提出无需修改模型的通用神经网络归因方法，可适配任意架构。
Always Keep Your Promises: A Model-Agnostic Attribution Algorithm for Neural Networks
- 在计算图操作级别实现归因，不依赖特定模型结构。
- 在多个模型上达到或超越专用方法的归因精度，最高达95.06%。
- 适用于复杂架构如Mamba、Whisper，无需额外代码扩展。
层间相关性传播（LRP）通过守恒性质和深度泰勒分解基础为神经网络提供合理的归因。然而现有实现仅在模块层面运行，需针对不同架构设计传播规则并修改模型，限制了通用性和可持续性。本文提出DynamicLRP，一种在张量操作层面运行的模型无关LRP框架。通过将归因分解至计算图中的单个操作，并引入名为‘承诺系统’的新机制以延迟激活解析，该方法在保持LRP理论保证的同时实现真正的架构无关性。其设计独立于反向传播机制，无需模型修改，可与梯度反向传播并行执行。基于计算图，该方法理论上可扩展至支持自动微分的其他深度学习库。我们在多种模型上验证：在VGG上达到1.77的ABPC，优于专用实现（1.69）；在ViT上性能相当；在SQuADv2上对RoBERTa-large和Flan-T5-large的归因准确率分别为93.70%和95.06%。在包含状态空间模型（Mamba）、音频Transformer（Whisper）及多模态系统（DePlot）在内的15种架构中，覆盖31,465个计算图节点，节点覆盖率高达99.92%，仅用47种基础操作规则，无任何模型特异性代码。操作级分解与承诺系统为未来架构演进提供了可持续、可扩展的LRP基础。所有代码已开源：https://github.com/keeinlev/dynamicLRP。
原文摘要 · Abstract (English)
Layer-wise Relevance Propagation (LRP) provides principled attribution for neural networks through conservation properties and foundations in Deep Taylor Decomposition. However, existing implementations operate at the module level, requiring architecture-specific propagation rules and model modifications. These limit the generality of target model and sustainability of implementations as architectures evolve. We introduce DynamicLRP, a model-agnostic LRP framework operating at the tensor operation level. By decomposing attribution to individual operations within computation graphs and introducing a novel mechanism for deferred activation resolution, named the Promise System, our approach achieves true architecture agnosticity while maintaining LRP's theoretical guarantees. This design operates independently of backpropagation machinery, requiring no model modification, enabling side-by-side execution with gradient backpropagation. Being based on computation graphs, this method is theoretically extensible to other deep learning libraries that support auto-differentiation. We demonstrate faithfulness matching or exceeding specialized implementations (1.77 vs 1.69 ABPC on VGG, equivalent performance on ViT, 93.70% and 95.06% top-1 attribution accuracy for explaining RoBERTa-large and Flan-T5-large answers on SQuADv2, respectively) while maintaining practical efficiency on models with 100M-1B parameters. We achieved 99.92% node coverage across 31,465 computation graph nodes from 15 diverse architectures, including state-space models (Mamba), audio transformers (Whisper), and multimodal systems (DePlot) without any model-specific code with rules for 47 fundamental operations implemented. Our operation-level decomposition and Promise System establish a sustainable, extensible foundation for LRP across evolving architectures. All code is available at https://github.com/keeinlev/dynamicLRP .
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。