Research map / Architectures and efficiency / Attention and transformers
ATMAN: Understanding Transformer Predictions Through Memory Efficient Attention Manipulation
Björn Deiseroth, Mayukh Deb, Samuel Weinbach, Manuel Brack, Patrick Schramowski, Kristian Kersting
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 |
|---|---|---|---|
| use AtMan attention-manipulation instead of Chefer gradient method | GPT-J · SQuAD | mAP: 72.7 -> 73.7 | Improved |
| use AtMan instead of Chefer | GPT-J · SQuAD | mAR: 96.6 -> 93.4 | Hurt |
| use AtMan instead of Chefer | MAGMA-6B · OpenImages | mAP: 58.3 -> 65.5 | Improved |
| use AtMan instead of Chefer | MAGMA-6B · OpenImages | mAR: 11.7 -> 15.7 | Improved |
| add cosine-similarity correlated token suppression (ablation) | AtMan · OpenImages | precision/recall | Improved |
| scale MAGMA from 6B to 13B/30B | AtMan · OpenImages | explanation quality | Mixed |
| apply AtMan to BLIP image encoder | AtMan · OpenImages | explanation quality | No clear effect |
| use AtMan instead of gradient methods for memory | Transformers · GPU memory | memory consumption | Improved |
| use AtMan instead of Chefer for runtime | MAGMA · explaining image | runtime | Hurt |
Later papers that compared with this method
1 later accepted paper reports AtMan as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- AttnLRP: Attention-Aware Layer-Wise Relevance Propagation for Transformers (ICML 2024 · better on 2 of 2 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 0.87)
- An Evolved Universal Transformer Memory (ICLR 2025 · similarity 0.86)
- Parallelizing Linear Transformers with the Delta Rule over Sequence Length (NeurIPS 2024 · similarity 0.86)
- Transformers Learn Faster with Semantic Focus (NeurIPS 2025 · similarity 0.86)
- LaVin-DiT: Large Vision Diffusion Transformer (CVPR 2025 · similarity 0.85)
- Pool Me Wisely: On the Effect of Pooling in Transformer-Based Models (NeurIPS 2025 · similarity 0.85)
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.
