How to enjoy your first ML conference

I went to my first ML conference the last few days and I realised there is a lot of stuff I didn’t know about how to attend a large ML conference. So here is a list of information that is useful to you, a first time ML conference attendee.

Before we start, let’s explain the three parts of a ML conference

  1. Posters: this is where most papers of the conference come. When a paper gets accepted into a ML conference, it becomes a poster. The conference over its few days will have several hundred posters in the exhibition hall, in multiple sessions every day.
  2. Orals: some of the papers reviewers decide are good enough to get airtime in a conference are become orals. Basically this means that there are oral sessions where papers in bunches (I think about eight?) are presented to the audience in quick order. Each oral gets a short time, maybe less than ten minutes.
  3. Workshops: These are after the main conference and are basically focused on one specific topic. They’re more chill, less chaotic and more focused on work that is in-progress as compared to the conference which is focused on work that is complete.

Pre-conference preparation

The first thing you don’t realise before going here is that ML conferences are huge. No matter which university you come from, you probably won’t realise the scale of this. There are over 3,700 papers at ICLR 2025, and over 370 of those got a spotlight. Even if you spend five minutes per paper, that’s 308 hours overall to read all the papers. So, you have to pick and choose which posters you want to go to and which orals you want to see.

How do you choose? The first thing you can do is go on the site of the conference or openreview and skim through titles of all the accepted papers. Or do a keyword search for them. Another thing you can do is use tools made by researchers to better locate the papers you want (ICLR had this great one) . There are also tools like Scholar Inbox (which I only learnt of during the conference) that help you do it. Paper discovery is a fairly idiosyncratic process, so I’ll leave it to you. One thing I’d recommend is if you have a group of people you trust, you could split up the work and see what all of you find interesting to maximise the area you cover.

Regardless, start early. I’d say about two weeks before the conference. I started just a few days before, and trust me I was very tired just looking at the large number of papers before.

The point while skimming or reading is to see

  1. Is the paper relevant to your subfield of interest
  2. Does it tell you something new you didn’t know or expect (does it violate your priors in some way or another)

Remember that the objective is to find the papers that are relevant and useful for you. Maybe put them into a google sheets if that’s your style

During the conference

Now let’s talk about surviving the actual conference days. ML conferences are physically demanding - you’ll be on your feet a lot, walking between sessions, and constantly engaging with new ideas and people.

Physical survival

The basics matter more than you’d think. Stay hydrated - the conference environment can be surprisingly dehydrating. Keep your setup minimal - I found a water bottle and notebook to be the perfect combination. Avoid lugging around a heavy backpack or laptop if you can help it.

Take breaks when you need them. Find places to sit down between sessions. Remember, the goal isn’t just to survive one day - you need to maintain energy through both the main conference and workshop days.

Poster sessions require strategy. Before each session, consult your prepared list and map out an efficient route. Think of it like solving a traveling salesman problem - you want to minimize walking time between posters you care about.

The poster layout has some loose thematic clustering - I noticed protein papers tended to be near each other, as did diffusion papers. However, the clustering isn’t perfect, so planning your route matters.

Keeping track

You’ll discover interesting papers during the conference that weren’t on your radar before. Have a system ready - I sent interesting finds to a dedicated chat/notes where I could review them later. This lets you capture opportunities without breaking your flow.

Energy management

Sleep is crucial. You might be tempted to skip it for networking or preparation, but don’t. The physical and mental demands of the conference compound each day.

A note about keynotes and orals: I personally found poster sessions gave me better returns on time invested - the direct interaction with authors was more valuable than watching presentations. However, this is quite personal - your mileage may vary depending on your learning style and goals.

The key is being intentional about your time and energy. You can’t do everything, so focus on what gives you the most value.

Social interaction

A big part of what makes conferences valuable isn’t just the academic content - it’s the connections you make. Many organizations host social events during the conference:

Workshop networking

In my experience, workshops are excellent for networking. They’re more relaxed than the main conference sessions and tend to attract people with shared, specific interests. The environment is more conducive to real conversations and finding potential collaborators.

Making connections stick

Meeting people is just the first step. What matters is converting those brief interactions into meaningful professional relationships:

  1. Collect contact information - whether that’s email, Twitter, or whatever platform works for you
  2. Follow up after the conference while you’re still fresh in their memory
  3. Be memorable for the right reasons - thoughtful questions and genuine engagement beat aggressive networking

Even if you’re not actively job hunting (I wasn’t), these connections can lead to valuable collaborations and opportunities down the line.

Value proposition

The real value of conferences often isn’t in the rapid-fire poster sessions or talks - it’s in the meaningful conversations you have with other researchers. Prioritize creating spaces for these interactions, whether that’s during workshop breaks or at social events.