Batch jobs
Sort or tag a thousand items in a single request. Annotate, summarize or extract findings across hundreds of papers.
Beaver keeps a ledger of every item. The work is finished, and reported, item by item.
File each item into the existing subject collection matching what the paper is about.
Limits are per batch job. Multiple batch jobs can run in a single or across multiple requests.
Built to finish the job
A batch job is a ledger of decisions. Beaver systematically works through every item and either changes it or records it as needing no change.
File each item into the existing subject collection matching what the paper is about.
Approval
Nothing runs until you approve it. The approval card shows the exact count, the goal of the batch job, the maximum credit costs, and a warning about any destructive library changes.
Progress reporting
You watch the job move while Beaver works: a live progress bar with the count so far, and a running tally of where items are landing. An item only counts as done when the work actually happened.
Receipt on completion
When the last item has a decision, the job closes with a receipt: how many items went where, how many needed no change and why, what could not be read.
Benchmarks
Batch jobs are a powerful tool for large-scale library management. We systematically evaluate performance using large-scale Zotero libraries. Comparing Beaver assigned tags with a gold-standard set of labels, we see high tag quality particularily when tags are clearly defined.
| The job | Without batch jobs Beaver 0.23 | With batch jobs Beaver 0.24 |
|---|---|---|
Sort 680 unfiled items into existing collections Items sorted, across three runs | 0 · 206 · 332 | 600 · 607 · 617 |
Tag 616 untagged items by subject Items given a subject tag that tells them apart, across three runs | 0 · 0 · 0 | 615 · 616 · 615 |
Tag quality Agreement with gold-standard labels (F1), across three runs | — | 90% · 91% · 91% |
Tag quality with a one-line tag definition Agreement with gold-standard labels (F1), across three runs | — | 96% · 97% · 97% |
Each case runs against a frozen 866-item Zotero library, reset to the same state before every run, on Beaver’s default model budget. Coverage only counts when items receive a tag or collection that actually tells them apart from the rest. Tag quality is scored against gold-standard labels produced by multiple independent labelers. Only papers where every labeler agreed are counted, and disputed papers are excluded.
What to ask for
Beaver automatically uses batch jobs for large-scale sorting, tagging, metadata editing, annotating, or for the systematic extraction of information across many items. Or just ask and Beaver will complete your request as a batch job.
Four habits that pay off
Name the population clearly
Clearly naming the population, like “All unfiled items”, works better than vague requests like “a few of my unfiled items”. That leaves less room for Beaver to guess what you mean.
Say what a good outcome looks like for one item
The goal is judged one item at a time. “File it into the existing subject collection that matches what the paper is about” gives Beaver something it can hold a single item against. “tidy up my library” does not.
Define your tags and collections, not just name them
Beaver tries to derive what a tag or collection means from its name and contents. A one-line definition maybe in your custom prompt gives every job more than the name to go on. That results in more consistent tags and collections across hundreds of items.
One kind of work at a time
Tagging, sorting, and writing notes are three different kinds of work. Use three separate requests so that each has clearly defined goals.