Lifting 3D Gaussians: A Wavelet Multiresolution Representation for Scalable Gaussian Splatting Streaming

arXiv:2606.07179 · cs.CV, cs.MM, eess.IV · Submitted 2026-06-05 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: AI Radio. Generated commentary on the latest Artificial Intelligence papers.

Tom: We're looking at the paper: "Lifting 3D Gaussians".

Title and authors: Tom: So we're talking about this new paper today called "Lifting three dee Gaussians: A Wavelet Multiresolution Representation for Scalable Gaussian Splatting Streaming" <ref:2606.07179#pg1>. It looks like they've been tackling the problem of making these huge three dee scenes stream smoothly and efficiently <ref:2606.07179#pg1>.

Jane: That’s right, Tom. It seems like they are moving away from the older ways of doing this, which have some pretty big issues with quality jumps and wasted data when you try to stream things progressively.

Lu: The core idea here is that they replace the discrete layering approach with something continuous called an Evolution Tree, where each splat actually refines into two children. This structure is designed to handle those error accumulations we see in the old methods, so child nodes can structurally correct errors from their parents.

Meng: Structurally correcting errors sounds ambitious; what does that actually mean for the practical side? Does it just make the file bigger, or does it actually help with how fast we can download it?

Tom: Well, on the mechanics side of things, they are using these specific equations to define that refinement. They have C one equals P plus theta one and C two equals P minus alpha times theta. This means they use a shared direction, called theta, and an asymmetry factor called alpha to decide how the splats split when they evolve.

Jane: That's a very technical way of saying they are mathematically defining how the refinement happens between layers. It sounds like this asymmetry factor is key to getting those specific gains in quality and storage efficiency that they mention later.

Lalam: I can confirm that the paper states that using an asymmetric collinear residual, which is option D, gives them five parameters per split to capture complex details while keeping wavelet sparsity. This tells me they are being very deliberate about how much detail each step adds.

Tom: Exactly, they are not just adding random noise or detail at every level; they are using a specific structure that is inspired by wavelets to ensure the refinement signals themselves stay sparse. That sparsity is what makes it good for progressive coding.

Jane: So if I understand correctly, the main thing here is that this wavelet-inspired refinement mechanism creates these inherent sparse signals, which directly translates into better compression when you try to save the model down for streaming purposes.

Lu: It does; and they show some pretty solid numbers on that compression. They report that applying standard compression to those refinement residuals shrinks the storage footprint by a factor of eight point seven times compared to their previous work, which is pretty significant for transmission <ref:2606.07179#pg3>.

Meng: A factor of eight point seven times sounds promising when you’re shipping assets over a network with limited bandwidth <ref:2606.07179#pg3>. But what about the trade-off? Does this better compression come at the cost of visual quality?

Title and authors: Tom: That’s a fair question, Meng. They show that this compression is achieved with minimal visual degradation, and they even report a minimal point error of zero point one dB when they apply standard compression to those residuals.

Jane: So it seems like you get the benefit of much smaller files and better quality than before, all while maintaining a very smooth transition between qualities. That smooth quality progression is another big feature they highlight for streaming applications.

Lu: The continuous layering structure itself is what enables that smoothness; transitioning from one layer to the next isn't just replacing splats randomly; it’s replacing parents with correlated children offset by those sparse refinement signals, which keeps the visual updates minute and localized.

Meng: I worry about how this scales on real hardware. If we have a million splats, does the tree structure keep rendering costs manageable when we are doing per-region adaptive rendering?

Tom: They address that by saying the model is traversed as a tree, which lets clients render different spatial regions at mixed resolutions simultaneously. This fine-grained spatial adaptation is something discrete layering simply can't do structurally.

Jane: That’s a huge deal for real-time applications because it means the user can see high quality where they are looking and lower quality in the periphery without having to load massive files everywhere.

Lalam: And I've looked at the results on certain datasets, like Playroom, and it seems like this continuous layering approach achieves a better visual quality and a smaller model size compared to methods like LapisGS, which is what they’re comparing against.

Lu: Yes, looking at the sample renderings for Drjohnson or Playroom, you can see that instead of the errors just piling up in one layer, the continuous layering structure ensures that higher layers actively correct those ancestral splats through those learned refinements.

Meng: So if I were to deploy this right now, what’s the biggest immediate concern? Is it training time or just deployment complexity?

Tom: The training part is intensive because you are learning all these refinement directions and asymmetry factors, but they show that the resulting model is highly compressible, which helps with storage. The main operational win seems to be the reduction in transmission payload and GPU memory footprint by factors of up to two point five times and five point five times.

Jane: It sounds like for someone who just wants to see cool three dee scenes streamed on a mobile device, this paper suggests a much more robust way than what we've seen before <ref:2606.07179#pg1>. But they also mention that the symmetric variant of this approach has a noticeably higher redundancy rate at level three compared to the asymmetric model because forcing equal magnitude constraints can sometimes put child splats in less optimal geometric positions.

Lalam: That’s an important caveat; it means you have to choose between a slightly more redundant but potentially simpler asymmetric structure and the symmetric one, depending on what you prioritize.

Title and authors: Lu: Ultimately, the paper demonstrates that this continuous layering paradigm provides a way to scale visual fidelity smoothly and proportionally with the data received, unlike discrete layering which has those stagnant improvements punctuated by abrupt jumps at LOD boundaries.

Tom: It seems like they've found a solid way to solve that inherent structural inefficiency in progressive representations by making the refinement explicit between layers.

Jane: So we’re talking about a representation called "Lifting three dee Gaussians: A Wavelet Multiresolution Representation for Scalable Gaussian Splatting Streaming" that uses an evolution tree to make streaming much smoother and smaller <ref:2606.07179#pg1>.

Meng: It sounds like the efficiency gains in storage and rendering footprint are substantial, which is what we need for deployment on constrained hardware.

Lalam: I'm confident that the findings regarding the reduction in splat redundancy from over sixty five percent down to under twenty five percent are a very strong indicator of its success.

Tom: Yeah, it’s a concrete number showing that this isn't just theoretical; there is actual data on how much redundant geometry they are eliminating.

Jane: It really shows how important it is to have those explicit parent-child relationships when you are building a system designed for streaming rather than just static rendering.

Lu: And the way they use wavelet decomposition to generate those refinement residuals gives them that inherent sparsity, which is what makes the signals highly compressible.

Meng: So we’ve covered how it works and why it’s good for size and quality, but where does this leave things for future work? Are there limitations they flag?

Tom: They do mention that the method relies on training the base layer L zero on data D zero for thirty thousand iterations, and subsequent levels Li on Di for another thirty thousand iterations. That’s a lot of training data and time commitment upfront.

Jane: And they also note that the work is built on existing three deeGS codebases, which is good because it means it integrates well with what we already know about three dee Gaussian Splatting <ref:2606.07179#pg1>.

Lalam: The main limitation they point out seems to be that the symmetric variant has a higher redundancy rate at level three compared to the asymmetric model, suggesting a design choice needs to be made based on your specific application needs.

Lu: That’s the trade-off they are presenting; it’s not a perfect solution for every scenario, but it clearly shows how continuous layering resolves the core issues of error accumulation in discrete systems.

Tom: So, in summary, this paper proposes a continuous layering paradigm with an evolution tree to achieve smooth quality transitions and better quality-to-size efficiency by using wavelet-inspired refinement signals.

Jane: It’s a really interesting direction for scalable three deeGS streaming because it tackles the structural limitations of discrete methods head-on with this evolutionary approach <ref:2606.07179#pg1>.

Lalam: I think the findings on dramatic reductions in transmission payload and rendering memory footprint make this paper a very important piece of literature for anyone working on efficient three dee content delivery <ref:2606.07179#pg1>.

The paper's summary: Tom: So, we're diving into this paper now called "Lifting three dee Gaussians: A Wavelet Multiresolution Representation for Scalable Gaussian Splatting Streaming". Basically, they’re ditching the old way of layering three dee Gaussian Splatting because it causes errors to pile up and wastes a ton of data.

Jane: That's right. Think of it like building a house layer by layer where each new floor messes up the foundation below it, and this paper is proposing a way to build that house continuously, like one big tree structure instead of separate boxes.

Lu: The core idea is an Evolution Tree. Each splat node actually refines into two children through some learned refinement process. This structure lets the child nodes fix mistakes made by their parent layers directly.

Meng: So, structurally correcting errors means that when you look at a lower quality layer, the higher quality layers are actively moving and fixing the geometry underneath them? That sounds like a lot of computation.

Tom: It is computationally heavy to train because they're learning all these refinement directions and asymmetry factors, but they show it works by showing how much redundancy they eliminate. They cut that redundancy from over sixty-five percent down to under twenty-five percent.

Jane: That reduction in redundancy is huge because it means the file size shrinks dramatically when you stream it. But what about how smooth the quality jump is when someone is watching?

Tom: The transition between levels isn't a sudden jump anymore. It’s just replacing one set of parent splats with correlated children that are offset by these sparse refinement signals. It results in visual updates that are tiny and localized, which makes the quality progression look continuous instead of abrupt.

Lu: And that sparsity comes from the wavelet-inspired refinement mechanism they use, which means the signals between layers are inherently very compressible. They show applying standard compression shrinks the storage footprint by a factor of eight point seven times at level three compared to older methods like LapisGS.

Meng: Halving or quartering the size of a three dee model is exactly what I need for mobile deployment, and they’re saying this happens with negligible visual loss, only about zero point one decibel in PSNR penalty. That's solid evidence if it holds up in real-world traffic.

Jane: So, for someone just listening to the show, this means that instead of waiting for a huge file to download before you can see anything good, you get a continuous stream where the quality improves smoothly as the data arrives. It’s adaptive rendering across different spatial regions too.

Tom: Right. They can render high quality where you're looking and lower quality in the edges at the same time because they are treating it like a tree structure rather than just one fixed layer for every level of detail.

Lu: The main implication is that this continuous layering paradigm offers a way to scale visual fidelity proportionally with the data received, which is fundamentally different from discrete layering where you get stuck at certain quality levels until you download the next big chunk.

Jane: It really shows how they’ve solved that structural problem of error accumulation inherent in those older progressive methods by making the correction process explicit between every layer. This could be a real deal for immersive three dee streaming applications.

Tom: The training setup is pretty demanding, though, requiring thirty thousand iterations for the base layer and another thirty thousand for each subsequent level, but they show that the resulting model stays compact in terms of memory footprint compared to baselines like LapisGS.

Meng: I see the training time as a hurdle for getting this into production quickly, but if you can get that refinement learning done once, the operational savings from smaller files and faster streaming will be worth it.

Jane: So while there are some upfront costs in terms of training time and complexity, the payoff seems to be a representation that is dramatically more efficient for transmission and better at maintaining a smooth viewing experience. We’ll take a closer look at how this actually scales with bigger scenes next.

The paper's improvements: Tom: So, we’re looking at what they actually improved in terms of performance here in "Lifting three dee Gaussians". They aren't just adding a new layer; they are changing the fundamental way those layers talk to each other.

Jane: That means if you’re trying to stream something, the quality doesn't just jump when you get more data; it actually flows smoothly as the data arrives, which is a big deal for real-time applications.

Lu: The paper focuses on how the wavelet-inspired refinement residuals create this inherent sparsity. It means the signals between layers are structured in a way that is naturally very easy to compress without losing much detail.

Meng: So, if we look at deployment costs, this suggests we can cut down on the memory footprint of these three dee models significantly because they aren't storing all that redundant geometry anymore. How much smaller are we talking?

Tom: They show a dramatic reduction in storage. Applying standard compression to those residuals shrinks the file size by about eight point seven times at level three, which is what makes it viable for transmission over networks with limited bandwidth.

Jane: That compression benefit is massive because it means less data has to travel from the server to your device, and that translates directly into faster downloads and less battery drain on mobile phones.

Lu: And they also show better quality retention when you compress it down. They report only a zero point one decibel penalty in PSNR when applying standard compression, which is very low for this kind of reduction.

Tom: That’s the clean ablation we needed; they didn't sacrifice visual fidelity for the sake of smaller file sizes. It shows that this isn't just a trick to make things small; it’s a structural change that makes the data itself more efficient.

Meng: I gotta ask about those training requirements again because I can’t ignore them, Tom. They said thirty thousand iterations for the base layer and another thirty thousand for each subsequent level. How much does that slow down my pipeline when I'm trying to iterate on different scenes?

Jane: That training cost is definitely a hurdle; it means you have a significant upfront investment in learning those refinement parameters, but once trained, the inference process benefits from that structural efficiency we just talked about.

Lu: The asymmetry factor alpha plays a role here too. They compare the symmetric variant to the asymmetric one and note that the symmetric version actually has a higher redundancy rate at level three because forcing equal magnitude constraints can sometimes lead to less optimal geometric positions for the child splats.

Tom: That’s a real caveat; so you have to choose between a slightly more redundant but simpler structure and something asymmetric that might be geometrically better, depending on what you prioritize in your specific streaming use case.

Jane: It means there's no single perfect answer; the choice depends on whether you want the simplest math or the most accurate geometry correction at every step.

Lu: Overall, this continuous layering structure solves the core problem of error accumulation by allowing higher layers to actively fix lower layers, which is a major structural improvement over discrete methods.

Tom: So it’s not just about getting a better quality model; it’s about building a representation that scales visually and proportionally with the data you actually receive, which is what makes this work for streaming.

Jane: Exactly. It moves us away from those old systems where you either get poor quality or a huge file size, and towards something adaptive where the visual fidelity matches exactly what the network can deliver at that moment. We’ll keep digging into how they handle those per-region adaptive renderings next.

Conclusion: Tom: So, to wrap up, the paper "Lifting three dee Gaussians: A Wavelet Multiresolution Representation for Scalable Gaussian Splatting Streaming" introduces a continuous layering approach using an evolution tree to fix errors structurally and make streaming much smaller and smoother.

Jane: It’s essentially taking those messy discrete layers and turning them into a flowing evolution where the layers actively correct each other, leading to better quality while keeping the file size way down.

Lu: The core of it is that wavelet-inspired refinement signals are inherently sparse, which means we can use compression much more aggressively without hurting the visual quality at all.

Meng: For deployment purposes, it seems like the main trade-off is that you have to commit to a lot of training time upfront for those refinement directions and asymmetry factors before you can even start making models.

Lalam: I’m pretty sure this approach is sound; the reduction in splat redundancy from over sixty-five percent down to under twenty-five percent is a concrete number showing its success in cutting wasted geometry.

Tom: Yeah, that’s what tells me it’s not just theoretical stuff; there's real data on how much redundant geometry they are eliminating across different datasets.

Jane: It really shows how important those explicit parent-child relationships are for building systems designed for streaming, especially when you have limited bandwidth or memory on the client side.

Lu: The way this structure allows per-region adaptive rendering is a huge practical benefit; it means clients can load different quality levels depending on what they’re currently viewing in the scene.

Meng: That spatial adaptation is definitely something we need to think about for our next generation of resource-constrained applications. It changes how much memory we have to budget for rendering at any given moment.

Tom: So, this paper moves us toward a representation that scales visual fidelity smoothly and proportionally with the data you receive, rather than having those sudden quality drops at level boundaries anymore.

Jane: That smooth progression is what makes it so appealing for immersive environments; it’s not just about getting a high-quality image, it’s about the experience of seeing that quality improve gracefully as you wait for more data to arrive.

Lu: This work sets a new baseline by showing how continuous layering can resolve the error accumulation problems that plague discrete representations in scalable three dee Gaussian Splatting.

Tom: It’s a robust and efficient foundation for moving three dee models out of the research lab and into actual streaming applications, provided you can handle the initial training complexity.

Jane: We’ll definitely be watching how this continuous layering paradigm evolves as researchers try to push those refinement mechanisms even further in future work.

Yuang Shi, Simone Gasparini, Géraldine Morin, Wei Tsang Ooi

National University of Singapore · IRIT - University of Toulouse, France · IPAL, IRL2955, Singapore

cs.CV, cs.MM, eess.IV

Submitted: 2026-06-05

Updated: 2026-11-18

Comments: Accepted by MMSys'27; Project page: https://yuang-ian.github.io/liftgs/

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 96/100

The gist: Streaming 3D Gaussian Splatting requires highly scalable, progressive representations The paper proposes EvoGS, a continuous-layering representation organized as an Evolution Tree…

Key concepts

Continuous Layering Representation
Instead of fixed, separate quality levels (discrete), this approach uses an Evolution Tree where each splat refines into two children. This structure creates a continuous progression of detail, allowing higher quality layers to actively correct and refine the geometry inherited from lower layers without accumulating errors.
Evolution Tree
This is the organizational structure for the continuous layering representation. It functions like a family tree where an internal splat node evolves into two children through learned refinement equations. This hierarchy explicitly defines parent-child relationships, enabling structural error correction across quality tiers.
Wavelet Multiresolution Representation
The refinement mechanism uses principles from wavelet decomposition to generate the signals that guide layer evolution. This wavelet-inspired approach ensures that the inter-layer refinement residuals are inherently sparse, which is crucial for making the overall representation highly compressible and efficient for streaming.
Structural Error Correction
A core feature where child nodes are explicitly designed to correct errors from their ancestors. Using learned refinement directions ($ heta$) and asymmetry factors ($\alpha$), higher layers can actively reposition or adjust ancestral splats, resolving the error accumulation issues found in traditional discrete layering methods.

Terminology

Summary

Streaming 3D Gaussian Splatting requires highly scalable, progressive representations The paper proposes EvoGS, a continuous-layering representation organized as an Evolution Tree, which fundamentally resolves the error accumulation and splat redundancy problems inherent in existing discrete-layering approaches This approach enables smooth quality transitions and achieves better quality–storage efficiency for scalable 3D streaming. p. 1

Where the paper says it

We propose a continuouslayering approach named EvoGS, which is a fundamental shift away from the discrete-layering representations used in existing progressive scalable streaming. By structuring splats into an Evolution Tree with explicit parent-child lineage, our method fundamentally resolves the error accumulation and splat redundancy problems inherent in existing discrete-layering approaches.

Pruning redundant reasoning tokens

Existing progressive methods rely on discrete layering, which leads to error accumulation, severe splat redundancy, and uncontrolled quality transitions The authors propose EvoGS as the first continuous-layering representation organized as an Evolution Tree with explicit parent-child refinement. This structure empowers child nodes to structurally correct ancestral errors, yielding inherently sparse and highly compressible inter-layer signals. p. 0

Where the paper says it

Organized as an Evolution Tree, EvoGS generates finer details via an explicit, wavelet-inspired parent-child refinement. This empowers child nodes to structurally correct ancestral errors, yield inherently sparse and highly compressible inter-layer signals.

Extensive experiments show EvoGS eliminates splat redundancy from over 65% to under 25%. p. 0
Where the paper says it

This empowers child nodes to structurally correct ancestral errors, yield inherently sparse and highly compressible inter-layer signals. Extensive experiments show EvoGS eliminates splat redundancy from over 65% to under 25%.

How the training signal is built

The continuous layering paradigm structures splats into an Evolution Tree where each internal splat node evolves into two children through a learned refinement. This refinement is defined by the equations C1 = P + θ1 and C2 = P - α⊙θ, where θ is a shared refinement direction and α is an asymmetry factor. p. 1

Where the paper says it

(c) Continuous layering: each LOD is a structured refinement of the previous one; quality improves smoothly. Inspired by wavelet transformations [Sweldens 1996], we organize splats into a binary tree where each internal splat node 𝑷 evolves to two splat children 𝑪1, 𝑪2 through a learned refinement: 𝑪1 = 𝑷 + 𝝍, 𝑪2 = 𝑷 − 𝜶 ⊙ 𝝍, where 𝝍 is a shared refinement direction, and 𝜶 is an asymmetry factor.

Option D, the asymmetric collinear residual, uses D + 5 parameters per split to capture complex details while preserving wavelet sparsity. p. 1
Where the paper says it

We design a collinear refinement mechanism to generate child splats. Grounded in wavelet decomposition theory, this formulation yields inherently sparse refinement signals while utilizing a learned per-attribute asymmetry factor to flexibly capture complex, non-symmetric scene details without wasting representational capacity.

The leaf's final parameters S are computed by accumulating refinements from root to leaf via S = Proot + Σlk=1 sk⊙θk.

Where the gains come from

The primary gain comes from the explicit parent-child lineage, which ensures inter-layer relationships are explicit and traversable. This design resolves discrete layering limitations by allowing higher layers to actively reposition and correct ancestral splats via learned refinements. p. 6

Where the paper says it

This is paradigm-level evidence: continuous layering’s advantage is structural to the construction principle, not specific to any particular discrete-layering design. The advantage stems from mitigating the error accumulation inherent to discrete layering: both LapisGS and L3GS permanently freeze lower-layer splats, locking in geometric errors that propagate upward, whereas EvoGS’s explicit parentchild lineage allows higher layers to actively reposition and correct ancestral splats via the learned refinements 𝝍 and 𝜶.

The wavelet-inspired refinement mechanism generates inherently sparse refinement signals, which are favorable for progressive coding. p. 0
Where the paper says it

Organized as an Evolution Tree, EvoGS generates finer details via an explicit, wavelet-inspired parent-child refinement. This empowers child nodes to structurally correct ancestral errors, yield inherently sparse and highly compressible inter-layer signals.

This structural property naturally yields highly compressible signals, as evidenced by the compression of psi to 91. p. 8
Where the paper says it

1 and validated in Figures 5 and 6, the wavelet-inspired refinement residuals 𝝍 exhibit strong sparsity. This structural property naturally yields highly compressible signals.

75 MB for EvoGS at L3 compared to LapisGS's 801.80 MB.

Smooth Renderings in Between

EvoGS inherently supports smooth quality transition because transitioning from Fi to Fi+1 simply replaces individual parents with correlated children offset by a highly sparse psi. This results in visual updates that are minute and localized, yielding an inherently continuous quality progression rather than abrupt jumps. p. 4

Where the paper says it

In EvoGS, transitioning from F𝑖 to F𝑖+1 simply replaces individual parents with correlated children offset by a highly sparse 𝝍. As incremental refinement data arrives, visual updates are minute and localized, yielding an inherently continuous quality progression.

Furthermore, the model is traversed as a tree, allowing spatial regions to be rendered at mixed resolutions simultaneously through per-region adaptive rendering. p. 4
Where the paper says it

Per-region adaptive rendering. Because our model is traversed as a tree, spatial regions can be rendered at mixed resolutions simultaneously.

This fine-grained spatial adaptation is structurally impossible in discrete-layering representations. p. 4
Where the paper says it

𝑗 in a single pass. This fine-grained spatial adaptation is structurally impossible in discrete-layering representations.

Compressibility

The wavelet-inspired refinement residuals psi exhibit strong sparsity, which naturally yields highly compressible signals. Table 5 reveals that applying standard compression to psi aggressively shrinks this to 91. p. 9

Where the paper says it

80 MB) simply by eliminating ghost splats. Applying standard compression to 𝝍 aggressively shrinks this to 91.

75 MB for EvoGS at L3, which is 11. p. 0
Where the paper says it

EvoGS: Constructing Continuous-Layered Gaussian Splatting with Evolution Tree for Scalable 3D Streaming Yuang Shi1,2,3 , Simone Gasparini2,3 , Géraldine Morin2,3 , Wei Tsang Ooi1,3 1National University of Singapore, Singapore 2 IRIT - University of Toulouse, France 3 IPAL, IRL2955, Singapore 17. 81% Ghost Splats 57.

5% of the LapisGS size with a negligible PSNR penalty of 0. p. 9
Where the paper says it

75 MB, which is 11. 5% of the LapisGS, with a negligible PSNR penalty of 0.

10 dB. This confirms that continuous layering actively structures the scene into sparse, compressible residuals ideal for scalable transmission. p. 9
Where the paper says it

5% of the LapisGS, with a negligible PSNR penalty of 0. This confirms that continuous layering not only mitigates structural bloat, but actively structures the scene into sparse, compressible residuals ideal for scalable transmission.

The symmetric variant of EvoGS produces a noticeably higher redundancy rate (38. p. 8
Where the paper says it

Furthermore, this analysis empirically validates our asymmetric design choice. The symmetric variant of EvoGS produces a noticeably higher redundancy rate (38.

85% at LOD 3) than the asymmetric model because the rigid equalmagnitude constraint forces some child splats into sub-optimal geometric positions. p. 8
Where the paper says it

85% at LOD 3) than the asymmetric model. This occurs because the rigid equalmagnitude constraint forces some child splats into sub-optimal geometric positions when capturing non-symmetric local details in real-world scenes.

Conclusion

EvoGS introduces a continuous layering paradigm that resolves structural inefficiencies for scalable 3DGS streaming by establishing an explicit parent-child lineage across quality tiers. The wavelet-inspired refinement mechanism enables finer layers to structurally correct lower-layer geometry, naturally eliminating redundant splats and yielding dramatic reductions in transmission payload and rendering GPU memory footprints. p. 9

Where the paper says it

The waveletinspired refinement mechanism enables finer layers to structurally correct lower-layer geometry. This design naturally eliminates redundant splats, yielding dramatic reductions in both transmission payload and rendering GPU memory footprints compared to current baselines.

EvoGS provides a robust, efficient, and adaptive foundation for immersive 6DoF streaming in resource-constrained environments. p. 9
Where the paper says it

Furthermore, the inherent sparsity of the formulation produces highly compressible inter-layer signals and ensures smooth rate-distortion performance. EvoGS provides a robust, efficient, and adaptive foundation for immersive 6DoF streaming in resource-constrained environments.

The local dependency and coherence inherent to EvoGS could offer opportunities for scene editing or animation. p. 9
Where the paper says it

EvoGS provides a robust, efficient, and adaptive foundation for immersive 6DoF streaming in resource-constrained environments. Besides, the local dependency and coherence inherent to EvoGS could offer opportunities for scene editing or animation.

The paper's findings demonstrate that continuous layering scales visual fidelity smoothly and proportionally with received data, unlike discrete layering which exhibits stagnant improvement punctuated by abrupt jumps at LOD boundaries. The structural advantage of EvoGS stems from mitigating the error accumulation inherent to discrete layering by allowing higher layers to actively correct ancestral splats via the learned refinements psi and alpha. p. 6
Where the paper says it

This is paradigm-level evidence: continuous layering’s advantage is structural to the construction principle, not specific to any particular discrete-layering design. The advantage stems from mitigating the error accumulation inherent to discrete layering: both LapisGS and L3GS permanently freeze lower-layer splats, locking in geometric errors that propagate upward, whereas EvoGS’s explicit parentchild lineage allows higher layers to actively reposition and correct ancestral splats via the learned refinements 𝝍 and 𝜶.

The paper's results show that EvoGS dramatically reduces both storage and rendering footprint compared to LapisGS and L3GS, demonstrating superior quality-to-size efficiency. This structural advantage is validated by the fact that the EvoGS cluster remains highly compact in the top-left of Figure 7 while other baselines drift rightward across levels. p. 6
Where the paper says it

The two baselines exhibit different growth patterns — LapisGS’s storage inflates unboundedly as densification adds splats per layer, while L3GS’s storage is capped by its pretrained-then-pruned initialization but grows linearly with layer count — yet EvoGS dominates both, because neither growth pattern reflects efficient representation. Figure 7 visualizes this: the EvoGS cluster (red stars) remains highly compact in the top-left, while LapisGS (orange) and L3GS (purple) drift rightward across levels.

The paper concludes that EvoGS provides a robust, efficient, and adaptive foundation for immersive 6DoF streaming in resource-constrained environments. p. 9
Where the paper says it

Furthermore, the inherent sparsity of the formulation produces highly compressible inter-layer signals and ensures smooth rate-distortion performance. EvoGS provides a robust, efficient, and adaptive foundation for immersive 6DoF streaming in resource-constrained environments.

The references cited include Kerbl et al. p. 9
Where the paper says it

Besides, the local dependency and coherence inherent to EvoGS could offer opportunities for scene editing or animation. References Jonathan T.

2023, Shi et al. p. 5
Where the paper says it

Our implementation is built on the official 3DGS and LapisGS codebases [Kerbl et al. 2023; Shi et al.

2025, and Sweldens 1996. p. 1
Where the paper says it

(c) Continuous layering: each LOD is a structured refinement of the previous one; quality improves smoothly. Inspired by wavelet transformations [Sweldens 1996], we organize splats into a binary tree where each internal splat node 𝑷 evolves to two splat children 𝑪1, 𝑪2 through a learned refinement: 𝑪1 = 𝑷 + 𝝍, 𝑪2 = 𝑷 − 𝜶 ⊙ 𝝍, where 𝝍 is a shared refinement direction, and 𝜶 is an asymmetry factor.

The paper also references work on Mip-NeRF360 and Tanks&Temples. p. 5
Where the paper says it

2020], Mip-NeRF360 (Mip-360) [Barron et al. 2022], Tanks&Temples (T&T) [Knapitsch et al.

The paper's methodology is built on the official 3DGS and LapisGS codebases. The references also include works by Collet and Kucherawy, Taubman and Marcellin, Kerbl et al. p. 3
Where the paper says it

The collinearity constraint is preserved, as the children still lie along a single axis through 𝑷. Paralleling how JPEG2000 [Taubman and Marcellin 2002] adopted biorthogonal filters over orthogonal Haar wavelets, our 𝜶 trades exact mean preservation for representational power, while preserving the essential properties of the wavelet framework.

2024, and Yuan-Chun Sun et al. p. 9
Where the paper says it

991–1000. Yuan-Chun Sun, Yuang Shi, Cheng-Tse Lee, Mufeng Zhu, Wei Tsang Ooi, Yao Liu, Chun-Ying Huang, and Cheng-Hsin Hsu.

2025. The paper's conclusion is supported by quantitative results showing EvoGS improves PSNR across various datasets at different quality levels. The paper's overall contribution is a novel paradigm for scalable 3DGS based on continuous layering. The paper's experimental setup involves training the base layer L0 on D0 for 30k iterations and subsequent levels Li on Di for 30k iterations. The paper also references work on ViVo and NETSPLAT. The paper's conclusion is supported by quantitative results showing EvoGS improves PSNR across various datasets at different quality levels.

--- Page 1 ---

EvoGS: Constructing Continuous-Layered Gaussian Splatting with Evolution Tree for Scalable 3D Streaming

Yuang Shi1,2,3, Simone Gasparini2,3, Géraldine Morin2,3 Wei Tsang Ooi1,3 National University of Singapore 1 IRIT - University of Toulouse France IPAL IRL2955 Singapore Conference’17 July 2017 Washington DC USA Shi et al.

The rapid surge of extended reality XR applications has created an urgent demand for efficient delivery of 3D content over dynamic networks. p. 0

Where the paper says it

5×, respectively, and achieves smooth quality transitions optimal for real-time adaptive streaming. Keywords Progressive Gaussian Splatting, Continuous Layering, Splat Refinement, Scalable Streaming 1 Introduction The rapid surge of extended reality (XR) applications has created an urgent demand for efficient delivery of 3D content over dynamic networks.

3D Gaussian Splatting (3DGS) [Kerbl et al. p. 0
Where the paper says it

Keywords Progressive Gaussian Splatting, Continuous Layering, Splat Refinement, Scalable Streaming 1 Introduction The rapid surge of extended reality (XR) applications has created an urgent demand for efficient delivery of 3D content over dynamic networks. 3D Gaussian Splatting (3DGS) [Kerbl et al.

2023] has emerged as a compelling representation for this setting, offering photorealistic quality and real-time rendering. p. 0
Where the paper says it

3D Gaussian Splatting (3DGS) [Kerbl et al. 2023] has emerged as a compelling representation for this setting, offering photorealistic quality and real-time rendering.

However, immersive streaming scenarios involve heterogeneous client devices, unpredictable interactive 6DoF navigation, and dynamic multi-dimensional budgets (e. p. 0
Where the paper says it

2023] has emerged as a compelling representation for this setting, offering photorealistic quality and real-time rendering. However, immersive streaming scenarios involve heterogeneous client devices, unpredictable interactive 6DoF navigation, and dynamic multi-dimensional budgets (e.

g., bandwidth, memory, latency) [Viola and Cesar 2023]. p. 0
Where the paper says it

However, immersive streaming scenarios involve heterogeneous client devices, unpredictable interactive 6DoF navigation, and dynamic multi-dimensional budgets (e. , bandwidth, memory, latency) [Viola and Cesar 2023].

These interleaved consumption and transmission patterns impose three strict requirements on the 3D representation [Gong et al. p. 0
Where the paper says it

, bandwidth, memory, latency) [Viola and Cesar 2023]. These interleaved consumption and transmission patterns impose three strict requirements on the 3D representation [Gong et al.

2025; Han et al. p. 0
Where the paper says it

These interleaved consumption and transmission patterns impose three strict requirements on the 3D representation [Gong et al. 2025; Han et al.

2020; Kim et al. p. 0
Where the paper says it

2025; Han et al. 2020; Kim et al.

2025; Shi 2025]. p. 0
Where the paper says it

2020; Kim et al. 2025; Shi 2025]: arXiv:2606.

--- Page 3 ---

Figure 1: Continuous vs. Discrete Layering for Scalable 3D Streaming EvoGS replaces discrete layered 3DGS layers with a continuous evolution tree, where each splat refines into correlated children. p. 0

Where the paper says it

Discrete Layering for Scalable 3D Streaming. EvoGS replaces discrete layered 3DGS layers with a continuous evolution tree, where each splat refines into correlated children.

This parent-child structure eliminates redundant ghost splats, enables smooth anytime rendering, and achieves better quality–storage efficiency for scalable 3D streaming. p. 0
Where the paper says it

EvoGS replaces discrete layered 3DGS layers with a continuous evolution tree, where each splat refines into correlated children. This parent-child structure eliminates redundant ghost splats, enables smooth anytime rendering, and achieves better quality–storage efficiency for scalable 3D streaming.

--- Page 5 ---

Figure 3: Top (discrete layers): To fit the target geometry (black outline) at LOD 1, the splat from LOD 0 remains frozen and is always rendered for LOD 1. Bottom (evolution tree): The parent splat at LOD 0 evolves into child splats that replace it at LOD 1, achieving a smooth quality transition. p. 2

Where the paper says it

Bottom (evolution tree): The parent splat at LOD 0 evolves into child splats that replace it at LOD 1. The children structurally adapt to fit the finer geometry, achieving a smooth quality transition.

--- Page 8 ---

Figure 8: Sample renderings of Playroom at four quality levels LapisGS [Shi et al. 2025] accumulates the geometry error from low layers, while continuous layering (both Symmetric and Asymmetric) achieves the best visual quality and smallest model size. p. 7

Where the paper says it

LapisGS [Shi et al. 2025] accumulates the geometry error from low layers, while continuous layering (both Symmetric and Asymmetric) achieves the best visual quality and smallest model size.

p. 7
Where the paper says it

LapisGS [Shi et al. 2025] accumulates the geometry error from low layers, while continuous layering (both Symmetric and Asymmetric) achieves the best visual quality and smallest model size.

--- Page 11 ---

Figure 10: Example of quality transition of Lego. The continuous layering construction of EvoGS inherently supports smooth quality transition. p. 8

Where the paper says it

Smooth Renderings in Between 𝐿" Figure 10: Example of quality transition of Lego. The continuous layering construction of EvoGS inherently supports smooth quality transition.

--- Page 4 ---

Figure 2: Three paradigms for LOD 3DGS (a) Space-based hierarchy, (b) Discrete-layered hierarchy, (c) Continuous-layered hierarchy.

--- Page 4 ---

Figure 5: Design space for parent-child refinement (a, b) Unconstrained constructions; (c) A symmetric collinear residual.

--- Page 5 ---

Figure 6: The distribution of psi for each attribute of splat of Playroom as an example.

--- Page 8 ---

Figure 9: Sample renderings of Drjohnson at four quality levels LapisGS [Shi et al. 2025] accumulates the geometry error from low layers, while continuous layering (both Symmetric and Asymmetric) achieves the best visual quality and smallest model size.

--- Page 9 ---

Table 3: The ratios of transparent splats at four quality levels (average across all datasets). EvoGS drops the transparent splat ratio to just 24. p. 8

Where the paper says it

In contrast, our continuous layering approach inherently mitigates this bloat. By refining parent splats rather than overlaying independent sets, EvoGS drops the transparent splat ratio to just 24.

92% at LOD 3 compared to LapisGS's 65.83%.

--- Page 10 ---

Table 5: Comparison of compressibility (averaged across all datasets). Uncompressed EvoGS already halves the storage footprint of LapisGS at L3 by eliminating ghost splats. p. 9

Where the paper says it

Uncompressed EvoGS already halves the storage footprint of LapisGS at 𝐿3 (347. 80 MB) simply by eliminating ghost splats.

The results reveal a dramatic delivery advantage, with uncompressed EvoGS already halving the storage footprint of LapisGS at L3 by eliminating ghost splats. p. 9
Where the paper says it

The results (Table 5) reveal a dramatic delivery advantage. Uncompressed EvoGS already halves the storage footprint of LapisGS at 𝐿3 (347.

--- Page 4 ---

Table 1: Quantitative comparison results on various datasets (Blender [Mildenhall et al. 2020], Mip-NeRF360 [Barron et al. p. 5

Where the paper says it

Table 1: Quantitative comparison results on various datasets (Blender [Mildenhall et al. 2020], Mip-NeRF360 [Barron et al.

p. 5
Where the paper says it

Table 1: Quantitative comparison results on various datasets (Blender [Mildenhall et al. 2020], Mip-NeRF360 [Barron et al.

2022], Tank&Temples [Knapitsch et al. p. 5
Where the paper says it

2020], Mip-NeRF360 [Barron et al. 2022], Tank&Temples [Knapitsch et al.

2606.07179>

--- Page 4 ---

Table 1: Quantitative comparison results on various datasets (Blender [Mildenhall et al. 2020], Mip-NeRF360 [Barron et al. 25]

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Continuous Layering Paradigm for Scalable 3DGS: EvoGS replaces discrete layered 3DGS layers with a continuous evolution tree, where each splat refines into correlated children. This structure eliminates redundant ghost splats and enables smooth anytime rendering.

  2. Structural Correction of Ancestral Errors: The parent-child structure allows child nodes to structurally correct ancestral errors via the learned refinement direction and asymmetry factor, which naturally eliminating redundant ghost splats.

  3. Enhanced Inter-Layer Correlation Exploitation: EvoGS exposes inter-layer correlation that discrete layering cannot, allowing higher layers to actively reposition and correct ancestral splats, thus enabling smooth quality transitions during progressive streaming.

  4. Wavelet-Inspired Refinement for Compressibility: The collinear refinement mechanism is grounded in wavelet decomposition theory, yielding inherently sparse refinement signals which makes the inter-layer signals significantly more compressible than independent splat sets.

  5. Per-Region Adaptive Rendering: By traversing the model as a tree, clients can render different spatial regions at mixed resolutions simultaneously, allowing "foveal regions to be traversed to LN while peripheral regions stop at Lj (j < N)."

Abstract

3D Gaussian Splatting (3DGS) represents a scene as an unstructured set of primitives optimized at a single scale, offering no native support for progressive delivery to heterogeneous devices. Existing scalable approaches organize these primitives into independent quality layers or spatial hierarchies. However, maintaining separate splat representations across scales introduces redundancy, while independent layers have limited ability to correct coarse geometry as finer detail arrives. We present LiftGS, a multiresolution representation that organizes a scene into a coarse Gaussian set and a tree of refinements. Each refinement replaces one coarse Gaussian with two finer children whose attributes are learned relative to their parent. Inspired by wavelet lifting, we encode these changes using a shared detail vector and per-attribute factors that allow the two children to differ asymmetrically from the parent. A client first receives the coarse set, then reconstructs and substitutes children as refinement data arrives. This construction reuses previously transmitted attributes to encode finer detail and removes replaced parents from the active rendering set, reducing both refinement payload and rendering memory. Extensive experiments on four benchmark datasets show LiftGS cuts splat redundancy from over 65% to under 25%, reduces transmission payload and GPU memory by up to 2.4x and 5.5x over state-of-the-art scalable-3DGS baselines, achieves smooth quality transitions in simulated progressive delivery, and compresses a further 8.7x with standard coding at negligible quality loss. Project website: https://yuang-ian.github.io/liftgs/

Sources

Related papers