Beyond Attention: Signed Integrated Gradients Attribution in a BiomeGPT-Style Microbiome Transformer
Oren Nelson
cs.LG
Submitted: 2026-08-06
Updated: 2026-08-10
Comments: 15 pages, 6 figures. Code: https://github.com/nohren/token-source-attribution
Code: https://github.com/nohren/token-source-attribution
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Towards better understanding of gradient-based attribution methods for Deep Neural Networks
- Attention is not Explanation
- Explaining Explanations: Axiomatic Feature Interactions for Deep Networks
- Captum: A unified and generic model interpretability library for PyTorch
- Uniform Discretized Integrated Gradients: An effective attribution based method for explaining large language models
- Axiomatic Attribution for Deep Networks
- Attention is not not Explanation
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks