An Investigation of the NeurIPS and ICML 2025 Position Tracks

A recent analysis of the NeurIPS and ICML 2025 Position Tracks indicates a significant predominance of 'reformist critique' submissions within the publicly accessible reviewed pool. The paper argues that while these tracks effectively host critical analyses, they should explicitly seek out and balance 'direction-setting work' to foster a more diverse range of foundational contributions.

Why it matters

The composition of agenda-setting tracks in influential research venues directly impacts the future trajectory of a field, guiding funding, talent, and technological development. An overemphasis on critique without corresponding direction-setting content could limit innovation and the establishment of new research paradigms, potentially slowing progress in critical areas.

Key insights

• Machine learning (ML) venues significantly influence the types of research and evidence deemed legitimate.
• The NeurIPS and ICML Position Paper Tracks are designed for agenda-setting research.
• The 2025 reviewed pool for these tracks is predominantly composed of 'reformist critique'.
• The study audited accessible submissions using a pre-specified rubric and compared them to established agenda-shifting ML papers.
• Three-quarters of audited submissions align with reformist critique.
• The paper recommends that the tracks actively solicit 'direction-setting work' in addition to reformist critiques.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2608.16894