TMLR as a Publication Venue | Generated by AI
Question: What is TMLR (Transactions on Machine Learning Research), and how does it work as a publication venue for machine learning papers?
Answer:
Overview of TMLR
Transactions on Machine Learning Research (TMLR) is a new venue for disseminating machine learning research, intended to complement JMLR (Journal of Machine Learning Research) while supporting the unmet needs of a growing ML community.
TMLR began publication in June 2022, and has since grown into one of the most respected venues in the field.
Why TMLR Was Created
The founders identified several frustrations with the existing conference ecosystem: disenchantment with the exclusivity of top conferences, concerns that conferences were not highlighting the most impactful work, slow turnaround from submission to decision, high-stakes pressure around fixed deadlines, and a perceived decline in peer review quality. In that spirit, they founded TMLR as a new contribution to the ML publication ecosystem.
Key Characteristics
1. Journal, Not a Conference
TMLR emphasizes technical correctness over subjective significance, ensuring it facilitates scientific discourse on topics that may not yet be accepted in mainstream venues but may be important in the future.
2. Conference-Style Format
TMLR caters to shorter format manuscripts that are usually submitted to conferences, providing fast turnarounds and double blind reviewing. It employs a rolling submission process, shortened review period, flexible timelines, and variable manuscript length, to enable deep and sustained interactions among authors, reviewers, editors, and readers.
3. No Overlap Policy
TMLR does not accept submissions that have any overlap with previously published work.
4. Anytime Submission
Being a journal, TMLR accepts submissions throughout the year via a rolling review process executed on a per-paper timeline.
5. Fast Turnaround
TMLR targets providing reviews within 4 weeks of submission and decisions within 2 months.
6. Acceptance Criteria
Acceptance avoids judgments based on subjective, editorial, or speculative elements of typical conference decisions (such as novelty and potential for impact). Instead, acceptance is based on the validity and correctness of the claims.
7. Open & Transparent Review
TMLR’s review process is hosted by OpenReview, making it open and transparent to the community.
Certifications (Paper Awards)
TMLR uses a certification system to highlight exceptional papers:
- Reproducibility Certification – for papers focused on reproducibility studies.
- Featured Certification – awarded by Action Editors to stand-out papers.
- Outstanding Certification – the highest honor, awarded annually.
The selection process for the Outstanding Certification considers all papers published in TMLR up to a cutoff date. Papers are filtered based on whether they received a Featured Certification or a large number of citations, and then reviewed by a dedicated committee.
For example, the 2024 Outstanding Certifications were awarded to “Inversion by Direct Iteration: An Alternative to Denoising Diffusion for Image Restoration” and “Holistic Evaluation of Language Models” (HELM), with notable Featured Certifications going to BIG-bench, DINOv2, and an RLHF survey paper.
Conference Integration (Journal-to-Conference Track)
Selected papers published in TMLR are now eligible for presentation at the joint NeurIPS, ICML, and ICLR Journal-to-Conference (J2C) Track, giving authors an opportunity to share their research in person at top conferences.
To be eligible, a TMLR paper must have received a J2C Certification, Featured Certification, or Outstanding Certification, and must not contain contributions already published at a conference by the same authors.
Editorial Leadership
The current Editors-in-Chief of TMLR are Laurent Charlin (HEC Montréal), Gautam Kamath (University of Waterloo), Naila Murray (Meta), and Nihar B. Shah (CMU). Founding Editors-in-Chief include Kyunghyun Cho, Raia Hadsell, and Hugo Larochelle.
Summary Table
| Feature | Details |
|---|---|
| Type | Journal (not a conference) |
| Started | June 2022 |
| Submission | Anytime (rolling) |
| Review | Double-blind, via OpenReview |
| Turnaround | ~2 months target |
| Focus | Technical correctness, not novelty |
| Certifications | Reproducibility, Featured, Outstanding |
| Conference link | NeurIPS, ICML, ICLR J2C Track |
References:
- TMLR Official Site – jmlr.org
- Announcing TMLR – Hugo Larochelle on Medium
- TMLR joins NeurIPS/ICML/ICLR J2C Track
- 2024 TMLR Outstanding Certification
- 2025 TMLR Outstanding Certification
- NeurIPS Journal-to-Conference Track