Don't Go Breaking My LLM: The Impact of Pruning Attention Layers on Explanation Faithfulness and Confidence Calibration

arXiv cs.LG Papers
pruning attention-layers llm interpretability faithfulness calibration explainability

Summary

This paper studies how pruning attention layers in LLMs affects explanation faithfulness and confidence calibration, finding that accuracy often remains high but interpretability and reliability degrade, highlighting a misalignment between model confidence, interpretability, and accuracy.

arXiv:2606.24970v1 Announce Type: new Abstract: Pruning Large Language Models (LLMs) reduces memory and inference costs by removing parts of the network, producing smaller models that retain most of their accuracy. As attention layers are the most resource-intensive parts of LLMs, pruning them is a promising compression strategy. Prior work shows that up to 33% of attention layers can be pruned with minimal accuracy loss. Nevertheless, the impact of attention pruning on model interpretability, specifically faithfulness and confidence calibration, remains unstudied. To address this gap, we study how pruning attention layers affects explanation faithfulness and confidence calibration across five LLMs and eight datasets. While the pruned models often maintain high accuracy, we find that their faithfulness and calibration often degrade. Notably, faithfulness and calibration can fluctuate significantly, even when accuracy remains stable, highlighting a misalignment between model confidence, interpretability, and accuracy. Our findings suggest that layer pruning can affect LLMs' interpretability and reliability in ways not captured by accuracy and efficiency measures alone. We recommend including explainability and calibration metrics when evaluating pruned models.
Original Article
View Cached Full Text

Cached at: 06/25/26, 05:08 AM 

# Don't Go Breaking My LLM: The Impact of Pruning Attention Layers on Explanation Faithfulness and Confidence Calibration
Source: [https://arxiv.org/abs/2606.24970](https://arxiv.org/abs/2606.24970)
[View PDF](https://arxiv.org/pdf/2606.24970)

> Abstract:Pruning Large Language Models \(LLMs\) reduces memory and inference costs by removing parts of the network, producing smaller models that retain most of their accuracy\. As attention layers are the most resource\-intensive parts of LLMs, pruning them is a promising compression strategy\. Prior work shows that up to 33% of attention layers can be pruned with minimal accuracy loss\. Nevertheless, the impact of attention pruning on model interpretability, specifically faithfulness and confidence calibration, remains unstudied\. To address this gap, we study how pruning attention layers affects explanation faithfulness and confidence calibration across five LLMs and eight datasets\. While the pruned models often maintain high accuracy, we find that their faithfulness and calibration often degrade\. Notably, faithfulness and calibration can fluctuate significantly, even when accuracy remains stable, highlighting a misalignment between model confidence, interpretability, and accuracy\. Our findings suggest that layer pruning can affect LLMs' interpretability and reliability in ways not captured by accuracy and efficiency measures alone\. We recommend including explainability and calibration metrics when evaluating pruned models\.

## Submission history

From: Pietro Tropeano \[[view email](https://arxiv.org/show-email/87a97d7a/2606.24970)\] **\[v1\]**Tue, 23 Jun 2026 11:07:59 UTC \(2,526 KB\)

Similar Articles