NeurIPS and ICML 2025 position tracks favor critique over new direction, study finds

A new paper investigates the position tracks of the NeurIPS and ICML 2025 conferences, arguing that they are dominated by reformist critiques rather than direction-setting work. The audit of accessible submissions found that while many papers critique existing benchmarks or methodologies, their evidentiary depth did not strongly correlate with reviewer ratings. The paper suggests that agenda-shifting research, exemplified by foundational works like AlexNet and the Transformer, typically provides new artifacts for the community to build upon, test, or contest. To address this, the authors propose four interventions for the conference calls for papers to encourage a broader mix of submissions. AI

IMPACT This analysis of conference tracks could influence how AI research agendas are set and evaluated in the future.

RANK_REASON Academic paper analyzing conference submission trends. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• AlexNet
• arXiv
• Concrete Problems in AI Safety
• NeurIPS
• Transformer

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →