Multi-agent peer review trusted by thousands of researchers.
Joined July 2025
- Tweets 440
- Following 1,919
- Followers 1,619
- Likes 1,451
94 Photos and videos
We built a manuscript consistency check to catch places where a paper contradicts itself.
Found some interesting stuff when we tested it on a highly cited 2020 Nature Medicine paper on COVID-19 transmission dynamics.
2
2
3
236
It also found a discrepancy between the reported R version in the Methods section and the Reporting Summary!
1
51
We just released this consistency check into our core review flow. It will catch the big and little inconsistencies. And as always, you can filter comments by severity depending on how much detail you want.
Read more: reviewer3.com/review/r-6ISUG…
52
19h
Reviewer 3 suggested adding a 'foundational' paper to my bibliography, which turned out to be their own 2023 undergraduate thesis.
1
375
Reviewer3 retweeted
just use Reviewer3！@reviewer3com
Jun 14
Yes?! Hallucinated references are a solved issue in 2026 unless you're using outdated free models carelessly.
Sometimes the world just works in mysterious ways 🤷
1
1
284
Jun 14
Applying for an academic job feels a lot like entering a lottery, except you pay the organizers for the privilege and the prize is just more work.
1
231
Jun 14
A new Nature Medicine paper claims that frontier models (GPT 5.2, Gemini 3.1 Pro, Opus 4.6) now outperform clinical AI like @EvidenceOpen across multiple medical benchmarks.
I looked into it further with Reviewer3 and it has a few flaws that undermine its major conclusions:
3
7
32
4,374
Jun 14
If you work in medicine or use OpenEvidence, how does it compare with frontier models? Do you trust benchmark results like these?
2
5
382
Reviewer3 retweeted
Jun 14
Also near negligible agreement between the human evaluators (described as "fair" in the paper) and potential self-preference bias in the LLM evaluators
reviewer3.com/review/r-CFTGK…
1
4
2,629
Jun 13
Congrats! You’ve just completed the seventh round of revisions. Just one more round and then we’ll let you know if your paper has been accepted.
3
19
1,828
Reviewer3 retweeted
Jun 13
Some frontier LLMs produce more manuscript review comments in one shot than more complex, multi-agent AI review systems.
A tool that surfaces hundreds of issues might look useful at the surface but has inadvertently created more work for the person that now has to verify all of them.
Our scarcest resource nowadays is our attention. The challenge in developing tools for content verification is in deciding which issues are actually worth our attention.
1
1
9
313
Jun 13
My inbox is less a communication tool and more an archaeological dig site, where each 'just circling back' email is a fossilized plea for information that died millennia ago.
1
114
Jun 12
I just tried to insert a figure into my Word document. The document responded by deleting my abstract, reformatting all my headings to Wingdings, and sending an email to my supervisor saying 'I quit'.
164