aiqbench analysis
This research introduces SparseLoCo, a novel training algorithm that resolves the communication-performance bottleneck in distributed LLM pretraining by combining Top-k sparsification with extreme quantization. The method achieves up to 97-99% compression of pseudo-gradients while outperforming dense baselines like DiLoCo across various model scales and MoE architectures, making cross-datacenter training significantly more efficient without sacrificing accuracy.
Overcoming the Communication-Performance Tradeoff in LLM Pretraining
arXiv:2508.15706v3 Announce Type: replace Abstract: Communication-efficient distributed training algorithms (e.g., DiLoCo) have received considerable interest due to their benefits for training large language models (LLMs) in bandwidth-constrained settings, such as across datacenters and over the internet. While these local-update methods achieve communication reduction through reduced synchronization, they still require communicating dense model-sized pseudo-gradients, resulting in a communication bottleneck even for cross-datacenter links. While quantization is often applied to reduce the pseudo-gradient's size, in the context of LLM pre-training, existing approaches have not been able to leverage sparsification without incurring significant performance degradation. In this work, we introduce SparseLoCo, a communication-efficient training algorithm for LLMs that can effectively leverage Top-k sparsification and 2-bit quantization to reach extreme sparsity in the communicated pseudo-gradient, as high as 97-99%, while achieving lower final loss than dense DiLoCo. In our empirical study of language model pre-training, we demonstrate that SparseLoCo's improvements over DiLoCo in performance and pseudo-gradient compression are maintained across dense model scales (178M-2B), an MoE transformer (645M-A273M), increasing the number of workers, and increasing communication intervals.
Abstract:Communication-efficient distributed training algorithms (e.g., DiLoCo) have received considerable interest due to their benefits for training large language models (LLMs) in bandwidth-constrained settings, such as across datacenters and over the internet. While these local-update methods achieve communication reduction through reduced synchronization, they still require communicating dense model-sized pseudo-gradients, resulting in a communication bottleneck even for cross-datacenter links. While quantization is often applied to reduce the pseudo-gradient's size, in the context of LLM pre-training, existing approaches have not been able to leverage sparsification without incurring significant performance degradation. In this work, we introduce SparseLoCo, a communication-efficient training algorithm for LLMs that can effectively leverage Top-k sparsification and 2-bit quantization to reach extreme sparsity in the communicated pseudo-gradient, as high as 97-99%, while achieving lower final loss than dense DiLoCo. In our empirical study of language model pre-training, we demonstrate that SparseLoCo's improvements over DiLoCo in performance and pseudo-gradient compression are maintained across dense model scales (178M-2B), an MoE transformer (645M-A273M), increasing the number of workers, and increasing communication intervals.
Machine Learning (cs.LG)
arXiv:2508.15706 [cs.LG]
(or arXiv:2508.15706v3 [cs.LG] for this version)
Full text under the publisher's open licence · 179 words
Related coverage
Research Note: Filtering Subversion-Relevant Information From Pretraining Data Is Feasible
LessWrong · Oct 5, 2026
Latent reasoning architectures would likely undermine CoT, our strongest oversight tool
AI Alignment Forum · Sep 23, 2026
Problems of Proliferation
LessWrong · Sep 29, 2026
[AINews] Claude Opus 5.5, the new default model for AINews — and everybody cuts prices 40-50%
Latent Space · Sep 23, 2026