EDITORIAL PILOT / RES—001
One question.
Three independent answers.
Our first research benchmark records unedited output from Perplexity, Elicit and Consensus. This workspace collects evidence; it does not generate AI results or publish anything.
TEST PROTOCOL
Caffeine and sleep — literature evidence
Find three relevant peer-reviewed studies published before 2025 investigating caffeine intake and sleep quality in adults. For each: provide the exact title, authors, year, publication venue, DOI or stable identifier, main finding, study limitations and a working source link. If a study cannot be verified, explicitly say so. End with a cautious 120-word synthesis distinguishing correlation from causation.
Test all tools on the same date where possible. Note paid/free tiers, web access, search settings, retries and whether the software supports the exact task.
UNEDITED OUTPUT CAPTURE
Enter the real responses
Only results you actually generated belong here. These fields stay on your device until you export them.
0 / 3 responses collected
Verification checklist
- Every paper title, year and DOI resolves to the same publication.
- Every cited claim actually follows from the cited study; note the limits and study type.
- Results are unedited; record the first attempt and explain any retries.
- Mode, plan, date, attempts and costs are transparent.
- Publishing requires a second deliberate editorial approval.
This protocol is a software research benchmark, not medical guidance. Do not enter passwords, private client information or personal medical data.