OpenCódice

Venues

What gets a paper rejected at NeurIPS, ICLR, or COLM? Instead of recycling folklore, we cluster the weaknesses reviewers actually flag in their public reviews and present them in a browsable way. Open science on open data.

ICLR 2024

2024

Flagship venue for representation learning. Single round, fully transparent review cycle.

7,404
submissions
39.1%
Acceptance rate
10
clusters
View analysis

ICLR 2025

2025

ICLR 2025 edition. Same format as 2024, an even larger dataset: the API exposes 11,672 submissions and 5,019 rejections. Lets us see how reviewer concerns evolve year over year.

11,672
submissions
42.5%
Acceptance rate
10
clusters
View analysis

NeurIPS 2024

2024

The largest ML venue. OpenReview only exposes submissions that reached the decision stage; what you see here is the tail of the process.

4,236
submissions
95.3%
Visible acceptance
10
clusters
View analysis

TMLR

2024

Transactions on Machine Learning Research. Journal-style continuous review with no deadline. Different incentives for authors and reviewers: does it translate into different rejection patterns?

6,661
submissions
71.8%
Acceptance rate
10
clusters
View analysis

Year-over-year: ICLR 2024 vs 2025

Same venue, two consecutive editions. Did what reviewers complain about change between 2024 and 2025?

See evolution

Raw evidence

Every cluster includes anonymous exemplar quotes and its most representative terms. We don't ask you to trust us; we show the evidence.

Practical takeaways

Every pattern ships with an actionable note. What you'd change in your paper if you knew the pattern was there.

Reproducible

Open pipeline: TF-IDF + LSA + KMeans over OpenReview public data. Documented in OC-TR-2026-007.

How is this built?

Methodology, data sources, and limitations. Designed for reproducibility and auditing.

Methodology