# CIERA: Cross-Iteration Exponent Reuse for Lossless Allgather in Sharded MoE Training

arXiv: https://arxiv.org/abs/2609.04609 · HTML: https://mlsanity.com/paper/2609.04609/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Efficient MoE training and inference systems, one of 12 directions in Mixture-of-experts.
- Builds on 32 of its 47 references that are in the map.
- First posted on arXiv 4 Sep 2026.

## Placed in

- Mixture-of-experts / [Efficient MoE training and inference systems](https://mlsanity.com/moe/efficient-moe-training-and-inference-systems/)

## Builds on (most cited first, 30 of 32)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Character-level Convolutional Networks for Text Classification](https://mlsanity.com/paper/1509.01626/) (2015)
- [Qwen2 Technical Report](https://mlsanity.com/paper/2407.10671/) (2024)
- [Mixtral of Experts](https://mlsanity.com/paper/2401.04088/) (2024)
- [ZeRO: Memory Optimizations Toward Training Trillion Parameter Models](https://mlsanity.com/paper/1910.02054/) (2019)
- [DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model](https://mlsanity.com/paper/2405.04434/) (2024)
- [Adafactor: Adaptive Learning Rates with Sublinear Memory Cost](https://mlsanity.com/paper/1804.04235/) (2018)
- [DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models](https://mlsanity.com/paper/2401.06066/) (2024)
- [MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategies](https://mlsanity.com/paper/2404.06395/) (2024)
- [PyTorch FSDP: Experiences on Scaling Fully Sharded Data Parallel](https://mlsanity.com/paper/2304.11277/) (2023)
- [Deep Gradient Compression: Reducing the Communication Bandwidth for Distributed Training](https://mlsanity.com/paper/1712.01887/) (2017)
- [8-bit Optimizers via Block-wise Quantization](https://mlsanity.com/paper/2110.02861/) (2021)
- [OLMoE: Open Mixture-of-Experts Language Models](https://mlsanity.com/paper/2409.02060/) (2024)
- [Sparse Communication for Distributed Gradient Descent](https://mlsanity.com/paper/1704.05021/) (2017)
- [Comet: Fine-grained Computation-communication Overlapping for Mixture-of-Experts](https://mlsanity.com/paper/2502.19811/) (2025)
- [Torch.fx: Practical Program Capture and Transformation for Deep Learning in Python](https://mlsanity.com/paper/2112.08429/) (2021)
- [On the Utility of Gradient Compression in Distributed Training Systems](https://mlsanity.com/paper/2103.00543/) (2021)
- [Accordion: Adaptive Gradient Communication via Critical Learning Regime Identification](https://mlsanity.com/paper/2010.16248/) (2020)
- [Adaptive Compression for Communication-Efficient Distributed Training](https://mlsanity.com/paper/2211.00188/) (2022)
- [SDP4Bit: Toward 4-bit Communication Quantization in Sharded Data Parallelism for LLM Training](https://mlsanity.com/paper/2410.15526/) (2024)
- [NeuZip: Memory-Efficient Training and Inference with Dynamic Compression of Neural Networks](https://mlsanity.com/paper/2410.20650/) (2024)
- [GraVAC: Adaptive Compression for Communication-Efficient Distributed DL Training](https://mlsanity.com/paper/2305.12201/) (2023)
- [Efficient and Adaptable Overlapping for Computation and Communication via Signaling and Reordering](https://mlsanity.com/paper/2504.19519/) (2025)
- [ZipServ: Fast and Memory-Efficient LLM Inference with Hardware-Aware Lossless Compression](https://mlsanity.com/paper/2603.17435/) (2026)
- [Deep Optimizer States: Towards Scalable Training of Transformer Models Using Interleaved Offloading](https://mlsanity.com/paper/2410.21316/) (2024)
- [gZCCL: Compression-Accelerated Collective Communication Framework for GPU Clusters](https://mlsanity.com/paper/2308.05199/) (2023)
- [LSH-MoE: Communication-efficient MoE Training via Locality-Sensitive Hashing](https://mlsanity.com/paper/2411.08446/) (2024)
- [High Speed Hashing for Integers and Strings](https://mlsanity.com/paper/1504.06804/) (2015)
- [Kimad: Adaptive Gradient Compression with Bandwidth Awareness](https://mlsanity.com/paper/2312.08053/) (2023)
- [TAGC: Optimizing Gradient Communication in Distributed Transformer Training](https://mlsanity.com/paper/2504.05638/) (2025)

## Abstract (arXiv)

In training Mixture-of-Experts (MoE) models, sharded data parallelism partitions each expert's parameters across GPUs, requiring an Allgather operation to reconstruct the full weight matrix before each layer executes. This communication often dominates iteration time. Prior work often reduces this overhead using lossy compression methods that sacrifice numerical fidelity, while existing lossless methods do not exploit cross-iteration exponent stability. In this paper, we propose Cross-Iteration Exponent Reuse Allgather (CIERA), a lossless, system-aware communication method for sharded MoE training. We observe that after a brief warmup phase, the exponent values of most weights remain unchanged across iterations. Based on this, we cache exponents locally and transmit only the sign and mantissa when exponents are unchanged. The receiver reconstructs the original weights exactly by combining the cached exponents with the received data. Since parameter matrices vary in shape across layers, compression is applied only when it yields net time savings. Moreover, the compression operation is overlapped with both Allgather communication and computation. Our real experiments and large-scale trace-driven simulator show that on OLMoE-1B-7B at 16 GPUs, CIERA achieves a 3.70x speedup over the lossless baseline and 3.68x over the lossy baseline, projected to reach 4.28x and 4.42x respectively at 128 GPUs, while preserving bitwise-exact parameter reconstruction in all evaluated runs.