PaperFoxPaperFox
Track Chairs

Auto Assign Reviewers

Automatically match submissions with reviewers using Smart Matching

Auto Assign uses Smart Matching to find the best reviewers for each submission. It scores reviewer-submission compatibility using text similarity and keyword overlap, then finds the globally optimal assignment across all submissions simultaneously.

Getting Started

  1. Go to Submissions → Assignments for your track
  2. Click "Auto Assign" in the toolbar
Find the Auto Assign button on the Review Assignments page

This opens the auto-assignment page where you configure matching settings.


Settings

Settings with constraint mode toggle and feasibility indicator

Target Reviewers per Submission

How many reviewers to assign to each submission. Default: 3. This is a target — submissions may receive fewer reviewers if there aren't enough eligible reviewers available (due to COI filtering or capacity limits).

Max Submissions per Reviewer

Upper limit on submissions assigned to a single reviewer. Default: 5. Use the toggle to choose how strictly this limit is enforced:

ModeBehavior
Soft limit (default)Reviewers may exceed the limit by up to 2 submissions if needed to ensure all submissions get assigned
Hard limitStrict cap — no reviewer will exceed this number, even if some submissions get fewer reviewers as a result

Use Soft limit when full coverage matters most — you want every submission to get reviewers, even if a few reviewers go slightly over capacity. Use Hard limit when reviewer workload is the priority — no one should be overloaded, even if it means some submissions get fewer reviewers than the target.

Feasibility Indicator

Below the settings, a color-coded banner shows whether your current configuration is feasible:

  • Green — Capacity is sufficient. All submissions can be assigned the target number of reviewers.
  • Amber — Capacity is tight. Soft mode will allow overflow to fit, but hard mode may leave gaps.
  • Red — Insufficient capacity. Some submissions will get fewer reviewers than the target. Consider adding more reviewers or increasing the max submissions per reviewer.

Scoring Method

Choose how text similarity between submissions and reviewer expertise is computed.

MethodHow It WorksCostSpeed
Text SimilarityBM25 text matching against submission titles and abstractsFreeSeconds
AI SemanticOpenAI embeddings — understands synonyms and related conceptsCredits (based on submission count)Seconds

Both methods also use phrase-level keyword matching (comparing submission keywords with reviewer expertise tags) in addition to text similarity.

How scoring works

Each reviewer-submission pair gets a score from 0–100: up to 70 points from text similarity (BM25 or AI embeddings) and up to 30 points from keyword overlap. The system then uses network flow optimization (Min-Cost Max-Flow) to find the assignment that maximizes the total score across all submissions — not just one submission at a time.

When to use AI Semantic

Select AI Semantic when you need the algorithm to understand that related concepts match — for example, "NLP" and "natural language processing", or "deep learning" and "neural networks". Submission titles, abstracts, and reviewer expertise tags are sent to OpenAI for embedding. Use Text Similarity instead if data privacy is a concern.


Strategy

Fill Gaps Only (Default)

Only assigns reviewers to submissions that don't have enough reviewers yet. Existing assignments are preserved — no existing reviewer-submission pairs are changed.

Use when: You've already made some manual assignments and want to fill in the rest.

Fresh Start

Generates new assignments for all submissions, replacing existing ones. When you apply the suggestions:

  • Existing assignments without a submitted review are removed and replaced with the new suggestions
  • Existing assignments with a submitted review are kept — the system never deletes work that reviewers have already completed

The confirmation dialog shows exactly how many assignments will be removed and how many will be kept before you apply.

Use when: Starting over or regenerating all assignments from scratch.


Options

COI Enforcement

By default, reviewers who have conflicts with submission authors are automatically excluded before the matching algorithm runs. Three rules are checked: self-conflict, co-author conflict, and declared conflict. If the track has turned COI enforcement off, co-author and declared conflicts stay in the pool and their suggestions carry a warning instead — self-conflicted reviewers are always excluded.

See Conflict of Interest Enforcement for full details on COI rules and how to encourage reviewers to declare their conflicts.

Balance Workload

When enabled, the algorithm favors distributing submissions evenly across reviewers, even if it means slightly lower expertise matches.

Use when: You want to avoid overloading some reviewers while others have few submissions.


Reviewers Without Research Profiles

Reviewers who haven't completed their research profile cannot be matched by expertise. They will be assigned to fill remaining slots after all profiled reviewers are placed.

If you see a warning about reviewers without profiles:

  1. Go to the reviewers page
  2. Send reminder emails to reviewers without profiles
  3. Wait for profiles to be updated, then run auto-assign again

See Research Profiles for details on what reviewers should include.


Running Auto Assign

  1. Adjust Target reviewers per submission and Max submissions per reviewer if needed
  2. Choose a constraint mode (Soft limit or Hard limit) — check the feasibility indicator to make sure capacity is sufficient
  3. Select a scoring method (Text Similarity or AI Semantic)
  4. Choose a strategy (Fill gaps only or Fresh start)
  5. Toggle options (Enforce COI, Balance workload)
  6. Click "Generate Assignments"

After generation completes, you'll see a suggestions page where you can:

  • Review each proposed assignment
  • See match scores and reasoning
  • Accept or reject individual suggestions
  • Apply accepted assignments in bulk
The suggestions page with the Manage Assignments link in the top right

Each suggested reviewer has a checkbox — clear any you don't want, or use "Deselect All" / "Select All" to start from scratch. The Reviewer Workload Distribution chart shows how many submissions each reviewer would end up with so you can spot anyone getting overloaded before you commit.

When you're happy with the selection, click "Apply Selected". A confirmation dialog summarizes the impact — for Fill gaps, it confirms that existing assignments are preserved; for Fresh start, it shows how many existing assignments will be removed and how many with submitted reviews will be kept.

The Apply Selected bar noting that applying does not notify reviewers

Applying does not notify reviewers

Applying assignments only records who is assigned to each submission — reviewers are not emailed. You can keep adjusting afterward: add, remove, or reassign reviewers manually, then send notifications when you're ready. Reviewers only hear about their assignments when you explicitly send them from Notify Reviewers on the assignments page.

Adjusting Assignments After Applying

The applied assignments aren't final. To fine-tune them by hand, click "Manage Assignments" in the top right of the suggestions page (or go to Submissions → Assignments for your track). There you can add or remove reviewers submission by submission, and only send notifications once everything looks right.


Tips for Best Results

  • Ensure reviewers have profiles — The more complete the research profiles, the better the matching quality
  • Use keywords on submissions — Submissions with topic keywords get better matches than those with only titles
  • Start with Text Similarity — It's free and fast, so you can iterate quickly
  • Try AI Semantic for final round — When you need the highest quality matches and have credits available
  • Check feasibility first — If the indicator is red, adjust your settings before generating

On this page