This venue
Pretraining, fine-tuning, inference, and LM evaluation; for general NLP tasks beyond LMs, ACL/EMNLP are also strong.
arXiv is allowed, but the review PDF must stay anonymous. Reviewers need not read the appendix—keep key results in the main text.
Length
Main text strictly ≤ 9 pages; unlimited citations. Unlimited appendix (unread is allowed). Optional ethics, reproducibility, and acknowledgments: ≤ 1 page each, outside the cap.
Review
Double-blind. No acknowledgments or identity-revealing links. Third-person self-citation.
Typesetting
Official one-column Letter template, typically 10pt. Do not resize the text block. Number pages. Submission mode often has line numbers; switch to final after acceptance.
Before you write
- Main text ≤ 9 pages.
- Links anonymized.
- Ethics/repro statements outside the 9 pages.
- Key tables not appendix-only.
Official pages
Summarized from public calls (as of 2025-09). Always match the CFP you are submitting to.
FAQ
- COLM vs ICLR?
- COLM for LM training/eval; ICLR for broader deep learning representation work.
- Will reviewers read the appendix?
- Not guaranteed—keep main results and key ablations in the body.
- Where do ethics statements go?
- Optional, ≤ 1 page each, outside the body cap—follow that year’s CFP.
