Flash-Unified: A Training-Free and Task-Aware Acceleration Framework for Native Unified Models

Junlong Ke; Zichen Wen; Boxue Yang; Yantai Yang; Xuyang Liu; Chenfei Liao; Zhaorun Chen; Shaobo Wang

cs.CVMar 16, 2026

Flash-Unified: A Training-Free and Task-Aware Acceleration Framework for Native Unified Models

Authors: Junlong Ke, Zichen Wen, Boxue Yang, Yantai Yang, Xuyang Liu, Chenfei Liao, Zhaorun Chen, Shaobo Wang,

Organizations: Tsinghua University · Shanghai Jiao Tong University · Shanghai AI Laboratory · Sichuan University ·

Abstract

Native unified multimodal models, which integrate both generative and understanding capabilities, face substantial computational overhead that hinders their real-world deployment. Existing acceleration techniques typically employ a static, monolithic strategy, ignoring the fundamental divergence in computational profiles between iterative generation tasks (e.g., image generation) and single-pass understanding tasks (e.g., VQA). In this work, we present the first systematic analysis of unified models, revealing pronounced parameter specialization, where distinct neuron sets are critical for each task. This implies that, at the parameter level, unified models have implicitly internalized separate inference pathways for generation and understanding within a single architecture. Based on these insights, we introduce a training-free and task-aware acceleration framework, FlashU, that tailors optimization to each task's demands. Across both tasks, we introduce Task-Specific Network Pruning and Dynamic Layer Skipping, aiming to eliminate inter-layer and task-specific redundancy. For visual generation, we implement a time-varying control signal for the guidance scale and a temporal approximation for the diffusion head via Diffusion Head Cache. For multimodal understanding, building upon the pruned model, we introduce Dynamic Token Pruning via a V-Norm Proxy to exploit the spatial redundancy of visual inputs. Extensive experiments on Show-o2 demonstrate that FlashU achieves 1.78 $\times$ to 2.01 $\times$ inference acceleration across both understanding and generation tasks while maintaining SOTA performance, outperforming competing unified models and validating our task-aware acceleration paradigm. Our code is publicly available at https://github.com/Rirayh/FlashU.

Figures & tables

Figure 1: Task-specific FFN neuron analysis for generation and understanding. Each layer shows the fraction of neurons that are specific to generation, understanding, or shared between both tasks (excluding unimportant neurons). Percentages indicate the proportion of neurons in each layer ranked among the top 50% for task importance, as measured by OBD-inspired sensitivity scores $\Delta_{i}$ .

Figure 2: Understanding Redundancy. Left: Intra-layer cosine similarity between the input and output features of each layer for newly generated tokens in understanding tasks. Right: Inter-layer heatmap of the cosine similarity among layer outputs for newly generated tokens. All measurements were conducted on samples from the MME [ 19 ] .

Figure 3: Illustration of the FlashU acceleration framework. (a) Task-Specific FFN Pruning calculates an importance score $I_{j}$ based on activation norms and weight magnitude to statically mask redundant neurons for a specific task. (b) Dynamic Token Pruning utilizes the V-Norm proxy at shallow layers (the 2nd layer) to identify and prune visually spatially redundant tokens. (c) Dynamic Layer Skipping bypasses layers exhibiting high input-output cosine similarity ( $S_{i}$ ), re-evaluating the skipping mask every $T_{LS}$ steps. (d) Diffusion Head Cache exploits temporal coherence in the generative process, computing the full diffusion head only at fixed intervals and reusing cached hidden states for intermediate steps.

Figure 4: Redundancy in Generation Tasks. Left: Intra-layer cosine similarity between the input and output of each layer. Right: Inter-layer heatmap of hidden state cosine similarity. All measurements were conducted on samples from the GenEval [ 21 ] .

Figure 5: The relation between attention score and the norm of the value matrix for each token. Tokens with the largest attention score tend to have a lower norm in the value matrix.

Models	# Params.	MME $\uparrow$	GQA $\uparrow$	MMB $\uparrow$	MMMU $\uparrow$	MMStar $\uparrow$	AI2D $\uparrow$	Latency (s) $\downarrow$
Models	# Params.	(p)	GQA $\uparrow$	(en)	(val)	MMStar $\uparrow$	AI2D $\uparrow$	Latency (s) $\downarrow$
MetaMorph [ 58 ]	8B	-	-	75.2	-	-	-	-
ILLUME [ 60 ]	7B	1445.3	-	75.1	38.2	-	71.4	-
Show-o [ 71 ]	1.3B	1097.2	58.0	-	27.4	-	-	-
JanusFlow [ 46 ]	1.5B	1333.1	60.3	74.9	29.3	-	-	-
SynerGen-VL [ 33 ]	2.4B	1381.0	-	53.7	34.2	-	-	-

Table 1: Evaluation on multimodal understanding benchmarks. # Params. indicates the number of parameters of the base LLM. Results in gray indicate the vanilla Show-o2 performance. Latency is measured in seconds (s), conducted on samples from MMMU.

Method	# Params.	Single Obj.	Two Obj.	Counting	Colors	Position	Color Attri.	Overall $\uparrow$	Latency (s) $\downarrow$
SEED-X [ 20 ]	17B	0.97	0.58	0.26	0.80	0.19	0.14	0.49	-
TokenFlow-XL [ 48 ]	14B	0.95	0.60	0.41	0.81	0.16	0.24	0.55	-
ILLUME [ 60 ]	7B	0.99	0.86	0.45	0.71	0.39	0.28	0.61	-
Show-o [ 71 ]	1.3B	0.98	0.80	0.66	0.84	0.31	0.50	0.68	-
MUSE-VL [ 73 ]	7B	-	-	-	-	-	-	0.57	-
Transfusion [ 83 ]	3.5B	-	-	-	-	-	-	0.63	-

Table 2: Evaluation on the GenEval [ 21 ] . # Params. indicates the number of parameters of the base LLM. Results in gray indicate the vanilla Show-o2 performance. The subscript values show the speedup achieved by FlashU. Latency is measured in seconds (s).

Figure 6: Comparison of four unified multimodal models on generating a three-panel canvas with seamless cross-panel continuity. While baseline models fail to satisfy the structural or continuity requirements, our FlashU 1.5B (from Show-o2 [ 72 ] ) accurately produces all three aligned segments with the fastest inference time.

Method	# Params.	Global	Entity	Attribute	Relation	Overall $\uparrow$	Latency (s) $\downarrow$
Hunyuan-DiT [ 34 ]	1.5B	84.59	80.59	88.01	74.36	78.87	-
Playground v2.5 [ 32 ]	-	83.06	82.59	81.20	84.08	75.47	-
PixArt- $\Sigma$ [ 5 ]	-	86.89	82.89	88.94	86.59	80.54	-
DALL-E 3 [ 2 ]	-	90.97	89.61	88.39	90.58	83.50	-
SD3-Medium [ 18 ]	2B	87.90	91.01	88.83	80.70	84.08	-
Emu3-DPO [ 62 ]	8B	-	-	-	-	81.60	124.8

Table 3: Evaluation on DPG-Bench [ 23 ] . # Params. denotes the parameter count of the base LLM. Results in gray correspond to vanilla Show-o2. Subscripts indicate the speedup of FlashU. Latency is reported in seconds (s), averaged per sample.

Strategy	Latency (s)	Score
Show-o2 [ 72 ]	11.30	86.14
FlashU (Ours)	5.89 +1.92 $\times$	84.39
w/o Dynamic Layer Pruning	6.70 +1.69 $\times$	85.56
w/o Hybrid Network	5.65 +2.00 $\times$	83.68
w/o Adaptive Guidance	8.86 +1.28 $\times$	85.34
FFN Pruning Only	9.21 +1.23 $\times$	85.08

Table 4: Ablation study on the 7B model of Show-o2 using DPG-Bench [ 23 ] . The subscript values show the speedup relative to Show-o2 baseline. Latency is measured in seconds (s).

Strategy	Latency (s)	Score
Show-o2 [ 72 ]	5.23	85.02
FlashU (Ours)	2.80 +1.87 $\times$	84.12
w/o Dynamic Layer Pruning	2.91 +1.80 $\times$	85.33
w/o Hybrid Network	2.48 +2.11 $\times$	83.57
w/o Adaptive Guidance	3.92 +1.33 $\times$	83.42
FFN Pruning Only	3.84 +1.36 $\times$	84.27

Table 5: Ablation study on the 1.5B model of Show-o2 using DPG-Bench [ 23 ] . The subscript values show the speedup relative to Show-o2 baseline. Latency is measured in seconds (s).

Method	# Params.	Global	Entity	Attribute	Relation	Overall $\uparrow$	Latency (s) $\downarrow$
Hunyuan-DiT [ 34 ]	1.5B	84.59	80.59	88.01	74.36	78.87	20.1
Playground v2.5 [ 32 ]	-	83.06	82.59	81.20	84.08	75.47	7.67
TokenFlow-t2i [ 48 ]	7B	72.64	74.72	77.67	83.80	67.44	0.97
JanusFlow [ 46 ]	1.5B	81.16	85.16	85.32	91.05	77.57	1.09
VILA-U [ 70 ]	7B	89.92	82.73	80.80	87.83	74.26	9.78
Liquid [ 69 ]	8B	79.64	87.61	85.24	91.01	80.26	37.05

Table 6: Latency Comparison on DPG-Bench [ 23 ] . Results in gray correspond to vanilla Show-o2. Subscripts indicate the speedup of FlashU. Latency is reported in seconds (s), averaged per sample.

Appendix figures & tables8 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 7: Latency vs. Performance Trade-off. Comparison of FlashU against native unified models (Show-o2, VILA-U, Emu3-DPO, Liquid) and tailored models (JanusFlow, TokenFlow-t2i). The x-axis represents inference latency (lower is better), and the y-axis represents overall performance on DPG-Bench(higher is better). Circle sizes denote model parameter counts (1.5B, 7B, 8B).

Hyperparameter	Symbol	Value
Total Inference Steps	$T$	$0.7T_{0}$
Hybrid Threshold	$\tau$	0.3
FFN Pruning Ratio	$r_{p}$	0.20
Layer Skipping Ratio	$r_{LS}$	0.20
Recalculation Interval	$T_{LS}$	10
Diffusion Head Cache Interval	$\mathcal{T}_{cache}$	5

Appendix

Table 7: Hyperparameters for Generation Tasks.

Hyperparameter	Symbol	Value
Max Generated Tokens	$N_{max}$	128
Hybrid Threshold	$\tau$	0.3
FFN Pruning Ratio	$r_{p}$	0.10
Layer Skipping Ratio	$r_{LS}$	0.10
Calibration Samples	$N_{calib}$	20
Recalculation Interval	$T_{LS}$	10

Appendix

Table 8: Hyperparameters for Multimodal Understanding Tasks.

Figure 8: Sensitivity Analysis of FFN Pruning Ratio. The trade-off between generation quality (DPG-Bench score) and inference speedup is shown as the FFN pruning ratio (

r_{p}

) varies from 0.1 to 0.5.

Figure 9: Sensitivity Analysis of Layer Skipping Ratio. The relationship between generation quality and speedup across different layer skipping ratios (

r_{LS}

Figure 10: Comparative Analysis of Hybrid Threshold Strategies. The efficiency-quality trade-off curves for Adaptive Guidance versus Baseline strategies across different hybrid threshold (

\tau

) values.

Figure 11: Component-wise Latency Profiling Analysis. Inference time breakdown by component for Show o2 and FlashU on the 1.5B model.

Figure 12: Qualitative results of text-to-image generation. We showcase samples generated by FlashU across diverse prompts. The model successfully handles complex descriptions, rendering high-quality textures and correct object placements, demonstrating that our acceleration framework preserves the generative performance of the backbone model.

Efficient Multimodal Inference

Efficient multimodal inference reduces the compute, memory, latency, and energy needed to process image, video, audio, text, and sensor inputs.

See all 227 papers

Accelerating Unified Multimodal Models with Core-Expansion Routing and Unified Computation Scheduling
Aug 29, 2026Wengyi Zhan, Chenqian Yan, Songwei Liu +2Efficient Multimodal Inference Unified Multimodal Models
1Xiamen University · 2ByteDance
Lance: Unified Multimodal Modeling by Multi-Task Synergy
May 18, 2026Fengyi Fu, Mengqi Huang, Shaojin Wu +10Unified Multimodal Models Multi-Task Learning
Intelligent Creation Lab, ByteDance
UniCompress: Token Compression for Unified Vision-Language Understanding and Generation
Mar 11, 2026Ziyao Wang, Chen Chen, Jingtao Li +4Multimodal Token Compression Visual Token Compression
Sony AI · University of Maryland, College Park

Aug 29, 2026J/K move · Enter open · S save

Accelerating Unified Multimodal Models with Core-Expansion Routing and Unified Computation Scheduling

Wengyi Zhan, Chenqian Yan, Songwei Liu, Mingbao Lin+1

1Xiamen University · 2ByteDance

Efficient Multimodal Inference Unified Multimodal Models Multimodal Token Compression Diffusion Model Inference Acceleration Multimodal Model Routing

Open paper AskPDF arXiv

Unified multimodal models jointly support understanding and generation, but incur substantial redundant computation across tokens, layers, and generation timesteps. Through token-importance probing, we identify an asymmetric core-expansion structure: understanding exhibits a stable importance component, while generation largely shares this component but requires progress-dependent corrections. We therefore propose CE-Router, which uses a task-shared core scorer and progress-conditioned generation expansions, optimized through generation decomposition and cross-task core alignment. At inference, CE-Router compacts token computation and supplies a learned routing signal to Unified Computation Scheduling, which coordinates layer skipping, FFN pruning, diffusion-head cache reuse, and denoising-step early exit. Experiments on two representative UMM architectures demonstrate consistent quality--efficiency improvements across both tasks, retaining 98.03% of dense understanding performance with a 1.93

\times

end-to-end inference speedup.

Abstract

Efficient Multimodal Inference

Explore similar work

Accelerating Unified Multimodal Models with Core-Expansion Routing and Unified Computation Scheduling

Lance: Unified Multimodal Modeling by Multi-Task Synergy

UniCompress: Token Compression for Unified Vision-Language Understanding and Generation