SPARK: Input-Conditioned Sparse Activation Modulation for Frozen DiT-based Super-Resolution

• URL: http://arxiv.org/abs/2609.03813v1
• Date: Thu, 03 Sep 2026 13:17:29 GMT
• Title: SPARK: Input-Conditioned Sparse Activation Modulation for Frozen DiT-based Super-Resolution
• Abstract summary: We investigate whether dominant channels can serve as a compact adaptation interface for frozen DiT-based SR models.<n>We introduce SPARK, a lightweight input-conditioned controller that predicts bounded per-channel affine transformations for only the selected channels.<n>Experiments on three DiT-based SR backbones across DIV2K, RealSR, and DRealSR show consistent gains in both fidelity and perceptual quality while modulating only eight channels per stream and block.
• Score: 22.659619627342156
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Real-world image super-resolution (SR) increasingly relies on Diffusion Transformer (DiT) backbones, whose internal activations can be dominated by a small number of massive channels. Yet improving perceptual quality in these models still typically requires fine-tuning the network or attaching additional adapters, leaving this structured activation space largely unexplored for adaptation. We investigate whether dominant channels can instead serve as a compact adaptation interface for frozen DiT-based SR models. We first characterize their behavior in pretrained SR backbones and show through controlled interventions that they strongly affect reconstruction quality. Building on this observation, we introduce SPARK, a lightweight input-conditioned controller that predicts bounded per-channel affine transformations for only the selected channels, while keeping the SR backbone and VAE frozen. Dominant channels are identified through an online activation-ranking procedure, and only a small predictor conditioned on the low-resolution VAE latent is optimized. Experiments on three DiT-based SR backbones across DIV2K, RealSR, and DRealSR show consistent gains in both fidelity and perceptual quality while modulating only eight channels per stream and block. Controlled comparisons further show that these gains cannot be explained by parameter budget or access to the selected channels alone.

Related papers

• BASC : Behavior-Aligned Quantization and Pruning for Low-Bit Spiking Neural Networks [6.5713207503171285]
  Spiking Neural Networks (SNNs) encode information through binary spikes and compute in an event-driven manner.<n>High-performance SNNs incur substantial memory and timestep-wise costs that hinder deployment on resource-constrained devices.<n>We propose Behavior-Aligned SNN Compression (BASC), a unified framework with two lightweight modules.
  arXiv  Detail & Related papers  (2026-08-12T03:39:24Z)
• TRaM-VSR: Importance-Aware Token Routing and Merging for One-Step Diffusion Video Super-Resolution [52.96050253456634]
  TRaM-VSR is a framework for adaptive token allocation using context-aware video priors and network-levels.<n>It is shown that TRaM-VSR accelerates inference significantly while preserving state-of-the-art reconstruction quality and robust temporal consistency.
  arXiv  Detail & Related papers  (2026-07-24T11:57:26Z)
• Lightweight CNN-Based Anomaly Detection for High Voltage Converter Modulators in the Spallation Neutron Source [0.788278500923222]
  Unscheduled trips of high-power pulsed converters are a leading source of downtime at large accelerator facilities.<n>We show that architectural inductive bias, specifically the ordering of temporal filtering and cross-channel mixing, plays a central role in detection performance.<n>Our best variant achieves a pooled AUC-PR of 0.816 and AUC-ROC of 0.934, outperforming the state of the art on most subsystems and five of the six fault families.
  arXiv  Detail & Related papers  (2026-05-29T12:54:13Z)
• Dual-Channel Attention Guidance for Training-Free Image Editing Control in Diffusion Transformers [11.772150619675527]
  Existing attention manipulation methods focus exclusively on the Key space to modulate attention routing.<n>We propose Dual-Channel Attention Guidance (DCAG) to simultaneously manipulate both the Key channel and the Value channel.<n>DCAG consistently outperforms Key-only guidance across all fidelity metrics.
  arXiv  Detail & Related papers  (2026-02-20T06:24:20Z)
• One-Step Diffusion Transformer for Controllable Real-World Image Super-Resolution [54.485573602613535]
  We present ODTSR, a one-step diffusion transformer that performs Real-ISR considering fidelity and controllability simultaneously.<n>ODTSR employs Fidelity-aware Adversarial Training (FAA) to enhance controllability and achieve one-step inference.<n>Experiments demonstrate that ODTSR achieves state-of-the-art (SOTA) performance on generic Real-ISR.
  arXiv  Detail & Related papers  (2025-11-21T11:00:59Z)
• Rethinking Diffusion Model-Based Video Super-Resolution: Leveraging Dense Guidance from Aligned Features [51.5076190312734]
  Video Super-Resolution approaches suffer from error accumulation, spatial artifacts, and a trade-off between perceptual quality and fidelity.<n>We propose a novelly Guided diffusion model with Aligned Features for Video Super-Resolution (DGAF-VSR)<n>Experiments on synthetic and real-world datasets demonstrate that DGAF-VSR surpasses state-of-the-art methods in key aspects of VSR.
  arXiv  Detail & Related papers  (2025-11-21T03:40:45Z)
• MSLoRA: Multi-Scale Low-Rank Adaptation via Attention Reweighting [6.335488846185043]
  MSLoRA is a backbone-agnostic, parameter-efficient adapter that reweights feature responses rather than re-tuning the underlying backbone.<n>MSLoRA unifies adaptation for both convolutional neural networks (CNNs) and vision transformers (ViTs)
  arXiv  Detail & Related papers  (2025-11-16T00:35:37Z)
• AdaFortiTran: An Adaptive Transformer Model for Robust OFDM Channel Estimation [22.40154714677385]
  We introduce the Adaptive Fortified Transformer (AdaFortiTran) to enhance channel estimation in challenging environments.<n>AdaFortiTran achieves up to 6 dB reduction in mean squared error (MSE) compared to state-of-the-art models.
  arXiv  Detail & Related papers  (2025-05-14T02:22:37Z)
• Rethinking Transformer-Based Blind-Spot Network for Self-Supervised Image Denoising [94.09442506816724]
  Blind-spot networks (BSN) have been prevalent neural architectures in self-supervised image denoising (SSID)<n>We build a Transformer-based Blind-Spot Network (TBSN) which shows strong local fitting and global perspective abilities.
  arXiv  Detail & Related papers  (2024-04-11T15:39:10Z)
• DAIS: Automatic Channel Pruning via Differentiable Annealing Indicator Search [55.164053971213576]
  convolutional neural network has achieved great success in fulfilling computer vision tasks despite large computation overhead. Structured (channel) pruning is usually applied to reduce the model redundancy while preserving the network structure. Existing structured pruning methods require hand-crafted rules which may lead to tremendous pruning space.
  arXiv  Detail & Related papers  (2020-11-04T07:43:01Z)
• Operation-Aware Soft Channel Pruning using Differentiable Masks [51.04085547997066]
  We propose a data-driven algorithm, which compresses deep neural networks in a differentiable way by exploiting the characteristics of operations. We perform extensive experiments and achieve outstanding performance in terms of the accuracy of output networks.
  arXiv  Detail & Related papers  (2020-07-08T07:44:00Z)
• Deep Denoising Neural Network Assisted Compressive Channel Estimation for mmWave Intelligent Reflecting Surfaces [99.34306447202546]
  This paper proposes a deep denoising neural network assisted compressive channel estimation for mmWave IRS systems. We first introduce a hybrid passive/active IRS architecture, where very few receive chains are employed to estimate the uplink user-to-IRS channels. The complete channel matrix can be reconstructed from the limited measurements based on compressive sensing.
  arXiv  Detail & Related papers  (2020-06-03T12:18:57Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.