The paper introduces IsoMamba, a Vision Mamba variant that addresses state contention in underwater image enhancement by organizing recurrent states according to token compatibility. It employs a cluster–route–isolate paradigm to route update‑compatible tokens into independent state streams and adds a dynamic channel‑wise reweighting objective to handle wavelength‑dependent reconstruction challenges. Experiments show IsoMamba outperforms existing methods on multiple underwater enhancement benchmarks while producing cleaner hidden state responses.
arXiv:2603. 00198v2 Announce Type: replace-cross Abstract: Token reduction accelerates long-video vision--language models (VLMs), but existing methods target Transformers, where reduction is treated as token pruning.
Loop‑Mamba is a lightweight, loop‑based state‑space framework designed for restoring old photographs that suffer from multiple degradations such as scratches, cracks, fading, blur, noise, and missing regions. It models restoration as progressive state evolution, using a Semantic‑Guided Degradation Estimator to predict local degradation maps and global scores, and a Shared Structural Memory Mamba to maintain a persistent restoration state across iterations. The method employs first‑order state recursion and a multi‑directional scanning strategy to reduce gradient dilution and computational overhead, and introduces the Old Photo Damage Recovery Score (ODRS) to evaluate both degradation recovery and structural reconstruction, achieving superior performance on the SynOld benchmark.
arXiv:2610.16330v1 Announce Type: cross Abstract: Feed-forward transformer models have driven remarkable progress in 3D vision,yet their quadratic complexity renders them impractical for long video s...
arXiv:2607. 12112v1 Announce Type: cross Abstract: Federated fine-tuning of Multimodal Large Language Models (MLLMs) across distributed networks enables privacy-sensitive adaptation to evolving data streams, yet a fundamental obstacle prevents robust deployment in dynamic environments: catastrophic forgetting, wherein sequential task updates erase previously acquired knowledge across visual, linguistic, and cross-modal representations.
The paper introduces reViT, a recurrent vision transformer that uses a single Transformer block repeatedly to achieve the accuracy of a full-depth encoder while keeping inference FLOPs comparable. It achieves depth‑specific transformations by expressing each block’s feed‑forward network as a convex combination of a small shared expert bank, with a continuous depth coordinate guiding the mixture. Experiments on ImageNet‑1k and DINOv2 distillation show that reViT matches or surpasses state‑of‑the‑art models, reduces stored parameters by about 70%, and supports elastic‑depth deployment from a single checkpoint.
arXiv:2603. 25157v2 Announce Type: replace Abstract: Recent vision and multimodal foundation backbones, such as Transformer families and state-space models like Mamba, have achieved remarkable progress, enabling unified modeling across images, text, and beyond.
LongVU‑TTT is a causal test‑time training method for long‑video multimodal large language models that inserts a convolutional resampler with fast‑weight updates between the vision encoder and the LLM. The fast weights adapt per video and contextualize frame features before compression, while a hybrid selector keeps explicit visual evidence for downstream reasoning. Experiments show that TTT‑Conv outperforms TTT‑MLP and bidirectional Mamba2 on MLVU, and beats attention‑ and fixed‑state recurrent resamplers on three benchmarks, achieving competitive results on five video‑understanding tasks after reducing 512 frames to 128 LLM frames.
arXiv:2609.21391v1 Announce Type: cross Abstract: Neural Radiance Fields (NeRF) have attracted extensive attention in recent years due to their strong capability for high-quality 3D reconstruction an...
InfoMamba is an attention‑free hybrid model that combines a minimal‑bandwidth global interface with a selective recurrent stream. The architecture replaces token‑level self‑attention with a concept bottleneck linear filtering layer and integrates it via an information‑maximizing fusion (IMF) that injects global context into the state‑space dynamics. Experiments across classification, dense prediction, and non‑vision tasks show that InfoMamba outperforms strong Transformer and SSM baselines while maintaining near‑linear scaling and competitive accuracy‑efficiency trade‑offs.
The paper introduces a method to retrofit depth recurrence into pretrained vision‑language models, reducing the language backbone’s parameter count by pruning layers and looping over the remaining ones with linear adapters. A three‑stage recovery training—language‑only pretraining, vision‑language projector alignment, and full fine‑tuning—helps recover performance, with distillation from the non‑recurrent model further improving results. Experiments on TinyLLaVA‑1.4B show the retrofitted recurrent model outperforms the unpruned baseline while saving 23.3% of language parameters and performs better on reasoning‑oriented benchmarks.
RegCL is a non‑replay continual learning framework that adapts the Segment Anything Model (SAM) for visual grounding across evolving multi‑sensorial media domains. It consolidates domain‑specific segmentation knowledge into a single lightweight SAM adapter by incrementally merging LoRA‑style AugModules and preserving compact historical feature statistics. Experiments on five heterogeneous datasets demonstrate that RegCL retains performance while adapting to new domains, outperforming other non‑replay continual learning and merging baselines.
arXiv:2606.06158v2 Announce Type: replace Abstract: Adaptive video tokenisation seeks to dynamically allocate token budgets based on the underlying visual complexity of a sequence. Current continuous...