图像生成 - arXivDaily 专题

2603.11703 2026-04-09 cs.LG 50%

EvoFlows: Evolutionary Edit-Based Flow-Matching for Protein Engineering

EvoFlows：基于进化编辑的蛋白质工程流匹配

Nicolas Deutschmann, Constance Ferragu, Jonathan D. Ziegler, Shayan Aziznejad, Eli Bixby

机构 * Cradle Zürich, CH（Cradle 苏黎世，瑞士）

专题命中图像编辑：diffusion(abstract)

AI总结 EvoFlows通过编辑流学习蛋白质序列间的突变轨迹，实现可控的插入、删除和替换操作，生成与自然蛋白家族一致的变异体，优于现有方法。

Comments Accepted at Workshop on Foundation Models for Science: Real-World Impact and Science-First Design, ICLR 2026

2603.07990 2026-03-25 cs.LG 50%

MJ1: Multimodal Judgment via Grounded Verification

MJ1: 通过 grounded 验证实现多模态判断

Bhavesh Kumar, Dylan Feng, Leonard Tang

机构 * Haize Labs（哈泽实验室）

专题命中图像编辑：image editing(abstract)

AI总结 MJ1 通过结构化 grounded 验证链和反事实一致性奖励提升多模态判断准确性，训练后在 MMRB2 数据集上达到 77.0% 准确率，超越更大模型。

2603.07148 2026-03-10 cs.LG 50%

Agentic Planning with Reasoning for Image Styling via Offline RL

基于推理的图像风格化代理规划 via 离线强化学习

Subhojyoti Mukherjee, Stefano Petrangeli, Branislav Kveton, Trung Bui, Franck Dernoncourt, Arko Mukherjee

机构 * Adobe Research（Adobe研究）

专题命中图像编辑：image editing(abstract)

AI总结本文提出基于推理的代理规划框架，通过结构化规划和工具链提升图像风格化效果，采用合成数据生成和离线强化学习方法，验证在视觉质量和指令遵循上的优越性。

Comments 85 pages

2510.08570 2026-02-23 cs.LG 50%

Who Said Neural Networks Aren't Linear?

谁说神经网络不是线性的？

Nimrod Berman, Assaf Hallak, Assaf Shocher

专题命中图像编辑：diffusion(abstract)

AI总结本文提出线性化器架构，通过结构运输概念使神经网络在特定向量空间中表现为线性算子，从而应用线性代数工具处理非线性映射，并展示其在扩散模型和风格迁移中的应用。

2510.06050 2026-02-05 cs.LG 50%

Edit-Based Flow Matching for Temporal Point Processes

基于编辑的流匹配用于时间点过程

David Lüdke, Marten Lienen, Marcel Kollovieh, Stephan Günnemann

机构 * School of Computation, Information and Technology & Munich Data Science Institute（计算、信息与技术学院及慕尼黑数据科学研究所）

专题命中图像编辑：diffusion(abstract)

AI总结本文提出了一种基于编辑操作的流匹配方法，用于改进时间点过程的建模，通过连续时间马尔可夫链框架学习瞬时编辑率，提升生成效率和灵活性。

2602.01206 2026-02-03 cs.AI 50%

Addressing Explainability of Generative AI using SMILE (Statistical Model-agnostic Interpretability with Local Explanations)

通过SMILE（统计模型无关可解释性与局部解释）解决生成AI的可解释性问题

Zeinab Dehghani

机构 * School of Computer Science（计算机科学学院）； University of Hull（赫尔大学）

专题命中图像编辑：image editing(abstract)

AI总结 gSMILE通过受控扰动、Wasserstein距离和加权替代模型，提升生成AI的可解释性，实现细粒度归因和直观热图生成。

2601.22408 2026-02-02 q-bio.BM cs.LG 50%

Minimal-Action Discrete Schrödinger Bridge Matching for Peptide Sequence Design

最小作用离散薛定谔桥匹配用于肽序列设计

Shrey Goel, Pranam Chatterjee

机构 * Duke University（杜克大学）； University of Pennsylvania（宾夕法尼亚大学）

专题命中图像编辑：diffusion(abstract)

AI总结本文提出MadSBM框架，通过最小作用离散薛定谔桥匹配方法，实现肽序列设计的高效生成与优化。

2511.12174 2025-11-18 cs.LG 50%

TSGDiff: Rethinking Synthetic Time Series Generation from a Pure Graph Perspective

Lifeng Shen, Xuyang Li, Lele Long

专题命中图像编辑：diffusion(abstract)

Comments Accepted by AAAI 2026

2509.07627 2025-10-09 cs.CE q-bio.BM 50%

LSMTCR: A Scalable Multi-Architecture Model for Epitope-Specific T Cell Receptor de novo Design

Ruihao Zhang, Xiao Liu

专题命中图像编辑：diffusion(abstract)

Comments 13 main pages, 5 figures, 2 tables

2510.04738 2025-10-07 cs.SD cs.AI cs.CL cs.LG eess.AS 50%

Speak, Edit, Repeat: High-Fidelity Voice Editing and Zero-Shot TTS with Cross-Attentive Mamba

Baher Mohammad, Magauiya Zhussip, Stamatios Lefkimmiatis

机构 * MTS AI（MTS人工智能公司）； ITMO University（ITMO大学）

专题命中图像编辑：diffusion(abstract)

2509.02109 2025-09-30 cs.LG math.PR stat.ML 50%

Differentiable Expectation-Maximisation and Applications to Gaussian Mixture Model Optimal Transport

Samuel Boïté, Eloi Tanguy, Julie Delon, Agnès Desolneux, Rémi Flamary

机构 * Université Paris Cité, CNRS, MAP5（巴黎大学、国家科学研究中心、MAP5）； CNRS（国家科学研究中心）； ENS Paris-Saclay（巴黎-萨克雷大学）； ECOLE POLYTECHNIQUE（巴黎高等理工学院）； Institut Polytechnique de Paris（巴黎理工学院）

专题命中图像编辑：image generation(abstract)

2507.01456 2025-07-03 math.GN 50%

QC-OT: Optimal Transport with Quasiconformal Mapping

Yuping Lv, Qi Zhao, Xuebin Chang, Wei Zeng

专题命中图像编辑：image editing(abstract)

Comments 24 pages,18 figures

2411.03477 2025-04-22 cs.HC 50%

CrowdGenUI: Aligning LLM-Based UI Generation with Crowdsourced User Preferences

Yimeng Liu, Misha Sra, Chang Xiao

专题命中图像编辑：image editing(abstract)

2407.03992 2025-03-20 eess.IV 50%

Medical Image Fusion for High-Level Analysis: A Mutual Enhancement Framework for Unaligned PAT and MRI

Yutian Zhong, Jinchuan He, Zhichao Liang, Shuangyang Zhang, Qianjin Feng, Lijun Lu, Li Qi

专题命中图像编辑：image generation(abstract)

2409.17410 2025-01-31 cs.HC 50%

Copying style, Extracting value: Illustrators' Perception of AI Style Transfer and its Impact on Creative Labor

Julien Porquet, Sitong Wang, Lydia B. Chilton

专题命中图像编辑：text-to-image(abstract)

2501.11233 2025-01-22 cs.IR cs.CL cs.MA 50%

PlotEdit: Natural Language-Driven Accessible Chart Editing in PDFs via Multimodal LLM Agents

Kanika Goswami, Puneet Mathur, Ryan Rossi, Franck Dernoncourt

专题命中图像编辑：image editing(abstract)

Comments Accepted at ECIR 2025

2410.11908 2024-10-17 cs.HC cs.AI 50%

ChatHouseDiffusion: Prompt-Guided Generation and Editing of Floor Plans

Sizhong Qin, Chengyu He, Qiaoyun Chen, Sen Yang, Wenjie Liao, Yi Gu, Xinzheng Lu

专题命中图像编辑：diffusion(abstract)

2410.01971 2024-10-04 cs.RO cs.LG 50%

Run-time Observation Interventions Make Vision-Language-Action Models More Visually Robust

Asher J. Hancock, Allen Z. Ren, Anirudha Majumdar

专题命中图像编辑：image editing(abstract)

Comments Website: https://aasherh.github.io/byovla/

2409.06082 2024-09-17 cs.HC 50%

MemoVis: A GenAI-Powered Tool for Creating Companion Reference Images for 3D Design Feedback

Chen Chen, Cuong Nguyen, Thibault Groueix, Vladimir G. Kim, Nadir Weibel

专题命中图像编辑：image editing(abstract)

Comments In the Journal of ACM Transactions on Computer-Human Interaction

Journal ref ACM Transactions on Computer-Human Interaction, 2024

2408.14170 2024-08-27 eess.IV 50%

Image Provenance Analysis via Graph Encoding with Vision Transformer

Keyang Zhang, Chenqi Kong, Shiqi Wang, Anderson Rocha, Haoliang Li

专题命中图像编辑：image editing(abstract)

Comments 13 pages, 10 figures

2407.16732 2024-08-06 cs.SE cs.AI 50%

PyBench: Evaluating LLM Agent on various real-world coding tasks

Yaolun Zhang, Yinxu Pan, Yudong Wang, Jie Cai

专题命中图像编辑：image editing(abstract)

Comments 16 pages

2311.02247 2023-11-07 cs.LG 50%

PRISM: Progressive Restoration for Scene Graph-based Image Manipulation

Pavel Jahoda, Azade Farshad, Yousef Yeganeh, Ehsan Adeli, Nassir Navab

专题命中图像编辑：image generation(abstract)

2212.02953 2023-04-11 eess.IV 50%

Using Decoupled Features for Photo-realistic Style Transfer

Trevor D. Canham, Adrián Martín, Marcelo Bertalmío, Javier Portilla

专题命中图像编辑：diffusion(abstract)

2302.00079 2023-02-02 cs.HC cs.LG 50%

GANravel: User-Driven Direction Disentanglement in Generative Adversarial Networks

Noyan Evirgen, Xiang 'Anthony' Chen

专题命中图像编辑：image editing(abstract)

2112.09164 2022-08-17 cs.LG cs.AI 50%

High Fidelity Visualization of What Your Self-Supervised Representation Knows About

Florian Bordes, Randall Balestriero, Pascal Vincent

专题命中图像编辑：diffusion(abstract)

Comments Accepted at TMLR 2022

2207.08320 2022-08-16 cs.HC cs.AI cs.LG 50%

GANzilla: User-Driven Direction Discovery in Generative Adversarial Networks

Noyan Evirgen, Xiang 'Anthony' Chen

专题命中图像编辑：image editing(abstract)

2109.11139 2021-09-24 eess.IV 50%

Source Printer Identification using Printer Specific Pooling of Letter Descriptors

Sharad Joshi, Yogesh Kumar Gupta, Nitin Khanna

专题命中图像编辑：image editing(abstract)

Comments 34 pages, 5 figures, Journal

2006.06900 2020-11-02 cs.LG cs.CL stat.ML 50%

Improving GAN Training with Probability Ratio Clipping and Sample Reweighting

Yue Wu, Pan Zhou, Andrew Gordon Wilson, Eric P. Xing, Zhiting Hu

专题命中图像编辑：image generation(abstract)

Comments NeurIPS 2020 camera ready version (citations updated)

1912.01046 2020-06-02 cs.CL 50%

TutorialVQA: Question Answering Dataset for Tutorial Videos

Anthony Colas, Seokhwan Kim, Franck Dernoncourt, Siddhesh Gupte, Daisy Zhe Wang, Doo Soon Kim

专题命中图像编辑：image editing(abstract)

Comments Accepted at LREC 2020

1808.08703 2020-05-19 cs.CL cs.AI cs.LG 50%

Generating Text through Adversarial Training using Skip-Thought Vectors

Afroz Ahamad

专题命中图像编辑：image generation(abstract)

Comments NAACL 2019: https://www.aclweb.org/anthology/N19-3008

Journal ref "Proceedings of the 2019 Conference of the North {A}merican Chapter of the Association for Computational Linguistics: Student Research Workshop, Jun 2019, Pages 53-60"

1. 图像编辑 1264 篇

EvoFlows: Evolutionary Edit-Based Flow-Matching for Protein Engineering

MJ1: Multimodal Judgment via Grounded Verification

Agentic Planning with Reasoning for Image Styling via Offline RL

Who Said Neural Networks Aren't Linear?

Edit-Based Flow Matching for Temporal Point Processes

Addressing Explainability of Generative AI using SMILE (Statistical Model-agnostic Interpretability with Local Explanations)

Minimal-Action Discrete Schrödinger Bridge Matching for Peptide Sequence Design

TSGDiff: Rethinking Synthetic Time Series Generation from a Pure Graph Perspective

LSMTCR: A Scalable Multi-Architecture Model for Epitope-Specific T Cell Receptor de novo Design

Speak, Edit, Repeat: High-Fidelity Voice Editing and Zero-Shot TTS with Cross-Attentive Mamba

Differentiable Expectation-Maximisation and Applications to Gaussian Mixture Model Optimal Transport

QC-OT: Optimal Transport with Quasiconformal Mapping

CrowdGenUI: Aligning LLM-Based UI Generation with Crowdsourced User Preferences

Medical Image Fusion for High-Level Analysis: A Mutual Enhancement Framework for Unaligned PAT and MRI

Copying style, Extracting value: Illustrators' Perception of AI Style Transfer and its Impact on Creative Labor

PlotEdit: Natural Language-Driven Accessible Chart Editing in PDFs via Multimodal LLM Agents

ChatHouseDiffusion: Prompt-Guided Generation and Editing of Floor Plans

Run-time Observation Interventions Make Vision-Language-Action Models More Visually Robust

MemoVis: A GenAI-Powered Tool for Creating Companion Reference Images for 3D Design Feedback

Image Provenance Analysis via Graph Encoding with Vision Transformer

PyBench: Evaluating LLM Agent on various real-world coding tasks

PRISM: Progressive Restoration for Scene Graph-based Image Manipulation

Using Decoupled Features for Photo-realistic Style Transfer

GANravel: User-Driven Direction Disentanglement in Generative Adversarial Networks

High Fidelity Visualization of What Your Self-Supervised Representation Knows About

GANzilla: User-Driven Direction Discovery in Generative Adversarial Networks

Source Printer Identification using Printer Specific Pooling of Letter Descriptors

Improving GAN Training with Probability Ratio Clipping and Sample Reweighting

TutorialVQA: Question Answering Dataset for Tutorial Videos

Generating Text through Adversarial Training using Skip-Thought Vectors