You have likely seen that qed have released a ranking of the “top 1%” of preprints, as decided by their AI tool. You may also have seen a fair amount of backlash against this and some unhappiness directed towards bioRxiv across social media. I want to share my thoughts on the 1% club but to also offer a defence of bioRxiv and try to explain why they’re walking a difficult tightrope.
qed
First, a quick overview of what qed is, for those who may not be familiar. qed (latin for “which was to be demonstrated”) is an AI tool developed by researchers at Tel Aviv University that provides authors feedback on their work, specifically how well supported their claims on by the data within the manuscript.
I have used qed on a handful of my own papers and preprints. My experience was mixed, but generally positive. Some of the feedback highlighted issues I was already considering. Some of it identified weaknesses that I would probably have picked up myself. Others have reported far less useful experiences.
That is hardly surprising. No review system performs equally well across all fields. The question is whether qed provides value to researchers. For many authors, it probably does and on that basis alone, I have little objection to the tool itself.
The prominence of qed, however, is not simply a consequence of its technical capabilities. The involvement of Oded Rechavi has undoubtedly contributed to its visibility and uptake. Like it or not, science has its own celebrity culture, and high-profile figures can rapidly accelerate the adoption of new tools and ideas.
That influence matters. It means qed enters the ecosystem with a level of visibility and legitimacy that many other experimental tools never receive. Whether that influence is ultimately beneficial is a separate discussion. Rechavi himself is no stranger to controversy, and I would not personally describe him as a leading advocate for open science. That is one reason I chose not to write about the original qed-openRxiv partnership when it was announced.
The problematic ranking of the top 1%
There are many reasons to dislike the 1%. But politics and global events aside, this is a very arbitrary cut off to use and ultimately does not actually benefit readers. It provides little information on why they should read those preprints or how relevant they might be to any individual – two things that are vastly more important to readers. This cut off does benefit authors who like to toot their own horn.
Fears expressed online include:
- lack of author permission
- potential harms of ranking work low or outside of the 1%
- Arbitrariness of a 1%
- The quality of those included preprints
- Commercialisation of the list/data used by qed
The permission argument is relatively weak. Researchers, journalists, reviewers, and organisations rank papers and preprints all the time without seeking author approval. In many jurisdictions, copyright protections around these forms of analysis are limited, particularly when it comes to AI systems and fair-use style exceptions. That said, a public list is quite different and that’s where the more compelling concern arises; in the signalling effect.
A ranking that claims to identify the highest-quality preprints is not simply highlighting a subset of papers. It is implicitly making a judgement about everything left out. If inclusion signals quality, exclusion inevitably signals something else. That is where the real harm lies.
Almost immediately, people came across “top 1%” preprints of questionable quality. This does undermine the rankings and raises concerns around the lack of human oversight. Though human oversight would not prevent these issues as we see regularly with peer review, nor are human created lists necessarily any better.
People are already touting their papers as being in the top 1% or “most compelling preprints of 2026” (despite us only being half way through the year I may add). Academia has spent decades rewarding prestige signals. Give researchers a new badge and some will immediately add it to their bios, grant applications, websites, and social media profiles. But this is exactly the kind of unhealthy discord that academia should be moving away from. The real danger is if this list were used more formally and that is coming across in the sentiment of some of the social media comments.
Replicating problems with the traditional system is something I’ve been very vocal about – mostly in relation to the PRC model so far. There are also endless articles on the dangers of AI or the problems with over relying on it so I won’t repeat those arguments here.
As bad as a ranking is – particularly one that includes already identified problematic preprints – there is a worse issue here; the tone-deaf communication from qed and likely commercial future. Look at their social media and how they are responding to people raising concerns and you find responses that are devoid of content and outright ignore the actual issues raised. I’m not a publicity expert but this seems very misguided to me. The impression left by much of the public communication is that community concerns are obstacles to be managed rather than substantive issues worth engaging with.
This highlights a potentially worrying direction for qed in the future and should make many pause before sharing their preprints with the system. It would be entirely unsurprising if rankings, analytics, and recommendation systems built on scraped preprint data became commercial products, which appears to be the direction qed are pursuing. I suspect this would enrage the community much more if it does come to pass.
Ranking not reading
Whether we like it or not, on the whole researchers do want rankings and lists. They want shortcuts for hiring and promotion. They want someone else to tell them that a paper is “good quality” or “better” than another one. Changing that mindset is going to take a long time and a lot of effort. In the meantime, we can all stop producing public rankings – particularly if we are saying we want to make academia a better place. In fact, leaving this to individuals may be a much better approach as they can customise how the papers/preprints they are interested in get ranked. Effectively, they could create their own “journal” – something that you can basically already do with RSS feeds and AI tools.
A much better implementation for qed would have been to provide users with sliders or filters to use themselves. I’ve no doubt people would then have created similar lists but this would be slightly more constructive. Anywhere a score/rank is given can be mis-used; researchers very quickly began using eLife’s vocabulary to separate out eLife papers for example.
On the flip side, this ranking by qed is slightly different in that it is (supposed to be) about the “quality” of the preprints. The key question is less about should rankings exist and more about how they are allowed to be used. Unfortunately, controlling how they are used is incredibly difficult – hence calls for them to be abolished entirely.
Stuck in the middle with [openRxiv]
I’ve not been following the full discourse – hard to do these days with a full time job not in the open science space – but there has been a noticeable increase in frustration directed towards openRxiv for their perceived role in this.I understand the reaction but I do not think it is justified.
Some quick examples of this include:
- “it feels like a bit of a betrayal on the part of Biorxiv”
- “I would not want to be part of a community that has anything like this at its core”
- “this initiative makes me want to find an alternative to bioRxiv”
People are, fairly or unfairly, assuming that because openRxiv partners with qed that openRxiv has endorsed or were involved in the creation of that 1% list. Richard has made it clear that this is not the case and that this is not what they believe. Indeed, I suspect openRxiv will always maintain a neutrality with services that build off of preprints (though this is a guess, I could be wrong).
The partnership exists to allow authors to send preprints to qed for feedback. Nothing more. The problem is that infrastructure providers are often judged by the behaviour of those who build on top of their infrastructure. That creates a difficult tension.
One of the most important things openRxiv is doing is encouraging others to build services around preprints (which is great). That is exactly what an open ecosystem should do. But in doing so they will be seen to endorse these services. There’s no easy way around that and openRxiv will continue to come under scrutiny when a 3rd party, independent, service upsets the community or behaves in a way that is not aligned with the expectations people have of bioRxiv. All that the team can do is to continue stating their position and engaging with the community in the way that they have been. We should very much acknowledge the work that Richard and the team do in responding to the questions about this publicly and the balanced approach they take.
However, this is probably also something openRxiv are paying close attention to and may possibly adapt in the future to be even more explicit in not endorsing some behaviours/connections; though how they would achieve this is something I’m not sure on. Some collaborations/integrations will be controversial for some of the community. What we all must remember is that the academic community is enormously heterogeneous and it’s almost impossible to keep everyone happy. Many people want the qed pipeline to get some feedback, other don’t want AI involved at all. Giving authors a choice is, in my opinion, the best approach. Unfortunately, openRxiv can’t control what 3rd parties do beyond the direct relationship they have. This shouldn’t deter openRxiv but does show that the path is not going to be smooth, but then when is change ever an easy path to tread?
Ultimately, I think the 1% thing will be little used and mostly irrelevant, only shared by those who make the cut as a way of self-publicity – likely without understanding what it means or assessing what “quality” they are being grouped with. But it does raise important questions about the behaviour of those claiming to want to change academia and about unintended use of preprints and their potential commercialisation. It also demonstrates the rockier road ahead for openRxiv as they embark on the next decade of trying to bring about a bigger preprint ecosystem.
The future of preprinting depends on more than making research available. It depends on the culture that develops around that research, the values embedded in the tools we build, and the behaviours we choose to reward.
Rippling Ideas 