English(EN)EditCaption: Human-Refined SFT and HAE-DPO for Image Editing Instruction Synthesis
新AI研究聚焦多模态微调、图像编辑和验证
作者PulseAugur 编辑部·[24 个来源]·
研究人员开发了TRACER，一种新颖的鲁棒多模态微调方法，通过使用加权移动平均（WMA）教师来解决灾难性遗忘问题。该方法提高了CLIP等模型在分布外准确性和校准性。另外，OmniVerifier-M1引入了一种多模态元验证器，它使用符号输出来实现更可靠和细粒度的基础模型验证。此外，BlazeEdit提供了一种高效、紧凑的图像到图像扩散模型，用于设备端编辑，而Alterbute则能够编辑内在对象属性（如颜色和形状），同时保持身份不变。
AI
arXiv:2605.29380v1 Announce Type: cross Abstract: Mainstream strategies for finetuning pretrained multimodal models often degrade out-of-distribution (OOD) robustness, a phenomenon known as catastrophic forgetting. In this paper, we develop a theoretical framework for multimodal …
Mainstream strategies for finetuning pretrained multimodal models often degrade out-of-distribution (OOD) robustness, a phenomenon known as catastrophic forgetting. In this paper, we develop a theoretical framework for multimodal contrastive finetuning, yielding closed-form solut…
arXiv:2605.28649v1 Announce Type: cross Abstract: LLMs increasingly require surgical model editing to enhance domain-specific capabilities without incurring the computational cost or catastrophic forgetting associated with full fine-tuning. Sparse Autoencoders (SAEs) have emerged…
arXiv:2605.28067v1 Announce Type: new Abstract: The remarkable generation quality of modern diffusion models often comes at the cost of massive parameter counts, which necessitate server-side inference with significant computational costs and potential privacy risks. Consequently…
arXiv:2605.28805v1 Announce Type: cross Abstract: Visual outcomes are increasingly central to multimodal large language models, making reliable and fine-grained verification essential for scaling generalist foundation models. In this work, we investigate multimodal meta-verificat…
LLMs increasingly require surgical model editing to enhance domain-specific capabilities without incurring the computational cost or catastrophic forgetting associated with full fine-tuning. Sparse Autoencoders (SAEs) have emerged as a promising tool in this setting, in principle…
Multimodal meta-verification using symbolic rationales and decoupled reinforcement learning enables robust visual verification and fine-grained error localization in generalist foundation models.
arXiv:2604.18170v2 Announce Type: replace-cross Abstract: LLMs edit text and code by autoregressively regenerating the full output, even when most tokens appear verbatim in the input. We study Copy-as-Decode, a decoding-layer mechanism that recasts edit generation as structured d…
arXiv:2604.08213v2 Announce Type: replace-cross Abstract: High-quality source-target image pairs with precise editing instructions are essential for instruction-guided image editing, yet constructing such training triplets at scale remains costly. Recent pipelines often rely on v…
arXiv cs.CV
TIER_1English(EN)·Tal Reiss, Daniel Winter, Matan Cohen, Alex Rav-Acha, Yael Pritch, Ariel Shamir, Yedid Hoshen·
arXiv:2601.10714v2 Announce Type: replace Abstract: We introduce Alterbute, a diffusion-based method for editing an object's intrinsic attributes in an image. We allow changing color, texture, material, and even the shape of an object, while preserving its perceived identity and …
Visual outcomes are increasingly central to multimodal large language models, making reliable and fine-grained verification essential for scaling generalist foundation models. In this work, we investigate multimodal meta-verification, which leverages verifier-generated rationales…
arXiv cs.CV
TIER_1English(EN)·Han Zou, Yan Zhang, Ruiqi Yu, Cong Xie, Jie Huang, Zhenpeng Zhan·
arXiv:2512.14140v2 Announce Type: replace Abstract: Sketch editing requires jointly handling high-level semantic changes and precise local redrawing, a combination that is particularly challenging for sparse, style-sensitive line art. Unlike natural images, sketches rely on minim…
arXiv:2605.14664v2 Announce Type: replace Abstract: Reference-guided video editing takes a source video, a text instruction, and a reference image as inputs, requiring the model to faithfully apply the instructed edits while preserving original motion and unedited content. Existi…
arXiv cs.CV
TIER_1English(EN)·Yuanye Liu, Siyuan Zhou, Ke Zhang, Lei Li, Wei Chen, Xiahai Zhuang·
arXiv:2605.24932v1 Announce Type: new Abstract: Pre-trained Vision Transformers (ViTs) are increasingly deployed for medical image classification. However, correcting their inevitable failure cases in dynamic clinical scenarios poses a critical challenge. Conventional fine-tuning…
arXiv cs.CV
TIER_1English(EN)·Yuke Li, Lianli Gao, Ji Zhang, Pengpeng Zeng, Lichuan Xiang, Hongkai Wen, Heng Tao Shen, Jingkuan Song·
arXiv:2512.01382v4 Announce Type: replace Abstract: Exemplar-guided Image Editing (EIE) aims to modify a source image according to a visual reference. Existing approaches often require large-scale pre-training to learn relationships between the source and reference images, incurr…
arXiv cs.CV
TIER_1English(EN)·Mingyi Xu, Jinpeng Lin, Min Zhou, Tiezheng Ge, Ming Zeng·
arXiv:2605.25568v1 Announce Type: new Abstract: Scribble-guided image editing allows users to combine simple scribble annotations with text prompts to specify both where and how an image should be edited, enabling flexible interaction with precise spatial control. However, existi…
arXiv:2605.24805v1 Announce Type: new Abstract: Large-scale controllable 3D assets are critical for computer graphics, embodied AI, robotics, and interactive content creation, yet creating diverse 3D assets remains challenging due to the high cost of manual modeling and rigging. …
arXiv:2605.23518v1 Announce Type: new Abstract: Directly editing ultra-high-resolution (UHR) images is valuable but underexplored, primarily due to the lack of high-quality data and the challenge in modeling high-frequency texture details. We introduce VINS-120K, the first large-…
Directly editing ultra-high-resolution (UHR) images is valuable but underexplored, primarily due to the lack of high-quality data and the challenge in modeling high-frequency texture details. We introduce VINS-120K, the first large-scale dataset for instruction-based UHR image ed…
arXiv:2602.00122v2 Announce Type: replace Abstract: In recent years, image editing models have made significant progress, enabling users to manipulate visual content in a flexible and interactive manner through natural language instructions. However, an important yet underexplore…
arXiv:2602.01851v2 Announce Type: replace Abstract: Recent generative models have achieved remarkable progress in image editing. However, existing systems and benchmarks remain largely text-guided. In contrast, human communication is inherently multimodal, where visual instructio…
