{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T20:40:58Z","timestamp":1782938458210,"version":"3.54.5"},"reference-count":133,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"7","license":[{"start":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T00:00:00Z","timestamp":1782864000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T00:00:00Z","timestamp":1782864000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T00:00:00Z","timestamp":1782864000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Artif. Intell."],"published-print":{"date-parts":[[2026,7]]},"DOI":"10.1109\/tai.2025.3648376","type":"journal-article","created":{"date-parts":[[2025,12,25]],"date-time":"2025-12-25T18:29:15Z","timestamp":1766687355000},"page":"3653-3670","source":"Crossref","is-referenced-by-count":1,"title":["Methodologies for Diffusion Model Interpretability: A Systematic Review"],"prefix":"10.1109","volume":"7","author":[{"ORCID":"https:\/\/orcid.org\/0009-0006-0230-6873","authenticated-orcid":false,"given":"Tina","family":"Lakhani","sequence":"first","affiliation":[{"name":"School of Computer Science and Informatics, Cardiff University, Cardiff, U.K."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5123-9861","authenticated-orcid":false,"given":"Jing","family":"Wu","sequence":"additional","affiliation":[{"name":"School of Computer Science and Informatics, Cardiff University, Cardiff, U.K."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2094-5680","authenticated-orcid":false,"given":"Yu-Kun","family":"Lai","sequence":"additional","affiliation":[{"name":"School of Computer Science and Informatics, Cardiff University, Cardiff, U.K."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8968-9902","authenticated-orcid":false,"given":"Ze","family":"Ji","sequence":"additional","affiliation":[{"name":"School of Engineering, Cardiff University, Cardiff, U.K."}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","first-page":"2256","article-title":"Deep unsupervised learning using nonequilibrium thermodynamics","volume-title":"Proc. 32nd Int. Conf. Mach. Learn., Ser. Proc. Mach. Learn. Res.","volume":"37","author":"Sohl-Dickstein","year":"2015"},{"key":"ref2","article-title":"Denoising diffusion probabilistic models","author":"Ho","year":"2020"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2023.3261988"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1145\/3626235"},{"key":"ref5","article-title":"Diffusion models for reinforcement learning: A survey","author":"Zhu","year":"2023"},{"key":"ref6","article-title":"Auto-Encoding Variational Bayes","author":"Kingma","year":"2013"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1145\/3422622"},{"key":"ref8","article-title":"A survey on mechanistic interpretability for multi-modal foundation models","author":"Lin","year":"2025"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1706.03762"},{"key":"ref10","article-title":"GLIDE: Towards photorealistic image generation and editing with text-guided diffusion models","author":"Nichol","year":"2021"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01042"},{"key":"ref12","article-title":"Exploring the latent space of diffusion models directly through singular value decomposition","author":"Wang","year":"2025"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1145\/3746252.3761318"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.52202\/079017-1165"},{"key":"ref15","article-title":"Diffusion models already have a semantic latent space","volume-title":"Proc. 11th Int. Conf. Learn. Represent. (ICLR)","author":"Kwon","year":"2023"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/FG59268.2024.10581912"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1109\/ICAIIC64266.2025.10920714"},{"key":"ref18","first-page":"21696","article-title":"Variational diffusion models","volume":"34","author":"Kingma","year":"2021","journal-title":"Proc. Adv. Neural Inf. Process. Syst."},{"key":"ref19","first-page":"8780","article-title":"Diffusion models beat GANs on image synthesis","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"34","author":"Dhariwal","year":"2021"},{"key":"ref20","first-page":"6629","article-title":"GANs trained by a two time-scale update rule converge to a local Nash equilibrium","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"30","author":"Heusel","year":"2017"},{"key":"ref21","article-title":"Denoising diffusion implicit models","volume-title":"Proc. Int. Conf. Learn. Representations (ICLR)","author":"Song","year":"2021"},{"key":"ref22","article-title":"Latent consistency models: Synthesizing high-resolution images with few-step inference","author":"Luo","year":"2023"},{"key":"ref23","article-title":"Lcm-Lora: A universal stable-diffusion acceleration module","author":"Luo","year":"2023"},{"key":"ref24","article-title":"Zero-shot text-to-image generation","author":"Ramesh","year":"2021"},{"key":"ref25","first-page":"8599","article-title":"Grad-TTS: A diffusion probabilistic model for text-to-speech","volume-title":"Proc. 38th Int. Conf. Mach. Learn., Ser. Proc. Mach. Learn. Res.","volume":"139","author":"Popov","year":"2021"},{"key":"ref26","article-title":"Leveraging content-based features from multiple acoustic models for singing voice conversion","author":"Zhang","year":"2023"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/SLT61566.2024.10832255"},{"key":"ref28","first-page":"4195","article-title":"Scalable diffusion models with transformers","volume-title":"Proc. IEEE\/CVF Int. Conf. Comput. Vis. (ICCV)","author":"Peebles","year":"2022"},{"key":"ref29","article-title":"An image is worth 16x16 words: Transformers for image recognition at scale","volume-title":"Proc. Int. Conf. Learn. Represent. (ICLR)","author":"Dosovitskiy","year":"2021"},{"key":"ref30","article-title":"Generative modeling by estimating gradients of the data distribution","volume":"32","author":"Song","year":"2019","journal-title":"Proc. Adv. Neural Inf. Process. Syst."},{"key":"ref31","article-title":"Score-based generative modeling through stochastic differential equations","author":"Song","year":"2020"},{"key":"ref32","first-page":"1","article-title":"Energy-based models for continual learning","volume-title":"Proc. Conf. Lifelong Learn. Agents","author":"Li","year":"2022"},{"key":"ref33","article-title":"Learning energy-based models by diffusion recovery likelihood","author":"Gao","year":"2020"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1145\/3236386.3241340"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1109\/DSAA.2018.00018"},{"key":"ref36","article-title":"Towards a rigorous science of interpretable machine learning","author":"Doshi-Velez","year":"2017"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v38i19.30095"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1789"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1109\/ISDFS60797.2024.10527277"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1201\/b19467"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1136\/BMJ.N71"},{"key":"ref42","article-title":"Systematic and scoping review protocol template","year":"2020"},{"key":"ref43","article-title":"IEEE Xplore digital library"},{"key":"ref44","article-title":"Scopus"},{"key":"ref45","article-title":"Arxiv.org e-print archive"},{"key":"ref46","article-title":"Schloss Dagstuhl Leibniz Center for Informatics,"},{"key":"ref47","article-title":"Elicit: The AI research assistant","year":"2023"},{"key":"ref48","volume-title":"Python 3 Reference Manual.","author":"Van Rossum","year":"2009"},{"key":"ref49","article-title":"Guidelines for performing systematic literature reviews in software engineering","author":"Kitchenham"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01118"},{"key":"ref51","first-page":"24166","article-title":"Analyzing diffusion as serial reproduction","volume-title":"Proc. 40th Int. Conf. Mach. Learn. (ICML)","volume":"202","author":"Marjieh","year":"2023"},{"key":"ref52","article-title":"The journey, not the destination: How data guides diffusion models","author":"Georgiev","year":"2023"},{"key":"ref53","article-title":"Cross-attention head position patterns can align with human visual concepts in text-to-image generative models","volume-title":"Proc. 13th Int. Conf. Learn. Represent. (ICLR)","author":"Park","year":"2025"},{"key":"ref54","article-title":"Interpretable diffusion via information decomposition","volume-title":"Proc. 12th Int. Conf. Learn. Represent. (ICLR)","author":"Kong","year":"2024"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1185"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.1111\/cgf.70301"},{"key":"ref57","doi-asserted-by":"publisher","DOI":"10.52202\/079017-0868"},{"key":"ref58","article-title":"Unraveling the temporal dynamics of the UNet in diffusion models","volume-title":"Proc. Deployable AI@ AAAI","author":"Prasad","year":"2023"},{"key":"ref59","article-title":"Generalization in diffusion models arises from geometry-adaptive harmonic representations","volume-title":"Proc. 12th Int. Conf. Learn. Represent. (ICLR)","author":"Kadkhodaie","year":"2024"},{"key":"ref60","article-title":"ToddlerDiffusion: Interactive structured image generation with cascaded Schr\u00f6dinger bridge","volume-title":"Proc. 13th Int. Conf. Learn. Represent. (ICLR)","author":"Bakr","year":"2024"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2897"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-024-54281-3"},{"key":"ref63","doi-asserted-by":"publisher","DOI":"10.52202\/068431-1901"},{"key":"ref64","article-title":"Interpreting and improving diffusion models from an optimization perspective","author":"Permenter","year":"2023"},{"key":"ref65","article-title":"In search of dispersed memories: Generative diffusion models are associative memory networks","volume-title":"Proc. Associative Memory {&} Hopfield Netw.","author":"Ambrogioni","year":"2023"},{"key":"ref66","first-page":"27474","article-title":"Critical windows: Non-asymptotic theory for feature emergence in diffusion models","volume-title":"Proc. 41st Int. Conf. Mach. Learn., Ser. Proc. Mach. Learn. Res.","volume":"235","author":"Li","year":"2024"},{"key":"ref67","doi-asserted-by":"publisher","DOI":"10.1109\/iccv51701.2025.00443"},{"key":"ref68","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.2408799121"},{"key":"ref69","first-page":"36336","article-title":"InfoDiffusion: Representation learning using information maximizing diffusion models","volume-title":"Proc. 40th Int. Conf. Mach. Learn. (ICML)","author":"Wang","year":"2023"},{"key":"ref70","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00713"},{"key":"ref71","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-31975-4_1"},{"key":"ref72","doi-asserted-by":"publisher","DOI":"10.1109\/WACV61041.2025.00351"},{"key":"ref73","article-title":"Diffusion-based generation of neural activity from disentangled latent codes","author":"McCart","year":"2024"},{"key":"ref74","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00710"},{"key":"ref75","first-page":"172","article-title":"Concept sliders: LoRA adaptors for precise control in diffusion models","author":"Gandikota","year":"2024","journal-title":"Proc. Eur. Conf. Comput. Vis. (ECCV)"},{"key":"ref76","first-page":"6808","article-title":"Prompting hard or hardly prompting: Prompt inversion for text-to-image diffusion models","volume-title":"Proc. IEEE\/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR)","author":"Mahajan","year":"2023"},{"key":"ref77","doi-asserted-by":"publisher","DOI":"10.52202\/079017-2423"},{"key":"ref78","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1048"},{"key":"ref79","article-title":"Unsupervised discovery of semantic latent directions in diffusion models","author":"Park","year":"2023"},{"key":"ref80","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00191"},{"key":"ref81","doi-asserted-by":"publisher","DOI":"10.52202\/075280-3424"},{"key":"ref82","doi-asserted-by":"publisher","DOI":"10.1016\/j.eswa.2024.123231"},{"key":"ref83","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1102"},{"key":"ref84","article-title":"Beyond surface statistics: Scene representations in a latent diffusion model","author":"Chen","year":"2023"},{"key":"ref85","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01389"},{"key":"ref86","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1535"},{"key":"ref87","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00189"},{"key":"ref88","first-page":"416","article-title":"Viewpoint textual inversion: Discovering scene representations and 3D view control in 2D diffusion models","volume-title":"Proc. Eur. Conf. Comput. Vis. (ECCV)","author":"Burgess","year":"2024"},{"key":"ref89","article-title":"TIDE: Temporal-aware sparse autoencoders for interpretable diffusion transformers in image generation","author":"Huang","year":"2025"},{"key":"ref90","article-title":"Label-efficient semantic segmentation with diffusion models","author":"Baranchuk","year":"2021"},{"key":"ref91","article-title":"Concept bottleneck generative models","volume-title":"Proc. 12th Int. Conf. Learn. Represent. (ICLR)","author":"Ismail","year":"2024"},{"key":"ref92","first-page":"21548","article-title":"Cones: Concept neurons in diffusion models for customized generation","volume-title":"Proc. 40th Int. Conf. Mach. Learn. (ICML)","volume":"202","author":"Liu","year":"2023"},{"key":"ref93","doi-asserted-by":"publisher","DOI":"10.52202\/075280-3026"},{"key":"ref94","article-title":"Unsupervised discovery of interpretable directions in h-space of pre-trained diffusion models","author":"Zhang","year":"2023"},{"key":"ref95","article-title":"The hidden language of diffusion models","volume-title":"Proc. 12th Int. Conf. Learn. Represent. (ICLR)","author":"Chefer","year":"2024"},{"key":"ref96","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.02285"},{"key":"ref97","article-title":"A generative framework for self-supervised facial representation learning","author":"He","year":"2024"},{"key":"ref98","article-title":"Enabling local editing in diffusion models by joint and individual component analysis","author":"Kouzelis","year":"2024","journal-title":"Proc. The Brit. Mach. Vis. Conf. (BMVC)"},{"key":"ref99","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01141"},{"key":"ref100","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-short.24"},{"key":"ref101","first-page":"371","article-title":"Unleashing the potential of the semantic latent space in diffusion models for image dehazing","volume-title":"Proc. Eur. Conf. Comput. Vis. (ECCV)","author":"Yang","year":"2024"},{"key":"ref102","article-title":"Decoding diffusion: A scalable framework for unsupervised analysis of latent space biases and representations using natural language prompts","author":"Zeng","year":"2024"},{"key":"ref103","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51701.2025.01484"},{"key":"ref104","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52734.2025.00767"},{"key":"ref105","doi-asserted-by":"publisher","DOI":"10.52202\/079017-4363"},{"key":"ref106","first-page":"5644","article-title":"What the DAAM: Interpreting stable diffusion using cross attention","volume-title":"Proc. 61st Annu. Meeting Assoc. Comput. Linguistics (Volume 1: Long Papers)","author":"Tang","year":"2022"},{"key":"ref107","doi-asserted-by":"publisher","DOI":"10.1016\/j.neucom.2025.129591"},{"key":"ref108","article-title":"P+: Extended textual conditioning in text-to-image generation","author":"Voynov","year":"2023"},{"key":"ref109","article-title":"Localizing and editing knowledge in text-to-image generative models","volume-title":"Proc. 12th Int. Conf. Learn. Represent. (ICLR)","author":"Basu","year":"2024"},{"key":"ref110","first-page":"3224","article-title":"On mechanistic knowledge localization in text-to-image generative models","volume-title":"Proc. 41st Int. Conf. Mach. Learn. (ICML)","author":"Basu","year":"2024"},{"key":"ref111","doi-asserted-by":"publisher","DOI":"10.52202\/079017-0066"},{"key":"ref112","article-title":"ConceptAttention: Diffusion transformers learn highly interpretable features","author":"Helbling","year":"2025"},{"key":"ref113","doi-asserted-by":"publisher","DOI":"10.1109\/WIFS61860.2024.10810680"},{"key":"ref114","article-title":"Training data attribution for diffusion models","author":"Dai","year":"2023"},{"key":"ref115","article-title":"Detecting, explaining, and mitigating memorization in diffusion models","volume-title":"Proc. 12th Int. Conf. Learn. Represent. (ICLR)","author":"Wen","year":"2024"},{"key":"ref116","article-title":"Data attribution for diffusion models: Timestep-induced bias in influence estimation","volume":"2024","author":"Xie","year":"2024","journal-title":"Trans. Mach. Learn. Res."},{"key":"ref117","article-title":"Influence functions for scalable data attribution in diffusion models","volume-title":"Proc. 13th Int. Conf. Learn. Represent (ICLR)","author":"Mlodozeniec","year":"2025"},{"key":"ref118","article-title":"I2AM: Interpreting image-to-image latent diffusion models via attribution maps","author":"Park","year":"2024"},{"key":"ref119","doi-asserted-by":"publisher","DOI":"10.1016\/j.cviu.2025.104559"},{"key":"ref120","first-page":"27074","article-title":"TRAK: Attributing model behavior at scale","volume-title":"Proc. Int. Conf. Mach. Learn. (ICML)","author":"Park","year":"2023"},{"key":"ref121","first-page":"19920","article-title":"Estimating training data influence by tracing gradient descent","volume":"33","author":"Pruthi","year":"2020","journal-title":"Proc. Adv. Neural Inf. Process. Syst."},{"key":"ref122","article-title":"Improved techniques for training GANs","volume":"29","author":"Salimans","year":"2016","journal-title":"Proc. Adv. Neural Inf. Process. Syst."},{"key":"ref123","article-title":"A framework for the quantitative evaluation of disentangled representations","volume-title":"Proc. 6th Int. Conf. Learn. Represent. (ICLR)","author":"Eastwood","year":"2018"},{"key":"ref124","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1053"},{"key":"ref125","article-title":"Planning with diffusion for flexible behavior synthesis","author":"Janner","year":"2022"},{"key":"ref126","first-page":"2442","article-title":"Unsupervised anomaly detection via masked diffusion posterior sampling","volume-title":"Proc. 33rd Int. Joint Conf. Artif. Intell.","author":"Wu","year":"2024"},{"key":"ref127","first-page":"24271","article-title":"Attribute based interpretable evaluation metrics for generative models","volume-title":"Proc. Int. Conf. Mach. Learn. (ICML)","author":"Kim","year":"2024"},{"key":"ref128","article-title":"Image generation diversity issues and how to tame them","author":"Dombrowski","year":"2024"},{"key":"ref129","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.02337"},{"key":"ref130","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.emnlp-main.595"},{"key":"ref131","doi-asserted-by":"publisher","DOI":"10.1016\/j.cag.2024.104058"},{"key":"ref132","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2025.3569700"},{"key":"ref133","doi-asserted-by":"publisher","DOI":"10.1007\/s40860-024-00240-0"}],"container-title":["IEEE Transactions on Artificial Intelligence"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/9078688\/11589479\/11314921.pdf?arnumber=11314921","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T19:39:15Z","timestamp":1782934755000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11314921\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7]]},"references-count":133,"journal-issue":{"issue":"7"},"URL":"https:\/\/doi.org\/10.1109\/tai.2025.3648376","relation":{},"ISSN":["2691-4581"],"issn-type":[{"value":"2691-4581","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7]]}}}