# EAServe: Encode-Aware Disaggregated Serving for Multimodal Large Language Models

arXiv: https://arxiv.org/abs/2609.31551 · HTML: https://mlsanity.com/paper/2609.31551/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Visual token pruning and efficiency, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 10 of its 54 references that are in the map.
- First posted on arXiv 25 Sep 2026.

## Placed in

- Multimodal LLMs / vision-language models / [Visual token pruning and efficiency](https://mlsanity.com/mllm/visual-token-pruning-and-efficiency/)
- Efficient LLM inference and serving / [Serving systems and request scheduling](https://mlsanity.com/inference/serving-systems-and-request-scheduling/)

## Builds on (most cited first, 10 of 10)

- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [LLaVA-OneVision: Easy Visual Task Transfer](https://mlsanity.com/paper/2408.03326/) (2024)
- [Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond](https://mlsanity.com/paper/2308.12966/) (2023)
- [Gemma 2: Improving Open Language Models at a Practical Size](https://mlsanity.com/paper/2408.00118/) (2024)
- [Yi: Open Foundation Models by 01.AI](https://mlsanity.com/paper/2403.04652/) (2024)
- [Fast Distributed Inference Serving for Large Language Models](https://mlsanity.com/paper/2305.05920/) (2023)
- [semi-PD: Towards Efficient LLM Serving via Phase-Wise Disaggregated Computation and Unified Storage](https://mlsanity.com/paper/2504.19867/) (2025)
- [Nexus:Proactive Intra-GPU Disaggregation of Prefill and Decode in LLM Serving](https://mlsanity.com/paper/2507.06608/) (2025)
- [HydraInfer: Hybrid Disaggregated Scheduling for Multimodal Large Language Model Serving](https://mlsanity.com/paper/2505.12658/) (2025)
- [Niyama : Breaking the Silos of LLM Inference Serving](https://mlsanity.com/paper/2503.22562/) (2025)

## Abstract (arXiv)

Disaggregating the two stages, Prefill and Decode, onto separate GPU pools is now a standard optimization for (text-only) LLM serving. However, multimodal LLMs (MLLMs), which add a third phase, Encode, pose new challenges for resource allocation. Encode turns images, video, or audio into embeddings that the language model can consume, yielding a three-stage Encode-Prefill-Decode (EPD) pipeline. Existing frameworks offer only partial answers: text-only PD systems lack Encode, while EPD frameworks expose it as a separate service without regulating downstream request flow. The pipeline also carries a structural resource imbalance: every request enters through Encode before downstream work can begin, yet per-request execution leaves the encode GPU severely underutilized even at high loads, starving the downstream Prefill and Decode workers. Addressing this, we reposition Encode as the control point of the EPD pipeline, exposing three tightly coupled dimensions: when work enters downstream, where prefill executes, and how the GPU is shared. We instantiate this in EAServe across two co-designed layers. Its runtime manages load-adaptive micro-batching, rate-controlled partial offload to a co-resident prefill worker, and dynamic SM partitioning for predictable co-location. The configuration layer, Hybrid Auto Selection (HAS), navigates the joint space of GPU allocation, encode batch size, and offload ratio by pruning unbalanced allocations with per-stage capacity profiling and refining the remainder through TPE-based Bayesian optimization. Evaluated on three MLLM architectures spanning image, video, and audio, EAServe delivers up to 4.3x and 1.7x higher goodput than NVIDIA Dynamo and vLLM, respectively, under identical SLO constraints, sustains more balanced and higher GPU utilization across the EPD pipeline, and reaches near-optimal configurations faster than baseline search methods.