The paper introduces Meta‑Diff, a method that learns a reusable source‑to‑target relation in a compact task space, enabling rapid adaptation of a diffusion model to new targets with minimal or no target data. By encoding each distribution into a low‑dimensional coordinate and transporting a source coordinate to its predicted target, the approach avoids costly fine‑tuning and achieves significant performance gains on CIFAR‑100 and Office‑Home benchmarks. Experiments show that when task identity is preserved, the relation generalizes across tasks, and the source‑target shift is manageable, Meta‑Diff can outperform target‑only modeling and full fine‑tuning.
whyItMatters:"The method offers a practical solution for few‑shot adaptation in data‑scarce settings, such as healthcare, by reducing computational cost and improving reliability compared to traditional retraining approaches."
By Jinjian Jiang, Jing Peng, Rehan Deen, Yu Cao, Ziquan Liu, Muhammad Salman Haleem, David Mguni
arXiv:2605. 31191v2 Announce Type: replace Abstract: We investigate how teacher-student capacity relationships modulate knowledge distillation (KD) effectiveness in ResNet-based image classification on CIFAR-10.
By Umut Onur Yasar
Low-light image enhancement (LLIE) methods involve tunable parameters that are typically fixed, often leading to performance degradation when applied across scenes. Manually selecting the best configuration, however, can be time-consuming and not always practical.
arXiv:2610.34092v1 Announce Type: new
Abstract: Few-step distillation for autoregressive (AR) video generation commonly first initializes an AR student and then refines it with distribution matching...
By Zixuan Duan, Yabo Chen, Xunzhi Xiang, Xin Zhang, Haibin Huang, Yin Chen, Chi Zhang, Qi Fan
PairCLIP‑SWA is a single‑backbone deepfake detector that fine‑tunes an OpenCLIP ViT‑L/14 visual tower using a class‑balanced corpus created by CLIP‑similarity‑guided subsampling and then applies stochastic weight averaging over the fine‑tuning trajectory. Trained on the IEEE SP Cup 2025 (DF‑Wild‑Cup) benchmark, the weight‑averaged model achieves 97.04 % accuracy, 98.64 % AU‑ROC and a 4.07 % equal error rate—outperforming the best published three‑backbone ensemble. The study shows that averaging stabilizes validation performance, prevents misranking on disjoint test sets, and that CLIP pre‑training, rather than ensembling, is the key driver of accuracy.
By Md. Nakib, Sudipto Sarkar, A. J. A. Jubayer Talukder, Mohammad Ariful Haque
arXiv:2610.11146v3 Announce Type: replace
Abstract: Training-only teacher access lets a compact student deploy with local features and exported weights. This established distillation interface does n...
By Qiye Lu, Jiang Ji, Liang Zhang
arXiv:2608. 09091v1 Announce Type: cross Abstract: Transfer learning is particularly useful in settings with limited training data, and within image classification it is common to transfer learn upon massive datasets like ImageNet , CIFAR-100, or COCO .
By Jing Ning, James D. Braza
Scout-Refine-Correct (SRC) is a training recipe that equips flow-matching image generation models with an inference-time refinement capability. During sampling, the model performs a scout rollout to predict a full sample, uses this prediction as a self-conditioning signal, and then takes the actual step; this process can be repeated for multiple refinement rounds. On ImageNet-256, SRC improves LightningDiT-B’s FID from ~14.5 to ~7.26 with 550 normalised flow evaluations, matching a best-of‑N baseline that relies on an oracle verifier, and further reduces FID to 3.81 when combined with best-of‑N.
By Sergey Karpukhin, Aibek Alanov
The paper introduces TALON, a Task‑Adaptive LoRA‑Teacher framework for Few‑Shot Class‑Incremental Learning. TALON assigns a dedicated LoRA‑Teacher to each incremental task, then distills the frozen teachers into a single LoRA‑Student via Ensemble Knowledge Transfer, using a semantic‑guided weighting scheme to reduce forgetting and overfitting. Experiments on four FSCIL benchmarks show that TALON matches or surpasses state‑of‑the‑art accuracy while using up to 33× fewer deployment parameters and cutting inference time by 41.7%.
By Hongwei Zhao (School of Computer Science,Engineering, Beihang University), Rui Liu (School of Computer Science,Engineering, Beihang University), Yansong Liu (School of Computer Science,Engineering, Beihang University), Zhiyuan Zou (School of Computer Science,Engineering, Beihang University), Yong Chen (School of Computer Science, Beijing University of Posts,Telecommunications)
arXiv:2607. 16888v1 Announce Type: cross Abstract: Deploying a medical imaging model that must later accommodate a modality it has never seen is a recurring practical problem: retraining the shared representation is expensive and destroys performance on the modalities already in service.
By Ranat Das Prangon, Istiaque Ahmed, Shajid Hasan Naim, Waseem Mustak Zisan, Hossain Md Shakhawat
The paper introduces VDOT++, a unified distillation framework for few-step video generation across text-to-video, image-to-video, and condition-based tasks. It extends the previous VDOT method by incorporating an asymmetric unbalanced optimal transport distillation that allows student tokens to carry less mass, and an α1 ground cost that uses a weighted median to preserve modes. The approach combines distribution matching with adversarial refinement and cross-scale distillation, achieving competitive performance on multiple video benchmarks with only four sampling steps.
By Yutong Wang, Xingtong Ge, Enhuai Liu, Yunke Wang, Tianfan Xue, Yu Qiao, Yaohui Wang, Xinyuan Chen, Chang Xu
arXiv:2610.02516v1 Announce Type: cross
Abstract: Zero-shot classifiers are useful for routing user requests to specialized LLM tasks, but scoring every request against a large candidate set is expen...
By Haifeng Wu, Srinivasan Manoharan, Jian Wan, Fangbo Tu, Junhua Zhao, Xin Chen