---
title: Multi-Scale Mamba (MS-Mamba) Research
url: https://www.emergentmind.com/topics/multi-scale-mamba-ms-mamba-1061e75b-66e4-4fcc-8b98-f85e2bf826e0
type: topic
---

# Multi-Scale Mamba (MS-Mamba) Research

The literature suggests that **Multi-Scale Mamba (MS-Mamba)** denotes a family of Mamba- or selective state-space-model architectures that introduce explicit multi-scale processing into otherwise single-scale sequence models. Across vision, remote sensing, time-series forecasting, reinforcement learning, recommendation, and temporal action detection, these systems retain the linear-complexity selective scan paradigm of Mamba while altering how inputs are partitioned, sampled, scanned, fused, or hierarchically organized. The common objective is to couple long-range dependency modeling with scale diversity—spatial, temporal, spectral, frequency-domain, or modality-level—without reverting to quadratic self-attention [2504.07654] [2405.14174].

## 1. Conceptual scope and terminology

Taken together, the published uses of the term indicate that **“multi-scale” is not standardized in Mamba research**. In some works it means multiple temporal sampling rates, in others full- and low-resolution spatial scans, in others a U-Net pyramid, dilated state-space transitions, or frequency-domain decomposition. The term therefore names a design pattern rather than a single canonical block.

| Instantiation | Domain | Multi-scale realization |
|---|---|---|
| ms-Mamba [2504.07654] | Time-series forecasting | Parallel Mamba blocks with different sampling rates $\Delta_i$ |
| MSVMamba [2405.14174] | Vision backbone | One full-resolution scan plus three downsampled scans |
| MSFMamba [2408.14255] | Multi-source remote sensing | Full- and half-resolution spatial SSM routes inside MSpa-Mamba |
| M3SR [2601.08293] | Spectral reconstruction | U-Net structure with global, intermediate, and local scales |
| MS-Temba [2501.06138] | Temporal action detection | Block-wise dilated temporal convolution and dilated SSM |
| M2Rec [2505.04445] | Sequential recommendation | Temporal Mamba, FFT-based frequency modeling, and semantic fusion |

Related medical-imaging variants reinforce this plurality. MSV-Mamba introduces a decoder-side large-window multiscale Mamba module with pixel-level and patch-level branches [2501.07120]. MSVM-UNet inserts multi-scale convolutions into VSS blocks and combines them with large-kernel patch expanding [2408.13735]. A 3D volumetric study defines a multi-scale Mamba block, denoted MSv4, around parallel $3\times3\times3$, $5\times5\times5$, and $7\times7\times7$ depthwise convolutions followed by Mamba scanning [2503.19308].

## 2. State-space foundations and how scale is introduced

Most MS-Mamba variants preserve the standard state-space backbone. A continuous-time SSM is typically written as
$$
h'(t)=\mathbf{A}h(t)+\mathbf{B}x(t), \qquad y(t)=\mathbf{C}h(t),
$$
with discretization by a step size $\Delta$ yielding
$$
h_t=\overline{\mathbf{A}}h_{t-1}+\overline{\mathbf{B}}x_t,\qquad y_t=\mathbf{C}h_t.
$$
Mamba then makes key quantities such as $\mathbf{B}$, $\mathbf{C}$, and $\Delta$ input-dependent, producing a selective SSM with linear complexity in sequence length [2504.07654].

What changes in MS-Mamba is usually **not the recurrence itself, but the way sequences are formed or parameterized around it**. In ms-Mamba for forecasting, several Mamba blocks run in parallel with different sampling rates $\Delta_i$, and their outputs are averaged, so the same input is processed at several temporal scales [2504.07654]. In Multi-Scale VMamba, a feature map is split into a full-resolution branch and a downsampled branch; one scan is applied to the former and three scans to the latter, after which low-resolution outputs are interpolated and fused back [2405.14174]. In MS-Temba, dilation is pushed into the SSM by a reshaping operator $\Phi_\eta$ that partitions a sequence into subsequences such as odd and even indices, applies BiMamba to each, and then restores the original layout through $\Phi_\eta^{-1}$ [2501.06138]. In offline RL, MambaDM uses a parallel global–local fusion Mamba module in which one branch processes the full trajectory while another processes fixed-length windows, making multi-scale dependence explicit at the architecture level rather than inside a single scan kernel [2406.02013].

This suggests that MS-Mamba is best understood as **scale-aware selective scanning**: the SSM core remains intact, while scale enters through multiple $\Delta$s, dilated index mappings, multi-resolution feature maps, grouped channels, or explicit local/global branches.

## 3. Spatial, spectral, and hierarchical vision realizations

In image-like domains, MS-Mamba is frequently realized through **multi-resolution scanning plus hierarchical pyramids**. Multi-Scale VMamba is exemplary: it replaces four full-resolution SS2D scans with a “hierarchy-in-hierarchy” design that performs one scan at full resolution and three on a downsampled map, then supplements token mixing with a ConvFFN for channel mixing [2405.14174]. The motivation is explicit: downsampling reduces sequence length, improves effective long-range modeling, and lowers cost, while a retained full-resolution scan preserves fine detail.

MSFMamba, developed for joint classification of hyperspectral image (HSI) and LiDAR/SAR data, gives a more task-specific but closely related formulation. The network stacks spatial–spectral Mamba modules, each applying MSpa-Mamba to both modalities, Spe-Mamba to HSI only, and Fus-Mamba for cross-modal fusion. Its multi-scale component is the Multi-Scale Spatial SSM inside MSpa-Mamba: depth-wise convolutions with strides 1 and 2 generate full- and half-resolution maps, two scanning routes operate at full resolution, two at half resolution, and the half-resolution outputs are upsampled before aggregation. The same model also separates spatial, spectral, and fusion roles across MSpa-Mamba, Spe-Mamba, and Fus-Mamba, respectively [2408.14255].

M3SR, for RGB-to-HSI spectral reconstruction, embeds Mamba in a U-Net-like encoder–decoder with three conceptual scales—global, intermediate, and local—and assigns a multi-perceptual fusion block to each scale. That block combines a spatial branch built on VMamba’s VSS/SS2D, a frequency branch built on DWT–VSS–IDWT, and a spectral branch that treats per-pixel spectra as 1D sequences for Mamba processing. Here the multi-scale notion is inseparable from the encoder–decoder pyramid, while the multi-perceptual notion adds domain transforms inside each scale [2601.08293].

Medical imaging variants extend the same logic. MSV-Mamba uses a U-shaped model whose decoder contains an LMS module composed of PiM and PaM: PiM performs window-level pixel-sequence Mamba, whereas PaM pools each window to a token, applies bidirectional Mamba across tokens, and unpools back to full resolution. MSAA then fuses decoder features with spatial and channel attention, and auxiliary losses supervise multiple decoder depths [2501.07120]. MSVM-UNet instead augments VSS blocks with an MS-FFN that sums depth-wise convolutions of sizes $1\times1$, $3\times3$, and $5\times5$, and replaces ordinary patch expanding with large-kernel patch expanding so that spatial and channel information are jointly integrated during upsampling [2408.13735]. In 3D segmentation, the UlikeMamba study inserts multi-scale DWConv3D branches before Mamba and analyzes scan strategies ranging from single-scan to Tri-scan, thereby turning scale selection and scan complexity into explicit design variables [2503.19308].

## 4. Temporal, decision, and recommendation variants

In sequential domains, MS-Mamba is usually defined over **temporal scales rather than spatial resolutions**. ms-Mamba for multivariate forecasting is the most literal instance: each layer runs several bidirectional Mamba blocks with distinct sampling rates $\Delta_i$, and the outputs are averaged before normalization and an MLP. The paper evaluates fixed, learnable, and dynamic temporal scales, with the fixed multiplier set $\alpha=(1,2,4,8)$ and learnable $\Delta_i$ both performing strongly [2504.07654].

MambaDM for offline reinforcement learning uses a different temporal interpretation. Its GLoMa layer splits a reward-conditioned trajectory into a global branch over the full sequence and a local branch over fixed-length windows, applies Mamba to both, fuses them additively, and follows with an FFN. The motivation is specific to RL trajectories, which exhibit both local Markovian structure and global trajectory-scale correlations [2406.02013].

MS-Temba adopts a block-wise temporal hierarchy for densely labeled temporal action detection. Each Temba Block contains a Temporal Local Module, implemented as a 1D temporal convolution with dilation $\delta_b$, followed by a Dilated Temporal SSM in which the sequence is rearranged by $\Phi_{\eta_b}$ into stride-$\eta_b$ subsequences, processed by BiMamba, and restored to the original order. Outputs from several such blocks are projected to a common dimension and fused by a Temporal Mamba Fuser, which itself uses BiMamba after summation [2501.06138].

M2Rec shows that some authors extend “multi-scale” beyond time alone. Its Adaptive FFT–Powered Mamba encoder first transforms a user interaction sequence to the frequency domain, masks high-frequency components, reconstructs a denoised sequence by inverse FFT, and then applies Mamba. The resulting temporal-frequency representation is fused with LLM-based semantic item embeddings through a learnable gate. In this usage, the scales are temporal, frequency-domain, and semantic rather than multiple SSM kernels in the narrow sense [2505.04445].

## 5. Empirical behavior, efficiency, and representative results

The empirical record consistently links MS-Mamba designs to **improved accuracy-efficiency trade-offs**, though the exact benefit depends on domain and realization. On ImageNet, COCO, and ADE20K, MSVMamba-Tiny reports **82.8% top-1 accuracy**, **46.9% box mAP**, **42.2% instance mAP**, and **47.6% mIoU with single-scale testing on ADE20K** [2405.14174]. In time-series forecasting, ms-Mamba on ETTh2 at horizon 96 achieves **MSE 0.291, params 0.481M, memory 1.84MB, and MACs 0.165G**, compared with **MSE 0.296, params 1.150M, memory 4.40MB, and MACs 1.563G** for S-Mamba [2504.07654].

In multi-source remote sensing, MSFMamba reports the best overall accuracy and Kappa on Berlin, Augsburg, and Houston2018, with **OA 77.11% / 91.30% / 92.38%** and, on Augsburg, **1.5252M parameters** and **0.0377G FLOPs**, compared with **37.7M parameters and 3.56G FLOPs** for FusAtNet [2408.14255]. In spectral reconstruction, M3SR on NTIRE2022 reaches **RMSE 0.0343**, **PSNR 31.3995 dB**, **SAM 6.6199**, **MSSIM 0.9351**, with **2.166M** parameters and **100.924G** FLOPs for $482\times512$ input [2601.08293]. In long-video temporal action detection, MS-Temba reports **10M** temporal parameters and **0.6 GFLOPs**, versus **87M** and **6.6 GFLOPs** for MS-TCT, while remaining state-of-the-art on long-duration videos [2501.06138]. In 3D volumetric segmentation, the integrated UlikeMamba\_3dMT reports **Dice 89.95 on AMOS** and **90.60 on BraTS**, with **93.09G** FLOPs [2503.19308].

These results do not imply a single dominant recipe. Rather, they suggest that explicit scale handling—through downsampled scans, multiple $\Delta$s, dilated SSMs, or hierarchical feature pyramids—can convert Mamba’s linear-time recurrence into a more task-aligned inductive bias.

## 6. Misconceptions, limitations, and open directions

A common misconception is that “multi-scale” always means **multi-resolution images**. The literature does not support that reading. In MSFMamba it is full- and half-resolution spatial SSM routes; in ms-Mamba it is multiple temporal sampling rates $\Delta_i$; in M2Rec it includes frequency-domain and semantic scales in addition to time; and in MambaDM it is the coexistence of local windows and full trajectories [2408.14255] [2504.07654] [2505.04445] [2406.02013].

A second misconception is that more elaborate scanning or larger models are automatically beneficial. The 3D volumetric study explicitly finds that **simpler methods often suffice**, while **Tri-scan** yields notable advantages only in the most challenging scenarios [2503.19308]. MambaDM reports a scaling-law result in which **increasing model size does not bring performance improvement**, whereas **scaling the dataset amount by 2x for MambaDM can obtain up to 33.7% score improvement on Atari dataset** [2406.02013]. MSVM-UNet shows that the kernel set **$[1,3,5]$** outperforms **$[3,5,7]$** and **$[1,3,5,7]$**, indicating that larger or more numerous branches can degrade HD95 and increase compute [2408.13735]. MS-Temba similarly finds that three Temba blocks are optimal, while a fourth sharply reduces TSU mAP [2501.06138].

Open directions in the literature are correspondingly pragmatic. MSFMamba explicitly mentions **hybrid Mamba + Transformer frameworks** and improved interpretability for multi-source remote sensing [2408.14255]. M3SR also points to hybridization and suggests that the current L1-only objective could be extended by other losses, though it does not implement them [2601.08293]. This suggests that the next phase of MS-Mamba research is likely to concern **how scale should be allocated, fused, and interpreted**, rather than whether selective SSMs can replace attention in a uniform way.

Source: https://www.emergentmind.com/topics/multi-scale-mamba-ms-mamba-1061e75b-66e4-4fcc-8b98-f85e2bf826e0