MLC Affiliation Guidelines[edit]

When to Use MLC Affiliation[edit]

Use ML Collective affiliation when:

• Project started within an MLC focus group
• You were trained through MLC programs (CS197, reading groups)
• You got an idea from MLC reading group discussions
• You used MLC resources (compute grants, community feedback, mentors found through MLC)

Even if you did the work independently with zero collaboration within MLC, if you were trained through our programs, got ideas from reading groups, or used MLC compute grants, we encourage MLC affiliation. We also encourage presenting your work at Sunday calls, even if you don't need feedback—it helps the community learn.

Format:

[Your Name]^1,2
^1 [Your University, (if you have one)]
^2 ML Collective

When MLC Affiliation Is Optional[edit]

You decide when the project was primarily your independent work with minimal MLC involvement, or you found mentors independently. Consider using it anyway—it helps MLC's visibility and doesn't hurt you.

Alternative - Acknowledgment: If you were trained through MLC or used MLC resources but did the work independently (no MLC member collaboration), you can acknowledge instead: "We thank the ML Collective community for feedback and support."

In some cases, we have had members/publications who use the MLC's affiliation and also acknowledge the resources/feedback explicitly in the acknowledgment section.

When NOT to Credit MLC[edit]

Don't use MLC affiliation when you had zero interaction with MLC, weren't trained through our programs, and used no MLC resources.

How Committees View MLC Affiliation[edit]

Positives: Shows you're part of a research community, can collaborate across institutions, and are resourceful.

What Counts for Authorship?[edit]

Authorship vs. Acknowledgment[edit]

Authorship = Substantial intellectual or practical contribution
Acknowledgment = Helpful feedback, encouragement, minor assistance

The Two Most Important Contributions[edit]

IMPLEMENTATION and WRITING are the most important factors.

Implementation:

• Writing core methods (not just running existing code)
• Experimental design and execution
• Major debugging that fundamentally changed results
• Substantial dataset creation/curation

Writing:

• Drafting multiple sections (intro, methods, results, discussion)
• Major revisions based on feedback (not just typo fixes)
• Structuring the narrative

Contributing to the Idea[edit]

Ideas alone don't count for authorship. Early-career researchers learn best from actively implementing. If you don't do any part of implementation, you don't count for authorship. Ideas without execution are acknowledgment-level contributions.

Other Contributions[edit]

• Literature review: If comprehensive and shaped project direction (gray area; discuss with group)
• Figure creation: If complex and central (gray area)
• Data annotation/collection: If substantial and skilled (not mechanical work)

What Doesn't Count for Authorship[edit]

• Just attending meetings
• Emotional support and encouragement
• Creating meetings and secretarial work (scheduling, note-taking)
• Early brainstorming without follow-through
• Suggesting papers to read
• Minor proofreading or typo fixes
• Running someone else's code with minor modifications

Documenting Contributions[edit]

During the project: Create shared doc where members log contributions weekly (implementation, experiments, writing, other substantial tasks). Focus group leads and MLC leads can also judge contributions.

Before submission: Each member lists specific contributions, can anonymously rate others (1-5 scale). Ratings inform discussion but don't automatically determine authorship. Have explicit authorship conversation and get agreement.

Authorship Positions & What They Mean[edit]

First Authorship[edit]

First author = person who drove the project: Did majority of implementation and writing, led experimental design, coordinated team, made key decisions when stuck.

NOT: Just "whoever had the initial idea" or "whoever wrote the most code" or automatically shared "to be nice."

Our philosophy: Everyone should get at least one first-author paper from their time with MLC, so this should not be a sticking point, as the current middle author could lead the next project within the group.

Co-First Authorship[edit]

Denoted with * or †

Appropriate when: Two people contributed roughly equally across implementation AND writing, both feel genuine ownership, you can clearly articulate equal and substantial contributions.

NOT appropriate when: Being nice to someone who didn't pull weight, one person did 70%+ of work, default because you can't decide.

For grad school: Committees will ask "What was YOUR specific contribution?"

Second and Middle Authorship[edit]

Second author is valuable but much less impactful than first. Middle authors signal contribution but not leadership.

For grad school: Long list of middle authorships without firsts is a red flag. Committees wonder if you can drive projects independently.

Strategy: One strong first-author paper + one strong second-author paper is stronger than five middle-author papers.

Common Authorship Scenarios[edit]

Scenario A: One person clearly leads[edit]

Group starts together but Person A drives most work. Person A is first author, others are co-authors based on actual contributions. Don't make everyone co-first "to be nice."

Scenario B: Project splits into multiple papers[edit]

Best approach: Separate papers with different first authors (Paper 1: Person A first; Paper 2: Person B first). Cross-cite each other. Everyone gets first-authorship experience.

Scenario C: Workshop expanding to conference[edit]

Default: Same first author unless project direction changed drastically or someone else did 70%+ of new work. Person A already drove workshop version; expanding continues their leadership.

Scenario D: Someone isn't pulling their weight[edit]

Have honest conversation: "We need you to complete X by Y date to remain a co-author." If pattern continues: "Based on contributions, we're removing you from authorship. You'll be acknowledged." Document everything, defer to MLC leads if needed.

Scenario E: External mentor provides guidance[edit]

If mentor provides ongoing substantial guidance (multiple rounds of detailed feedback, helped design experiments, met regularly, brought key resources), they typically deserve co-authorship (often last author). One round of feedback or general encouragement = acknowledgment only.

How Grad School Committees View Authorship[edit]

The Hierarchy[edit]

Aside from having publications, committees scrutinize co-authors, author position, and venue quality. First-author paper at solid conference (tier-B) with external validation > middle-author paper from famous lab.

Conference Papers >>> Workshop Papers[edit]

Committees weight conferences much more heavily. Conferences have rigorous peer review (~25% acceptance), workshops are often lightly reviewed (>50% acceptance). Conferences show you can complete substantial projects.

For applications: - 2 workshop papers, 0 conference → weak - 1 conference paper → competitive
- Multiple conference papers → strong

Workshop as first-year researcher = good progress. Still only workshops by application time = weak signal.

First Author Matters[edit]

What committees see on your CV:

• Smith et al. → "Smith led this"
• Jones and Smith et al. → "Smith contributed massively but Jones led"
• Jones et al. [Smith is 4th of 8] → "Smith helped but unclear how much"

They're asking: Can this person independently drive research?

Yellow Flags[edit]

• Long author lists with unclear contributions (8 authors, you're 5th, all MLC)
• Only middle authorships (no leadership demonstrated)
• All MLC authors with no external validation (mitigate: external mentors, competitive venues, external collaborators)
• Workshop-only record (by application time, need conference paper)
• Inconsistent research direction (paper 1: NLP, paper 2: vision, paper 3: robotics—looks random)

What Strong Applicants Have[edit]

Minimum competitive:

• 1 first-author conference paper (tier-A or tier-B)
• 1-2 workshop papers OR
• 1-2 second-author papers

Strong:

• 2+ first-author conference papers
• Mix of MLC and external collaborations
• Mix of MLC and external letters

The Key Question[edit]

"What was YOUR specific contribution?"

Bad: "I worked on a project about NLP"

Good: "I designed and implemented the core attention mechanism, ran experiments across 5 languages, and wrote the methods and results sections. My co-author handled data collection."

Practice articulating: what you implemented, what you wrote, what experiments you designed/ran, what problems you solved, how you collaborated.

On Focus Groups Papers: Practical Requirements[edit]

ArXiv Submissions[edit]

Always upload to arXiv after workshop submission, even if not accepted. Why: Google Scholar indexing, shows productivity despite rejection, establishes precedence, portfolio building.

We'll endorse your submissions after reading them. Reach out to MLC leads.

Final Presentations[edit]

All group members should present completed work to MLC community with arXiv links. This creates public record, provides practice, enables feedback, and celebrates work.

Timeline Expectations[edit]

Aim for first workshop submission within 3 - 6 months. If approaching 8+ months without submission, reassess scope or pivot.

The progression: Workshop submission (validate idea) → incorporate feedback, find external mentors → expand with guidance → conference submission (the real goal).

Workshop acceptance is starting point, not endpoint.

When Disagreements Happen[edit]

Conflict Resolution[edit]

Step 1: Direct conversation. Each person states their view, use contribution logs, try to reach consensus.

Step 2: Escalate to MLC leads (Abraham, Steven, Busayo, Mardiyyah). We'll review logs, talk individually, mediate. We aim for consensus but will make final call.

Step 3 (rare): Bring to Sunday call (anonymized if requested). MLC leads make final decision.

Guiding principle: Whoever did the work gets the credit.

Prevention[edit]

• Discuss authorship before project starts
• Document contributions as you go
• Check in regularly about expectations
• Raise concerns early
• Defer to MLC leads when uncertain

FAQ[edit]

Q: Can I have two first-author papers from one focus group?
A: Yes, if projects are sufficiently different or you're leading follow-ups with new contributions.

Q: What if I disagree with authorship decision?
A: Talk to focus group lead first. If unresolved, escalate to MLC leads.

Q: Can I remove MLC affiliation if I do all the work?
A: If you started in MLC or were trained through our programs, keep affiliation—it helps future members. If you brought a mostly-complete external project for minor feedback, it's optional.

Q: What if someone leaves MLC mid-project?
A: Depends on contributions. Substantial work = co-author. If substantial work was done but doesn't make it into the restructured/submitted version (only an inspiration) = acknowledgment. Minimal contribution = acknowledgment. Leaving MLC doesn't automatically remove authorship.

Q: Should I list "under review" papers on applications?
A: Yes. List as "Conference 2026 (under review)."

Q: How many authors is "too many"?
A: 1-3 ideal; 4-5 common; 6-8 getting crowded; 9+ red flag unless it is a genuinely large-scale project.

Q: How do I know if my contributions warrant first vs. second authorship?
A: Ask honestly: Did I drive this forward when stuck? Could this paper exist without my contributions? Did I make most key decisions? Did I do bulk of implementation and writing? If yes to most: first author. If unsure: probably second.

Last updated: January 2026

This is a living document. Questions or suggestions? Bring them to Sunday calls or message MLC leads.