Publication strategy

Publication Strategy

Get your work published efficiently without wasted submissions.

Venue Selection

Note: Deadlines and acceptance rates shift year-to-year. Always verify current CFP at the official venue site before planning submissions.

AI/ML Top Venues (Tier 1)

Venue	Type	Deadline pattern*	Acceptance rate*
NeurIPS	Conference	May	~25%
ICML	Conference	Jan	~27%
ICLR	Conference	Oct	~30%
CVPR	Conference	Nov	~25%
ACL/EMNLP	Conference	Feb/Jun	~25%
AAAI	Conference	Aug	~23%

*Approximate — check official CFP each year.

Systems / Distributed Computing

Venue	Type	Focus
MLSys	Conference	ML systems, FL efficiency
EuroSys	Conference	Systems
IEEE TPDS	Journal	Parallel/distributed
IEEE IoT Journal	Journal	IoT applications

Choosing Your Venue

Answer these:

1. Contribution type: algorithm → ML venue; system → systems venue; application → domain venue
2. Novelty level: incremental → workshop or journal; significant → top conference
3. Timeline: PhD submission date → work backwards from deadlines
4. Audience: who needs to read this? → they attend which venue?

Strategy for PhD students:

First paper → workshop at top venue (NeurIPS/ICML workshop) — lower bar, builds CV
Second paper → Tier 2 conference (ECML, ICDM, PAKDD)
Third paper → Tier 1 attempt (NeurIPS, ICML, ICLR)
Thesis compilation → journal version (IEEE TPDS, JMLR)

Submission Checklist

• Paper within page limit (main + appendix)
• All author information removed (double-blind venues)
• Code anonymized (no usernames in comments, no personal paths)
• All figures are vector (PDF) or ≥300 DPI
• References formatted per venue style
• Supplementary material within size limit
• Ethics statement if required (NeurIPS mandates it)
• License statement if releasing dataset/code

Rebuttal Writing

Most top conferences have 5-14 day rebuttal windows (NeurIPS ~6 days, ICML ~7 days, ICLR ~2 weeks). Structure:

We thank all reviewers for their detailed feedback. We address the main concerns:

**[R1, R2] Missing baseline [Method X]:**
We ran [X] under identical conditions. Our method achieves Y% vs X's Z%
(see attached Table). We will include this in the final version.

**[R3] Concern about dataset diversity:**
We acknowledge this limitation. We plan experiments on [Dataset B] 
which we can include in revision. Preliminary results show [outcome].

**[R1] Unclear notation in Section 3:**
We will revise Section 3.1 to clarify [specific point]. The key insight is [explanation].

Minor issues raised by all reviewers will be addressed in revision.

Rebuttal rules:

• Don't be defensive — acknowledge valid concerns
• Promise only what you can deliver in revision timeline
• Focus on fatal flaws, not nitpicks
• Keep it concise — reviewers are busy

Revision Plan

After receiving "Major Revision" decision:

## Revision Plan for [Venue] [Year]

**Deadline:** [date]
**Review decision:** Major / Minor revision

| Reviewer | Concern | Action | Owner | Due |
|----------|---------|--------|-------|-----|
| R1 | Missing baseline X | Run experiments | Me | +2 weeks |
| R2 | Unclear Section 3 | Rewrite 3.1 | Me | +1 week |
| R3 | Needs ablation on Y | Design + run | Me | +3 weeks |

**New experiments needed:** [list]
**Sections requiring rewrite:** [list]
**Estimated revision time:** [X weeks]

After Rejection

Rejection ≠ bad paper. NeurIPS accepts ~25%, reject = 75% of all submissions.

1. Read reviews carefully — identify legitimate weaknesses
2. Rate each review: valid / partially valid / wrong
3. Decide: fix + resubmit to same tier OR resubmit to next venue
4. Don't submit same paper to next deadline without addressing reviewers

Typical path: NeurIPS reject → fix → ICML (3 months later)

Links to Other Skills

• Requires → internal-critique (paper must pass internal check first)
• Requires → paper-writing (clean draft)
• Feeds into → milestone-tracker (add submission deadlines)