ICML Experimental Program Using Google's Paper Assistant Tool (PAT) (blog.icml.cc)

The International Conference on Machine Learning (ICML) 2026 has announced an innovative initiative in collaboration with Google, introducing the Paper Assistant Tool (PAT). This Gemini-based AI tool is designed to provide authors with automated and actionable feedback on their manuscripts ahead of submission, fostering improved quality before the official review process. In a precedent set during the ACM Symposium on Theory of Computing, 94% of participants found the AI-generated pre-submission feedback beneficial, underscoring the potential impact of PAT on research clarity and rigor in the ML community. PAT operates in a secure, privacy-focused manner, strictly separate from the review process, ensuring that feedback is confidential and not accessible to reviewers or program committee members. The tool is particularly tailored for machine learning manuscripts, assisting authors in addressing common concerns such as experimental methodology and narrative clarity. By allowing only one submission per eligible author within a defined window, the program emphasizes quality over quantity, enabling authors to refine their work with insights drawn from advanced models that have excelled in reasoning tasks. This forward-thinking application of AI aims not only to bolster submission standards but also to pioneer a responsible approach to leveraging AI in academic writing.