Elastic Transformers: Training-Free Pruning at Any Compute Budget

University of Amsterdam's VISLab proposes an elastic Transformer method that learns a global importance ranking over MLP neurons and attention heads with frozen weights, enabling training-free pruning to fit any compute budget at inference time.

2026-10-15 ~ 2026-10-15 · 2 related posts

• Retraining-Free Method Turns Transformers Into Elastic Models Adapting to Any Compute Budget — VISLab-Amsterdam · 2026-10-15
• One Ranking, Any Budget: Elastic Transformer Prunes to 60% Sparsity Without Retraining — cgmsnoek · 2026-10-15