归因（DEA）方法，通过一致性加权融合梯度证据与区域干预信号，强化两源一致区域的注意力，并在梯度不稳定时保留因果支持。所有实验表明，DEA在删除法忠实性上持续优于仅用梯度的方法，且保持强鲁棒性，仅需额外干预计算开销。该基准揭示了归因方法间存在的可信度-稳定性权衡，此问题在纯视觉评估下完全不可见，为分割可解释性方法选择提供了理论基础。代码已开源。
原文摘要 · Abstract (English)
Attribution maps for semantic segmentation are almost always judged by visual plausibility. Yet looking convincing does not guarantee that the highlighted pixels actually drive the model's prediction, nor that attribution credit stays within the target region. These questions require a dedicated evaluation protocol. We introduce a reproducible benchmark that tests intervention-based faithfulness, off-target leakage, perturbation robustness, and runtime on Pascal VOC and SBD across three pretrained backbones. To further demonstrate the benchmark, we propose Dual-Evidence Attribution (DEA), a lightweight correction that fuses gradient evidence with region-level intervention signals through agreement-weighted fusion. DEA increases emphasis where both sources agree and retains causal support when gradient responses are unstable. Across all completed runs, DEA consistently improves deletion-based faithfulness over gradient-only baselines and preserves strong robustness, at the cost of additional compute from intervention passes. The benchmark exposes a faithfulness-stability tradeoff among attribution families that is entirely hidden under visual evaluation, providing a foundation for principled method selection in segmentation explainability. Code is available at https://github.com/anmspro/DEA.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。