Pith. sign in

REVIEW 5 cited by

AttnLRP: Attention-Aware Layer-Wise Relevance Propagation for Transformers

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2402.05602 v2 pith:F4Q75CTY submitted 2024-02-08 cs.CL cs.AIcs.CVcs.LG

Reduan Achtibat , Sayed Mohammad Vakilzadeh Hatefi , Maximilian Dreyer , Aakriti Jain , Thomas Wiegand , Sebastian Lapuschkin , Wojciech Samek This is my paper · ORCID
classification cs.CLcs.AIcs.CVcs.LG
keywords transformercomputationalefficiencylatentlayer-wisemethodmethodsmodels
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Large Language Models are prone to biased predictions and hallucinations, underlining the paramount importance of understanding their model-internal reasoning process. However, achieving faithful attributions for the entirety of a black-box transformer model and maintaining computational efficiency is an unsolved challenge. By extending the Layer-wise Relevance Propagation attribution method to handle attention layers, we address these challenges effectively. While partial solutions exist, our method is the first to faithfully and holistically attribute not only input but also latent representations of transformer models with the computational efficiency similar to a single backward pass. Through extensive evaluations against existing methods on LLaMa 2, Mixtral 8x7b, Flan-T5 and vision transformer architectures, we demonstrate that our proposed approach surpasses alternative methods in terms of faithfulness and enables the understanding of latent representations, opening up the door for concept-based explanations. We provide an LRP library at https://github.com/rachtibat/LRP-eXplains-Transformers.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Towards the Explainability of Temporal Graph Networks via Memory Backtracking and Topological Attribution

    cs.LG 2026-07 conditional novelty 6.5 of 10

    MemExplainer attributes TGN logits to neighboring and historical events via LRP on topology and memory trees, then optimizes event selection for fidelity, outperforming baselines on nine datasets.

  2. Causal Interpretation of Sparse Autoencoder Features in Vision

    cs.CV 2025-08 conditional novelty 6.0 of 10

    CaFE uses attribution-based effective receptive fields to explain sparse autoencoder features in vision transformers, recovering activations better than activation-ranked patches.

  3. The Hidden Dimensions of LLM Alignment: A Multi-Dimensional Analysis of Orthogonal Safety Directions

    cs.CL 2025-02 conditional novelty 6.0 of 10

    Safety refusal in Llama 3.1 8B is governed by a dominant activation direction plus smaller interpretable directions, and removing prompt tokens that activate these secondary directions can bypass fine-tuned safety.

  4. Attribution-Guided Continual Learning for Large Language Models

    cs.LG 2026-05 unverdicted novelty 5.0 of 10

    LRP-derived element-wise parameter importance scores gate gradients so parameters critical to earlier tasks receive smaller updates during continual LLM fine-tuning.

  5. Perspectives for Direct Interpretability in Multi-Agent Deep Reinforcement Learning

    cs.AI 2025-02 unverdicted novelty 4.0 of 10

    A perspective paper that advocates direct, post hoc interpretability for multi-agent deep reinforcement learning and offers a taxonomy of where those methods might apply.

Pith tools