Research map / Architectures and efficiency / Pruning and efficient tuning
Entropy-Based Block Pruning for Efficient Large Language Models
Liangwei Yang, Yuhui Xu, Juntao Tan, Doyen Sahoo, silvio savarese, Caiming Xiong, Huan Wang, Shelby Heinecke
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| prune attention with entropy criterion instead of LLMDrop (16 blocks) | Llama3.1-8B · Avg 11 benchmarks, 16 blocks pruned | average accuracy: 0.4354 -> 0.4781 | Improved |
| prune attention with entropy criterion instead of LLMDrop (12 blocks) | Mistral-7B-v0.3 · Avg 11 benchmarks, 12 blocks pruned | average accuracy: 0.5476 -> 0.5757 | Improved |
| prune attention with entropy criterion instead of LLMDrop (4 blocks) | Llama3.1-8B · Avg 11 benchmarks, 4 blocks pruned | average accuracy: 0.6283 -> 0.6271 | No clear effect |
| entropy layer pruning instead of ShortGPT (8 blocks) | Llama3.1-8B · Avg 11 benchmarks, 8 blocks pruned | average accuracy: 0.3824 -> 0.5133 | Improved |
| entropy layer pruning instead of ShortGPT (16 blocks) | Llama3.1-8B · Avg 11 benchmarks, 16 blocks pruned | average accuracy: 0.3251 -> 0.3199 | No clear effect |
| entropy layer pruning instead of ShortGPT (12 blocks) | Mistral-7B-v0.3 · Avg 11 benchmarks, 12 blocks pruned | average accuracy: 0.3397 -> 0.3777 | Improved |
| prune attention blocks instead of whole layers with entropy criterion (8 blocks) | EntroDrop on Llama3.1-8B · Avg 11 benchmarks, 8 blocks pruned | average accuracy: 0.5133 -> 0.6265 | Improved |
| change calibration dataset (C4, Wikitext, Medicine, Law) (ablation) | EntroDrop · pruning 12 attention layers | average accuracy | No clear effect |
| use Bucket or KNN entropy estimation instead of Renyi (ablation) | EntroDrop · deleting 12 attention layers | average accuracy | Improved |
Closest papers by meaning
- An Efficient Pruner for Large Language Model with Theoretical Guarantee (ICML 2025 · similarity 0.89)
- Compact Language Models via Pruning and Knowledge Distillation (NeurIPS 2024 · similarity 0.89)
- Toward Adaptive Large Language Models Structured Pruning via Hybrid-grained Weight Importance Assessment (AAAI 2025 · similarity 0.89)
- SlimGPT: Layer-wise Structured Pruning for Large Language Models (NeurIPS 2024 · similarity 0.89)
- You Only Prune Once: Designing Calibration-Free Model Compression With Policy Learning (ICLR 2025 · similarity 0.89)
- AlphaPruning: Using Heavy-Tailed Self Regularization Theory for Improved Layer-wise Pruning of Large Language Models (NeurIPS 2024 · similarity 0.89)
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.
