Research map / Architectures and efficiency / Attention and transformers
AttCAT: Explaining Transformers via Attentive Class Activation Tokens
Yao Qiang, Deng Pan, Chengyin Li, Xin Li, Rhongho Jang, Dongxiao Zhu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add self-attention weights to CAT (CAT -> AttCAT) (ablation) | AttCAT · Amazon (BERT) | AOPC: 0.409 -> 0.457 | Improved |
| add self-attention weights to CAT (CAT -> AttCAT) (ablation) | AttCAT · IMDB (BERT) | AOPC: 0.406 -> 0.528 | Improved |
| add self-attention weights to CAT (CAT -> AttCAT) (ablation) | AttCAT · SST2 (BERT) | AOPC: 0.352 -> 0.371 | Improved |
| AttCAT vs TransAtt | AttCAT · IMDB (BERT) | LOdds: -2.137 -> -3.671 | Improved |
| AttCAT vs TransAtt | AttCAT · MNLI (BERT) | AOPC: 0.130 -> 0.164 | Improved |
| AttCAT vs TransAtt | AttCAT · QQP (BERT) | AOPC: 0.145 -> 0.139 | Hurt |
| AttCAT vs RawAtt on QQP | AttCAT · QQP (BERT) | LOdds: 0.149 -> 0.073 | Improved |
| AttCAT vs best baseline confidence | AttCAT · Kendall-tau, SST2 | Kendall-tau: 0.101 -> 0.018 | Improved |
| AttCAT vs CAT confidence on QQP | AttCAT · Kendall-tau, QQP | Kendall-tau: 0.222 -> 0.349 | Hurt |
| summing layer impact scores instead of rollout (ablation) | AttCAT · BERT layer heatmap | impact score | Improved |
| attention-based methods (RawAtt, Rollout) vs gradient methods | RawAtt/Rollout · 6 datasets | AOPC and LOdds | Hurt |
Later papers that compared with this method
1 later accepted paper reports AttCAT as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- LibraGrad: Balancing Gradient Flow for Universally Better Vision Transformer Attributions (CVPR 2025 · better on 4 of 4 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- AttnLRP: Attention-Aware Layer-Wise Relevance Propagation for Transformers (ICML 2024 · similarity 0.89)
- XAI for Transformers: Better Explanations through Conservative Propagation (ICML 2022 · similarity 0.87)
- Self-Attention Attribution: Interpreting Information Interactions Inside Transformer (AAAI 2021 · similarity 0.87)
- Revisiting LRP: Positional Attribution as the Missing Ingredient for Transformer Explainability (NeurIPS 2025 · similarity 0.87)
- Trainable Transformer in Transformer (ICML 2024 · similarity 0.86)
- Token Transformation Matters: Towards Faithful Post-hoc Explanation for Vision Transformer (CVPR 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
