ICML 2026 is the 43rd International Conference on Machine Learning, taking place July 6–11 at the COEX Convention and Exhibition Center in Seoul, South Korea. The premier annual gathering for machine learning research draws researchers, engineers, and practitioners worldwide across a six-day programme of tutorials, main-conference sessions, and workshops. This marks the conference’s first appearance in South Korea.

Machine Learning / AI Research | Jul 6 – Jul 11, 2026 | Hybrid (In-person + Virtual), Seoul, South Korea | Hosted by the International Machine Learning Society (IMLS)

About This Event

ICML is one of the three most influential conferences in machine learning and AI research, alongside NeurIPS and ICLR. Organised by the International Machine Learning Society (IMLS), it has run annually since 1980, beginning as the International Workshop on Machine Learning at Carnegie Mellon University in Pittsburgh before moving to its current conference format in 1993.

The conference is globally recognised for presenting and publishing cutting-edge work across all areas of machine learning and adjacent fields, including statistics, data science, machine vision, computational biology, speech recognition, and robotics.

ICML traditionally carries more content on statistical learning theory, reinforcement learning, robotics, and optimization theory than its peer conferences. Proceedings are published open-access in the Proceedings of Machine Learning Research (PMLR) at proceedings.mlr.press.

Participants span academic and industrial researchers, entrepreneurs, engineers, graduate students, and postdocs. Technology companies including Google, Microsoft, Amazon, Meta, and Apple are regular sponsors who publish research and recruit at the conference, and financial firms including Citadel Securities, Jane Street Capital, and D.E. Shaw also maintain a presence.

What Usually Happens at ICML?

The conference follows a structured six-day format. The first day combines an Expo with tutorial sessions. The main conference runs across three days with oral presentations, poster sessions, invited talks, and a Test of Time Award lecture. The final two days are given over to workshops, which run as co-located events with their own programmes and independent registration caps.

Socials, mentoring sessions, and career events run alongside the technical programme throughout the week. Each year ICML selects a small number of outstanding papers from thousands of submissions, alongside a Test of Time Award recognising a paper from roughly a decade earlier whose impact has proved lasting.

When Is ICML 2026 Taking Place?

ICML 2026 runs July 6–11, 2026, at the COEX Convention and Exhibition Center, 513 Yeongdong-daero, Gangnam-gu, Seoul, South Korea. Tutorial and main conference in-person registrations are sold out. Workshop registrations remain open, and virtual registration, which is not subject to the in-person cap, is available at icml.cc/Register2.

ICML 2026: Latest Updates

The programme is confirmed and the conference is now weeks away. All updates below are sourced directly from icml.cc and the official ICML blog at blog.icml.cc.

Tutorial and Main Conference Registrations Sold Out

In-person registrations for the tutorials and main conference are sold out. Workshop registrations remain open, and authors of accepted posters can still register through the process on the Poster Instructions page. The organising team set a registration cap marginally below the venue’s legal limit, a decision confirmed by ICML Board President Kamalika Chaudhuri, who framed the cap as a way of ensuring attendees can hear the talks, have the conversations, and engage with the work once in Seoul.

Six Invited Speakers Announced

ICML 2026 has confirmed six invited speakers whose work spans machine learning theory, AI safety and ethics, economics and policy, computational biology, natural language processing, and human-computer interaction. The lineup was announced on the official ICML blog on May 18, 2026.

LLM Use in Reviewing Policy Published

The ICML 2026 policy on LLM use in reviewing has been published at icml.cc/Conferences/2026/LLM-Policy. It follows an experimental programme using Google’s Paper Assistant Tool (PAT) for ICML 2026, with a retrospective published on the ICML blog. See also the Peer Review FAQ.

Self-Ranking Policy Introduced

A new self-ranking policy for reviews was introduced for 2026. Authors with multiple submissions are asked, on a voluntary basis, to provide a partial ranking of their own submissions. Disagreements between author rankings and reviewer scores help flag papers needing more attention and guide the assignment of emergency reviewers.

WiFi Limitations Confirmed at COEX

The organising team has confirmed real limitations to wireless connectivity at COEX, particularly in the plenary hall during crowded sessions. Keynotes will be streamed into multiple rooms to distribute attendance, and the ICML mobile app stores the schedule locally without needing a connection. Attendees are encouraged to download the app and save their schedule before travelling.

ICML 2027 and 2028 Locations Confirmed

The Future Meetings page has been updated with confirmed locations for ICML 2027 and ICML 2028, continuing the conference’s long-term planning cycle.

Main Topics at ICML 2026

ICML 2026 covers the full breadth of machine learning research, from theoretical foundations to applied systems. The topics below reflect the standing research areas covered at every ICML and the specific invited-talk expertise announced for 2026.

Statistical Learning Theory and Optimization

ICML’s traditional strength in statistical learning theory, optimization theory, and mathematical foundations runs through the main conference, with sessions on generalization, sample complexity, convergence analysis, and the theoretical underpinnings of modern deep learning. Sham Kakade’s invited talk draws from this tradition, spanning scalable training methods for large language models, tensor methods for latent-variable estimation, and foundational deep learning theory.

Large Language Models and Foundation Models

LLMs and foundation models continue to dominate ICML submissions, covering training methods, evaluation, alignment, safety, and efficiency. A new Lay Summaries programme for 2026 aims to make accepted papers accessible to a broader audience. Verena Rieser’s invited talk on the responsible development and alignment of frontier AI models speaks directly to this track.

Reinforcement Learning and Decision Making

Reinforcement learning and sequential decision making remain a core ICML track, covering theoretical foundations, sample efficiency, offline RL, multi-agent systems, and RL applied to language models and robotic control. Sham Kakade’s foundational work on the statistical foundations of RL and provably efficient policy search provides mathematical backbone for much current work in the area.

AI Ethics, Safety and Societal Impact

The ethics, safety, and societal impact track reflects ICML’s expanded commitment to responsible research. Arvind Narayanan’s invited talk on the societal impact of digital technologies, drawing from his book AI Snake Oil, addresses AI agent evaluation, AI’s impact on institutions and professions, and algorithmic amplification. Pascale Fung’s work on ethical AI and affective computing, including her roles with the UN Advisory Body on AI Governance and the World Economic Forum, anchors the policy dimension.

Computational Biology and AI for Science

AI for science and computational biology is a growing ICML track, addressing machine learning applied to drug discovery, genomics, protein structure prediction, and clinical decision making. Aviv Regev’s invited talk draws from her leadership at Genentech, where she oversees the integration of AI and ML into drug discovery in a Lab in the Loop framework.

Economics, Policy and Social Systems

The intersection of machine learning with economics, policy, and social systems is represented by Susan Athey’s invited talk, drawing on her research on the economics of digitization, causal inference and AI, online advertising, news media, labour-market transitions, and digital technology for social impact.

Invited Speakers

Six researchers have been confirmed as invited speakers for ICML 2026. The lineup was announced on the official ICML blog on May 18, 2026. Talk topics and detailed scheduling will be announced closer to the conference.

Full speaker profiles are on the official ICML blog announcement of May 18, 2026.

Programme Format

ICML 2026 follows a six-day structure confirmed on the official ICML website and dates-and-deadlines page. The Expo and Tutorial day opens the conference, followed by three main-conference days, and two workshop days with their own independent registration cap that are not sold out.

Expo and Tutorial Day: July 6

July 6 combines an Expo with tutorial sessions. Tutorials were announced on April 2, 2026, via the official ICML blog. The Expo gives exhibitors and sponsors direct access to the ICML community, and Expo Presentations have their own call listed at icml.cc/Expo/applications.

Main Conference: July 7, 8 and 9

The main conference runs July 7–9 at COEX, covering oral presentations, poster sessions, all six invited talks, the Test of Time Award lecture, socials, mentoring sessions, and industry events. The Scholar Inbox conference planner provides a personal agenda and paper recommendation system, and all accepted papers are listed at icml.cc/virtual/2026/papers.html.

Workshops: July 10 and 11

Workshops run July 10–11 as co-located events with their own programmes, accepted papers, and participants. The workshop list, announced April 6, 2026, includes both standard and affinity workshops, with the full listing at icml.cc/virtual/2026/events/workshop. Workshop registrations remain open, and authors of accepted workshop papers may register through the standard process.

Virtual Access

Virtual registration provides access to all sessions and is not subject to the in-person capacity cap. It is required for at least one author of accepted papers not attending in person. The ICML mobile app is available on iOS and Android and stores the schedule locally without a live internet connection.

Why ICML 2026 Matters

ICML is where the next generation of machine learning paradigms first appears. Many of the techniques now powering large language models, reinforcement learning agents, and generative systems were first presented here. The 2026 edition marks the conference’s first appearance in South Korea and reflects the growing strength of the Asian ML research community.

For AI and ML Researchers

ICML acceptance is among the strongest forms of peer-reviewed validation in the ML community, with acceptance rates typically between 21% and 30% of submissions that now regularly exceed 10,000. The invited-talk programme gives researchers direct access to six leaders whose work spans ML theory, applied biology, economics, AI safety, and policy.

For AI Engineering and Product Teams

The papers accepted to ICML 2026 represent the research likely to shape production ML systems over the next one to three years. Teams tracking optimization, large language models, reinforcement learning, and efficient inference will find direct technical reference in the proceedings, published open-access in PMLR at proceedings.mlr.press.

For Founders and AI Leaders

ICML proceedings are one of the most important technical signals on where machine learning is heading. Arvind Narayanan’s talk on AI Snake Oil and Susan Athey’s work on the economics of AI give business leaders a direct view of the policy and economic dynamics shaping AI deployment at scale.

For Sponsors and Technology Companies

ICML is a primary recruiting venue for ML talent, with Google, Microsoft, Amazon, Meta, Apple, and financial firms including Citadel Securities and Jane Street Capital maintaining a regular presence through sponsorship, papers, and booths. The ICML Career Site is open during the conference week, connecting candidates with sponsor organisations.

Venue

COEX Convention and Exhibition Center — 513 Yeongdong-daero, Gangnam-gu, Seoul, South Korea

COEX is one of South Korea’s largest convention and exhibition complexes, located in the Gangnam district of Seoul. The venue is directly connected to Samseong subway station on Line 2 and sits within a large complex of shopping, dining, and hotels. It is reachable from Incheon International Airport (ICN) by airport express in approximately 60 minutes.

Agenda

Times shown in Korea Standard Time (KST, UTC+9). Full schedule at icml.cc/virtual/2026/calendar. The Scholar Inbox conference planner provides a personal agenda and paper recommendation system.

Full agenda at icml.cc/virtual/2026/calendar.

Registration and Access

Tutorial and main conference in-person registrations are sold out. Workshop registrations remain open, and virtual registration is available and not subject to the in-person cap. Full refunds are available until June 15, 2026; no refunds are issued after that date. Register at icml.cc/Register2. Pricing details are at icml.cc/Conferences/2026/Pricing.

Recap: 2025

ICML has run annually since 1980 and reaches its 43rd edition in 2026. The most recent edition sets the context for the scale and scope of the 2026 programme.

ICML 2025 Recap

ICML 2025, the 42nd International Conference on Machine Learning, took place July 13–19, 2025, at the Vancouver Convention Centre in Vancouver, Canada. Proceedings are available open-access at proceedings.mlr.press.

Main highlights included:

• Approximately 12,000 papers were submitted and 3,339 were accepted, reflecting continued rapid growth in ML research output.
• The Test of Time Award went to “Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift” by Sergey Ioffe and Christian Szegedy (ICML 2015), recognised as a forerunner of Layer Normalization.
• Six Outstanding Paper Awards were given in the main track and two in the position paper track, including work on token ordering in masked diffusions and the value of prediction in identifying the worst-off.
• Google presented over 140 papers as a Diamond Sponsor, with Google Research and Google DeepMind contributing across invited talks, 24 workshops, 7 orals, and in-booth demo sessions.
• Key research themes covered large language models, diffusion models, and robotics, with sessions addressing alignment, evaluation, efficient inference, and the societal implications of ML systems.

FAQs About ICML 2026

The answers below are drawn from the official ICML website and blog.

What is ICML and who is it for?

ICML is the International Conference on Machine Learning, one of the three most influential ML and AI research conferences alongside NeurIPS and ICLR. It is aimed at researchers, engineers, graduate students, and practitioners working across machine learning theory and applications.

When and where does ICML 2026 take place?

ICML 2026 runs July 6–11, 2026, at the COEX Convention and Exhibition Center in Seoul, South Korea.

Is ICML 2026 sold out?

In-person tutorial and main conference registrations are sold out. Workshop registrations remain open, and virtual registration is available and not subject to the in-person cap.

Can I still attend ICML 2026?

Yes. You can register for workshops in person, register virtually for full session access, or, if you are an accepted-paper author, register through the process on the Poster Instructions page.

Who are the invited speakers?

The six confirmed invited speakers are Pascale Fung, Susan Athey, Sham M. Kakade, Aviv Regev, Verena Rieser, and Arvind Narayanan, announced on the official ICML blog on May 18, 2026.

Where are ICML 2026 papers published?

Accepted papers are published open-access in the Proceedings of Machine Learning Research (PMLR) at proceedings.mlr.press, and are listed at icml.cc/virtual/2026/papers.html.

Where can I find the full schedule?

The full schedule is at icml.cc/virtual/2026/calendar, and the Scholar Inbox conference planner offers a personal agenda and paper recommendations