In this guide
A Recruiter’s Guide to AI Research Conferences: NeurIPS, ICML, CVPR and More
· ForesIQ
In computer science, the leading peer-reviewed conferences are where the most important research is published. NeurIPS, ICML and ICLR lead machine learning; CVPR, ICCV and ECCV lead computer vision; ACL, EMNLP and NAACL lead natural language processing; ICRA, IROS, RSS and CoRL lead robotics. Knowing the venues tells a recruiter where a field’s work appears and how to read a publication list.
In most sciences, journals carry the prestige and conferences are for presenting work in progress. Computer science is different. Its top conferences are selective, fully peer reviewed and treated as the primary place to publish. A paper at a leading conference is a finished contribution that has passed competitive review.
That makes conference names one of the most useful pieces of vocabulary a technical recruiter can learn. This guide covers the venues that matter in each field and how to read them.
The leading venues by field
| Field | Venues | What appears there |
|---|---|---|
| Machine learning | NeurIPS, ICML, ICLR | Learning theory, optimisation, deep learning, generative models, reinforcement learning. |
| Computer vision | CVPR, ICCV, ECCV | Recognition, 3D reconstruction, video understanding, image generation. |
| Natural language processing | ACL, EMNLP, NAACL | Language models, translation, information extraction, evaluation. |
| Robotics | ICRA, IROS, RSS, CoRL | Manipulation, locomotion, planning, robot learning. |
| Systems for machine learning | OSDI, SOSP, MLSys | Training and serving infrastructure, compilers, distributed systems. |
A few notes help with reading the table. ICCV and ECCV alternate, with ICCV held in odd years and ECCV in even years. CoRL focuses on the meeting point of robotics and machine learning. Work often crosses fields: a vision-language paper might appear at CVPR, ACL or NeurIPS depending on its emphasis.
How to read a venue on a publication list
The venue places the work. A run of papers at CVPR and ICCV tells you someone’s centre of gravity is computer vision, whatever their title says. A mix of NeurIPS and ICRA suggests someone working where learning meets robotics.
Author position shows the role. In most of machine learning, the first author carried the work and the last author is usually the senior researcher who guided it. Middle positions cover a wide range of contributions. Some theory communities list authors alphabetically, so check the convention before reading too much into order.
Dates show direction. Recent work tells you what someone is doing now. A publication list that moves from one topic to another over a few years shows where their interests are heading.
Presentation type is a bonus signal. Many conferences select a small share of accepted papers for oral or spotlight presentation. It is worth noticing and should never be the only thing you look at.
Main track, workshop and preprint
- Main track papers have passed the conference’s full peer review. This is what people mean when they say someone “has a NeurIPS paper”.
- Workshop papers are presented at smaller sessions held alongside the conference. Review is lighter and the work is often earlier stage. Workshops are valuable for spotting emerging topics.
- Preprints are papers shared publicly, most often on arXiv, before or without peer review. Much important work appears as a preprint first. Treat it as a claim awaiting review.
When you see a conference name on a CV or profile, check which of these it refers to. The difference is real, and researchers will notice if you understand it.
Using conferences in a search
Map the field through its venues. The authors publishing on a topic at the leading venues over the last few years are, in effect, the talent market for that topic.
Follow the workshops. A workshop on a new topic is often the first organised gathering of the people who will define it.
Use the programme. Proceedings and accepted-paper lists are public. They are a legitimate, rich starting point for understanding who is working on what.
Refer to the work in outreach. A message that mentions a specific paper and why it matters to your team reads very differently from one built on a job title.
Frequently asked questions
What are the top AI research conferences?
For machine learning, NeurIPS, ICML and ICLR. For computer vision, CVPR, ICCV and ECCV. For natural language processing, ACL, EMNLP and NAACL. For robotics, ICRA, IROS, RSS and CoRL.
Why do conferences matter more than journals in computer science?
The field moves quickly, and its leading conferences adopted rigorous, competitive peer review. Over time they became the primary place to publish finished work, a role journals play in most other sciences.
Is an arXiv paper peer reviewed?
No. arXiv is a preprint server. Papers there may later be accepted at a conference, and many influential works appeared there first, but posting alone involves no peer review.
What does first author mean on a machine-learning paper?
By convention the first author did most of the work, and the last author is often the senior researcher who supervised it. Some subfields list authors alphabetically instead.
Does a workshop paper count as a conference publication?
It is a real contribution, reviewed more lightly than the main track and usually earlier stage. Treat it as a signal of current interests.
Explore your research brief in ForesIQ
ForesIQ is a Research Talent Intelligence platform for frontier technology teams. Its specialist database brings expertise, field-relative standing, Momentum, Activity and supported career evidence into one workspace. Current areas include computer vision, machine learning, natural language processing, systems and infrastructure, and robotics.
Open a research area and see the people publishing in it, with the venues and topics behind each name. Request A Demo and we’ll send you a personalised demo showing how ForesIQ can support specialist research talent discovery.