Blinded comparison · Pilot study

Which edit works better?

Help us evaluate manuscript writing. Read a prepared fragment and compare two edits with the model identities hidden. Click the edit you would use in a paper to continue to the next sample.

We compare our fine-tuned model with its original base model using the same prompt and settings. Left and right are shuffled for every comparison. Both outputs retain their original wording; Markdown formatting markers are hidden.

Judge clarity, manuscript style, and whether the edit preserves the original facts and numbers. The set includes short fragments and longer, two-paragraph passages. These are fictional examples, not a general benchmark.

We save your random session ID, choices and timestamps to evaluate the models. These records expire after 90 days and are deleted on the next study request. A session token stays in this tab’s session storage. No personal manuscript text is requested. Privacy details

Which edit works better?

Original fragment Fictional example

Edit A

Edit B

Thank you for comparing.