Research map / Architectures and efficiency / Pruning and efficient tuning
LeSTD: LLM Compression via Learning-based Sparse Tensor Decomposition
Yi Li, Zhichun Guo, Miao Yin, Bingzhe Li
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 |
|---|---|---|---|
| use LeSTD instead of SVD-LLM at aggressive ratio 0.2 | OPT-30B compression · WikiText-2 | perplexity: 50.25 -> 42.70 | Improved |
| use LeSTD (sparse core) instead of TensorLLM (dense core) at ratio 0.2 | GPT-J compression · WikiText-2 | perplexity: 89.52 -> 80.37 | Improved |
| add closed-form refitting after each pruning step (ablation) | LeSTD · WikiText-2 GPT-J ratio 0.6 | perplexity: 11.25 -> 9.57 | Improved |
| add closed-form refitting after each pruning step (ablation) | LeSTD · WikiText-2 GPT-J ratio 0.6 | tokens/sec: 11932.45 -> 11925.90 | No clear effect |
| increase Tucker ranks R1 or R2 (ablation) | LeSTD · WikiText-2 GPT-J ~0.6 ratio | perplexity: 13.72 -> 9.26 | Improved |
| raise per-iteration pruning rate alpha from 0.30 to 0.50 (ablation) | LeSTD · WikiText-2 GPT-J ratio 0.6 | perplexity: 9.49-9.54 -> 9.82 | Hurt |
| use LeSTD instead of TensorLLM for throughput | GPT-J compression · WikiText-2 ratio 0.6 | tokens/sec: 8572.11 -> 12025.90 | Improved |
| use LeSTD instead of SVD-LLM at ratio 0.8 | GPT-J compression · WikiText-2 | tokens/sec: 11521.82 -> 11571.82 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SVD-LLM — SVD-LLM: Truncation-aware Singular Value Decomposition for Large Language Model Compression (ICLR 2025) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- LowRankArena: A Standardized Evaluation Platform for SVD-Based LLM Compression (NeurIPS 2026 · similarity 0.90)
- SoLA: Leveraging Soft Activation Sparsity and Low-Rank Decomposition for Large Language Model Compression (AAAI 2025 · similarity 0.89)
- Compressing LLMs: The Truth is Rarely Pure and Never Simple (ICLR 2024 · similarity 0.89)
- 3BASiL: An Algorithmic Framework for Sparse plus Low-Rank Compression of LLMs (NeurIPS 2025 · similarity 0.89)
- Basis Sharing: Cross-Layer Parameter Sharing for Large Language Model Compression (ICLR 2025 · similarity 0.89)
- MoDeGPT: Modular Decomposition for Large Language Model Compression (ICLR 2025 · 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.
