Skip to content

Research1 min read

Damage-Aware Bandit Pruning for Transformer Models

Research presents a damage-aware multi-armed bandit approach for structured post-training pruning of transformer models. Experiments across GPT-2, OPT, ViT-B/16, and Swin-Tiny show improved performance compared to other pruning methods, with statistically significant gains.

By OpenSmartRoute editorial · written through the router by writer-small

From arXiv cs.AI - “Damage-Aware Bandit Pruning for Vision and Language Transformers

The research introduces a method for pruning transformer models by framing it as a damage-aware multi-armed bandit problem. Attention heads and MLP channel groups are temporarily masked during calibration, measuring the resulting loss degradation. This paired damage metric is used to guide the selection process, employing either a UCB-style policy or fractional-Beta Thompson Sampling. The final mask is constructed sequentially, adding units based on the reward signal. The selected units are functionally zeroed in the original checkpoint, representing structural suppression rather than physical compression.

Experiments were conducted on WikiText-2, LAMBADA, and Imagenette using models including GPT-2, OPT, Pythia, Qwen2.5, SmolLM2, ViT-B/16, DeiT-Tiny, and Swin-Tiny. Comparisons were made against random, magnitude, static-saliency, and budgeted-greedy selection. Across five seeds, the bandit methods generally reduced degradation relative to budgeted greedy in language model comparisons.

Statistical analysis of 28 comparisons revealed significant gains. 23 bootstrap confidence intervals excluded zero, and 11 paired tests had p < 0.05. Matched-evaluation results for ViT-B/16 and Swin-Tiny indicated that the gains were not solely due to a larger candidate-evaluation budget.

This approach offers a structured method for pruning transformers, potentially reducing model size while maintaining performance. The use of a bandit algorithm allows for adaptive selection based on observed damage, leading to more effective pruning than traditional methods.

Source: https://arxiv.org/abs/2609.05448

Published Sep 9, 2026 · 212 words

Keep reading

Related posts

More in Research