# MaskCoFT: Masked Co-Adaptive Fine-Tuning for Memory-Efficient MoE Inference

arXiv: https://arxiv.org/abs/2609.34077 · HTML: https://mlsanity.com/paper/2609.34077/

## 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 25 of its 32 references that are in the map.
- First posted on arXiv 28 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, 25 of 25)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity](https://mlsanity.com/paper/2101.03961/) (2021)
- [Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation](https://mlsanity.com/paper/1308.3432/) (2013)
- [Mixtral of Experts](https://mlsanity.com/paper/2401.04088/) (2024)
- [GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding](https://mlsanity.com/paper/2006.16668/) (2020)
- [DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model](https://mlsanity.com/paper/2405.04434/) (2024)
- [Not All Experts are Equal: Efficient Expert Pruning and Skipping for Mixture-of-Experts Large Language Models](https://mlsanity.com/paper/2402.14800/) (2024)
- [Pre-gated MoE: An Algorithm-System Co-Design for Fast and Scalable Mixture-of-Expert Inference](https://mlsanity.com/paper/2308.12066/) (2023)
- [Fast Inference of Mixture-of-Experts Language Models with Offloading](https://mlsanity.com/paper/2312.17238/) (2023)
- [Fiddler: CPU-GPU Orchestration for Fast Inference of Mixture-of-Experts Models](https://mlsanity.com/paper/2402.07033/) (2024)
- [MoE-Infinity: Efficient MoE Inference on Personal Machines with Sparsity-Aware Expert Cache](https://mlsanity.com/paper/2401.14361/) (2024)
- [SwapMoE: Serving Off-the-shelf MoE-based Large Language Models with Tunable Memory Budget](https://mlsanity.com/paper/2308.15030/) (2023)
- [Taming Latency-Memory Trade-Off in MoE-Based LLM Serving via Fine-Grained Expert Offloading](https://mlsanity.com/paper/2502.05370/) (2025)
- [Let the Expert Stick to His Last: Expert-Specialized Fine-Tuning for Sparse Architectural Large Language Models](https://mlsanity.com/paper/2407.01906/) (2024)
- [ExpertFlow: Efficient Mixture-of-Experts Inference via Predictive Expert Caching and Token Scheduling](https://mlsanity.com/paper/2410.17954/) (2024)
- [Mixture of Cache-Conditional Experts for Efficient Mobile Device Inference](https://mlsanity.com/paper/2412.00099/) (2024)
- [DAOP: Data-Aware Offloading and Predictive Pre-Calculation for Efficient MoE Inference](https://mlsanity.com/paper/2501.10375/) (2024)
- [MoNDE: Mixture of Near-Data Experts for Large-Scale Sparse Models](https://mlsanity.com/paper/2405.18832/) (2024)
- [MaskConnect: Connectivity Learning by Gradient Descent](https://mlsanity.com/paper/1807.11473/) (2018)
- [Not All Models Suit Expert Offloading: On Local Routing Consistency of Mixture-of-Expert Models](https://mlsanity.com/paper/2505.16056/) (2025)
- [ReMoE: Boosting Expert Reuse through Router Fine-Tuning in Memory-Constrained MoE LLM Inference](https://mlsanity.com/paper/2605.27081/) (2026)
- [Cache-Aware Joint Router Adaptation for Memory-Efficient MoE Inference](https://mlsanity.com/paper/2609.04895/) (2026)
- [Context-Aware Mixture-of-Experts Inference on CXL-Enabled GPU-NDP Systems](https://mlsanity.com/paper/2512.04476/) (2025)
- [Sticky Routing: Training MoE Models for Memory-Efficient Inference](https://mlsanity.com/paper/2607.08780/) (2026)
- [BEAM: Binary Expert Activation Masking for Dynamic Routing in MoE](https://mlsanity.com/paper/2605.14438/) (2026)

## Abstract (arXiv)

Mixture-of-experts (MoE) language models often exceed the memory of a single GPU. Expert offloading keeps most experts in host memory and loads them on demand, so decoding speed depends on how many experts each token must fetch. Caching and prefetching reduce this cost only as far as the routing allows. Router-only fine-tuning can reshape the routing to reuse experts, but it keeps the experts frozen, so they cannot adapt to the tokens the new routing sends them. We propose MaskCoFT, a masked co-adaptive fine-tuning method that trains routers and experts together with the cross-entropy loss alone. During fine-tuning, a learnable binary mask restricts the Top-K routing of each layer to a subset of experts, and the experts adapt to the tokens redirected to them. At inference, the learned mask becomes a soft prior that re-ranks experts, so every expert remains selectable. We simulate a GPU cache of 4 experts per layer for Mixtral-8x7B and 12 for DeepSeek-V2-Lite. MaskCoFT cuts expert fetches per token by 23.7% and 10.1% relative to the base model. In real offloading system serving, it lowers the time per output token by up to 16.4% and 5.5%, respectively. Its average accuracy over nine benchmarks stays above the base model by 0.92 and 0.53 points.