Start with contribution shape
Conference papers fit one clear, reproducible result inside a hard page cap. Journals want depth: theory, ablations, multiple datasets, and a longer related-work conversation.
A thin story padded into a journal often fails desk review; a thick study crushed into eight pages reads incomplete.
Who is this for?
At NeurIPS, ICML, and ICLR, reviewers usually care whether the method holds up and the experiments reproduce. At CVPR and ICCV, they care more that the visual approach is clear and credible.
Chinese journals are another room: readers and tastes sit closer to domestic research contexts, so abstract style and citation habits rarely copy-paste from English flagships.
Picking a venue is picking who hears the first sentence of your contribution—and that sentence often has to change.
- Learning-systems / algorithm-first work → TNNLS, NeurIPS, ICML
- Vision methods with heavy experiments → TPAMI, CVPR, IJCV
- Chinese software and systems → Journal of Software, CJC
Deadlines and dual submission
Conferences have hard deadlines; journals roll but review slowly. Missing a deadline is not a reason to dual-submit.
Substantially similar parallel archival submissions are usually forbidden. Workshops without proceedings may be exempt—always read the live CFP.
How to use impact factor
Institutions care about lists and quartiles, but fit still comes first. IF changes yearly; conferences have none.
Stabler signals: CCF tier, CAS/JCR quartile, and whether recent issues still publish your topic.
FAQ
- Conference or journal for a first paper?
- Conferences are often more realistic: faster feedback and a clear page target. Go journal-first only if the work is already archival-depth.
- Sort only by impact factor?
- No. High IF with the wrong readers is a slow path. Fit first, then institutional lists.