The AI Research Job Market is Broken: Overemphasis on CVPR, NeurIPS, ICML, ECCV, ICCV, & ICLR is Hurting PhD Graduates

I have visited multiple universities as an external PhD examiner across Europe and Asia and have interacted with many PhD students. A common theme has emerged: apprehension, stress, and even thoughts of leaving AI research altogether—both in academia and industry.

Many of these PhD students have made novel contributions and published in reputable journals and conferences, but not in the six "elite" venues: CVPR, ECCV, ICCV, NeurIPS, ICML, and ICLR. Due to the hype around these conferences on social media, in hiring processes, and in top universities, they feel undervalued and overlooked.

Even more concerning is that some students who have published 5+ papers in these "elite" venues still feel inadequate because they see others with 20 or even 30 papers in these same venues struggling to secure top tech jobs. The result? A toxic cycle of endless publishing, stress, and uncertainty about career prospects.

Having worked in AI/ML/CV for over 20 years in academia, my advice to PhD students and early-career researchers is simple:

👉 Focus on doing good work in whatever project you are working on—whether as a PhD, postdoc, or intern.

👉 Seek diverse opportunities and diverse projects.

👉 Open your horizon beyond just these six venues and look for impactful work in industry, startups, and even applied research roles.

Challenges Faced by AI PhDs in the Current Job Market

1. The Unrealistic Emphasis on These Six Conferences

Top AI research labs and tech companies often require or strongly favor candidates with multiple first-author papers in these conferences. Many hiring managers and universities do not value reputable journal publications (like TPAMI, IJCV, JMLR) or contributions to applied AI research. If a candidate has no CVPR/NeurIPS papers, they might not even be considered, despite having impactful work elsewhere.

2. Extreme Competition for Paper Acceptance

These conferences have low acceptance rates (~25%) with oral presentations below 5%. Papers often get rejected multiple times for arbitrary reasons like lack of another ablation study or missing a small improvement margin (e.g., 0.5% gain over SOTA). Even strong research gets desk-rejected due to reviewer fatigue and inconsistencies. Even if all the reviewers accept a paper, the area chair can override their opinions and unilaterally reject it based on their own discretion, citing reasons they personally deem valid. Unlike some journals where the editor's decisions are transparent, conference area chairs remain anonymous, making it difficult to assess whether their decisions are unbiased or influenced by personal biases.

3. The "Paper Chase" Delays Graduation

Many PhD students delay their graduation by 1 to 3 years just to get papers accepted at these venues. To make matters worse, I recently heard from a visiting student at a Top 5 university, who was part of a research group led by one of the world's top 1000 researchers. The student mentioned that if they hadn’t published at least three papers in these six elite conferences, they were not even allowed to request permission to start writing their PhD thesis. This unofficial policy applied even if it meant delaying graduation by five years or, in some cases, never completing the PhD at all. I have even seen multiple instances in Asia during my own journey before I became faculty at UiT- The Arctic University of Norway that supervisor declined to support PhD degree even if student published other conference papers at WACV, MICCAI, KDD, ECML-PKDD, etc. I have seen this happen to multiple students. There is also an echo chamber among high-profile AI professors who publish exclusively in these six venues. If a paper is not accepted at these conferences, they often dismiss the work entirely, regardless of the effort students and staff have put into the research or the novelty of the contributions. Some students fall into a vicious cycle of endless resubmissions and revisions. Those with multiple rejected submissions feel demotivated and anxious about their career prospects.

4. Industry's Hiring Bias Toward "Elite" Publications

Applied AI and real-world impact are often undervalued—companies want benchmark-driven research. Journals are ignored in hiring, even though they allow for more detailed research. AI/ML engineering experience, open-source contributions, and real-world AI applications are overlooked in favor of conference papers.

5. The Mental Health Toll

Constant pressure to publish at these venues causes burnout, depression, and anxiety. Imposter syndrome is rampant, even among those with multiple publications. Many students feel "not good enough" and question whether to continue in research.

How Can AI PhDs Navigate This Toxic Job Market?

The students need to diversify their publication venues. Beyond the "Big Six", consider publishing in high-impact journals (TPAMI, IJCV, JMLR) and other strong conferences (AAAI, WACV, MICCAI, KDD, ECML-PKDD, etc.). ArXiv + GitHub + Open-Source Work can be just as valuable as a top-tier paper if widely adopted.

The students can focus on gaining real-world AI/ML engineering skills. Work on scalable AI systems, multimodal learning, edge AI, and real-world ML pipelines. Collaborate with startups, industry, and cross-disciplinary projects.

The students may look beyond "Top Tech Research Labs". There are incredible AI opportunities in healthcare, climate tech, robotics, retail, finance, and beyond. Startups and applied research roles can offer fulfilling careers outside of academia.

The students can work on building a strong personal brand. Write technical blogs, share insights on LinkedIn and Twitter/X, and engage in AI discussions. Contribute to open-source projects, Kaggle competitions, and Hugging Face models.

The students should prioritize their mental health. Seek mentorship from advisors who support a healthy work-life balance. Reject toxic "paper count" culture—impact matters more than quantity.

• Remember: A PhD is about learning and problem-solving, not just chasing conference acceptances.

Final Thoughts

The AI research job market has become broken due to its overemphasis on six conferences. This has created an ecosystem where:

✅ Talented PhD graduates feel undervalued, even if they have done great work.

✅ Hiring committees overlook researchers who don't have a "NeurIPS/CVPR" stamp.

✅ Mental health issues and burnout are rising among young AI researchers.

The reality is that there is a world beyond NeurIPS and CVPR. Many exciting, impactful AI jobs exist outside the Big Tech research bubble. By diversifying skills, broadening opportunities, and focusing on real-world impact, PhD graduates can build fulfilling, successful AI careers—without being trapped in the paper-count rat race.

If you're a PhD student or researcher struggling with this pressure, you are not alone. Let’s start a conversation—what are your experiences with the current AI research job market?

#AI #PhD #MachineLearning #ComputerVision #NeurIPS #CVPR #ICML #ICLR #MentalHealth #Research #TechCareers