analiza aiqbench
Sparsity Evolution Fine-Tuning (SEFT) addresses the critical challenge of adapting sparse LLMs to downstream tasks without compromising their efficiency gains. By dynamically reallocating updates and reactivating pruned weights based on parameter importance, the method outperforms existing baselines in both accuracy and resource utilization across major model architectures. This approach offers a practical solution for maintaining the deployment advantages of sparsity while enabling effective task-specific adaptation.
Leave it to the Specialist: Repair Sparse LLMs with Sparse Fine-Tuning via Sparsity Evolution
arXiv:2505.24037v3 Announce Type: replace Abstract: Sparse large language models (LLMs) offer an attractive direction toward efficient deployment, but adapting them to downstream tasks remains challenging. The central difficulty is to enable effective task adaptation without sacrificing the efficiency advantages of sparsity. Existing fine-tuning methods are not well-suited to this setting, as they either introduce additional dense parameters or assume a fixed sparse topology, limiting their compatibility with sparse LLMs. In this paper, we propose Sparsity Evolution Fine-Tuning (SEFT), a fine-tuning framework designed specifically for sparse LLMs. SEFT allows sparse structure to evolve during fine-tuning by periodically reallocating sparse task-specific updates and reactivating previously pruned weights when beneficial. At the same time, SEFT preserves the efficiency advantages of sparsity through topology adaptation based on parameter importance. Experiments on LLaMA, DeepSeek, and Mistral models across multiple benchmarks show that SEFT delivers stronger performance while offering superior memory and time efficiency compared to existing baselines. Our code is publicly available at: https://github.com/QiaoXiao7282/SEFT.
Abstract:Sparse large language models (LLMs) offer an attractive direction toward efficient deployment, but adapting them to downstream tasks remains challenging. The central difficulty is to enable effective task adaptation without sacrificing the efficiency advantages of sparsity. Existing fine-tuning methods are not well-suited to this setting, as they either introduce additional dense parameters or assume a fixed sparse topology, limiting their compatibility with sparse LLMs. In this paper, we propose Sparsity Evolution Fine-Tuning (SEFT), a fine-tuning framework designed specifically for sparse LLMs. SEFT allows sparse structure to evolve during fine-tuning by periodically reallocating sparse task-specific updates and reactivating previously pruned weights when beneficial. At the same time, SEFT preserves the efficiency advantages of sparsity through topology adaptation based on parameter importance. Experiments on LLaMA, DeepSeek, and Mistral models across multiple benchmarks show that SEFT delivers stronger performance while offering superior memory and time efficiency compared to existing baselines. Our code is publicly available at: this https URL.
Artificial Intelligence (cs.AI)
arXiv:2505.24037 [cs.AI]
(or arXiv:2505.24037v3 [cs.AI] for this version)
Text integral sub licența deschisă a publicației · 167 cuvinte
Subiecte conexe
RESCUE: Repairing Language Model Errors to Sparse Circuits via Reinforcement Learning
arXiv · 30 sept. 2026
Self-Modeling Interventions Modulate Emergent Misalignment
LessWrong · 5 oct. 2026
[AINews] High Taste Testers Ep. 1
Latent Space · 17 oct. 2026
The Model Organism Lottery: Model Organism Interpretability Strongly Depends on Training Methodology
LessWrong · 23 iul. 2026