Epstein Files Full PDF

This draft was nominated for deletion on 7 January 2026. The result of the discussion was redirect.

This draft does not require a rating on Wikipedia's content assessment scale. It is of interest to the following WikiProjects:

Biography This draft is within the scope of WikiProject Biography, a collaborative effort to create, develop and organize Wikipedia's articles about people. All interested editors are invited to join the project and contribute to the discussion. For instructions on how to use this banner, please refer to the documentation.BiographyWikipedia:WikiProject BiographyTemplate:WikiProject Biographybiography

Since @WestwoodHights573 has decided to remove a tag without addressing the reason the tag was placed, I'll start a conversation here. What source says that Dao is "best known" for creating Mamba? I don't see any sources here that indicate he's even "known" for it, just that he developed it. To say someone is "known" or "best known" for something requires secondary coverage indicating that a person is widely known, not a mere assertion of the underlying fact. "Best known" appears to be, at present, a claim of original research. Dclemens1971 (talk) 21:10, 15 January 2026 (UTC)[reply]

Source nr. 4 explicitly says "Albert Gu at Carnegie Mellon University and Tri Dao at Princeton University developed the Mamba architecture, a refinement of the earlier state space sequence architecture". Along with a dozen of peer reviewed articles based on the subject, many of which state the same fact. It clearly attributes the creation of Mamba to Dao as a known fact in the field. Feel free to change the phrase if you do not like it. So far it seems to be widely used in 99% of Wikipedia articles. WestwoodHights573 (talk) 21:52, 15 January 2026 (UTC)[reply]

Sure, the source says he developed it. You're saying he's "known" for developing it. Who knows him for his involvement in developing it? That's a conclusion not supported by your sources, which tend to be primary or affiliated, and independent secondary coverage is necessary to validate what you're trying to do here. Dclemens1971 (talk) 23:14, 15 January 2026 (UTC)[reply]

The citation closest to this is https://towardsdatascience.com/understanding-flash-attention-writing-the-algorithm-from-scratch-in-triton-5609f0b143ea/ which only gives Dao a passing mention. Thus the first fact that it is used to cite fails validation.

Additionally, that the phrase "best known for" is used elsewhere does not mean it should be used anywhere unless a citation passing WP:42 states it. No precedent is ever set by any article for any other. If it were we would have a brutally fast descent into idiocracy

However, so far, Dao remains non notable. We are in the midst of being drawn in to busywork. There is no point in editing this at all until notability is proven properly. Then it can and should be rewritten

I think you have probably written what you wish to say about the subject, and then sought references after writing in order to cite what you say. This is WP:BACKWARDS. Instead, please read this essay, one of several which outline a process which will succeed assuming the subject to be notable. If it isn't notable then no amount of editing can help. We use the references in the process described in the essay to determine and verify notability. No suitable references means the subject is not notable, and it is time to stop.

@Dclemens1971, I think your thoughts on this would be valuable. 🇵🇸‍🇺🇦 Fiddle^Timtrent Faddle^Talk to me 🇺🇦‍🇵🇸 23:25, 15 January 2026 (UTC)[reply]

@Dclemens1971 To your point - "Who knows him for his involvement in developing it?" The research community, the academic community, and those who work in machine learning, as these are the people who produce further research on his algorithms, use them, and continue to study them. It is an ongoing topic, and he continues to further develop it, while others continue studying whatever he is developing, according to journal summaries. We likely cannot walk up to someone on the street and ask whether they know who Tri Dao is or what he is known for. It is subject specific. As I said, if you do not like something, feel free to edit the phrasing. I am not going to argue with you about one phrase. I shared whatever I wanted to share. Feel free to go over thousands of other actual Wikipedia articles using the same phrasing and personally remove it, I could not care less. WestwoodHights573 (talk) 23:32, 15 January 2026 (UTC)[reply]

Your conclusions that he is known in the academic community for all these things is your own judgment based on the fact that his work is cited. It doesn't say that's what he's best known for, and it doesn't say how highly the academic community views his work. That would require secondary coverage to validate, not your own conclusions based on primary sources and raw citation numbers.

@Timtrent Indeed, I am not going to spend more time on this in draft form. I hoped pointing out where sourcing needed to be improved would provide some guidance about how to get this topic ready for mainspace, should it ever be. It's on my watchlist and I will notice if it is moved back to mainspace. @WestwoodHights573, if you submit this again I *strongly* recommend you go through AfC.

I know Tri Dao isn't subject to your topic ban, but you should view the AfD (where no one else thought this was notable right now, whereas 5 people thought that it might be possible to write an article demonstrating notability in the near future) as a sign that you should also let this topic alone for the time being and find other places to edit around the project. Immediately recreating a draft for a topic that was just found non-notable by consensus borders on I can't hear you. Please listen to what colleagues around the project are saying. Good luck! Dclemens1971 (talk) 23:36, 15 January 2026 (UTC)[reply]

@Dclemens1971 Thanks! Best of luck to you too! And here is the perfect place for you to start with. Definitely notable and know for. Michael W. Pfaffl WestwoodHights573 (talk) 00:21, 16 January 2026 (UTC)[reply]

But I actually would add the quote here: "Tri Dao is known for his groundbreaking work on Flash Attention at Stanford, enabling a fast and efficient implementation of the Attention mechanism in Transformers, opening up possibilities for much longer sequence length in GPT-4, Claude Anthropic as well as in images, video and audio." https://www.vectara.com/business/resources/podcasts/generation-ai-podcast-episode-3-tri-dao WestwoodHights573 (talk) 01:19, 16 January 2026 (UTC)[reply]

The point is not that you know this. It requires a citation in WP:42 grade sourcing. The burden is yours, WestwoodHights573, because it is you who wishes this to be an article. The choice is binary: cite it or remove it. 🇵🇸‍🇺🇦 Fiddle^Timtrent Faddle^Talk to me 🇺🇦‍🇵🇸 23:36, 15 January 2026 (UTC)[reply]

@WestwoodHights573, I'd like to ask about this section of the draft.

In 2024, he introduced part 2 of his research, Mamba Vision, a continuous-time model that excels at modeling continuous data. It has been primarily used in Large Language Models, as well as other sequence modeling applications such as audio, processing, genomics and time series analysis by Mistral AI, AI21 Labs, IBM Research, academic labs and open source.

For which you have cited two sources:

• Soares, Eduardo Almeida; Brazil, Emilio Ashton Vital; Shirasuna, Victor Yukio; Zubarev, Dmitry; Cerqueira, Renato; Schmidt, Kristin (2025-06-09). "A Mamba-Based Foundation Model for Materials". npj Artificial Intelligence. 1 8. doi:10.1038/s44387-025-00009-7.
• Bergmann, Dave (2025-07-07). "What Is A Mamba Model? | IBM". IBM Research. Retrieved 2026-01-16.

Neither of those sources mention Mamba Vision. In fact, one of the first hits when I searched for the topic was:

• Ali Hatamizadeh; Jan Kautz (2024). "MambaVision: A Hybrid Mamba-Transformer Vision Backbone". arXiv:2407.08083 [cs.CV].

That is, a paper not written by Tri Dao. Could you clarify this apparent discrepancy, and why you attributed Mamba Vision to Tri Dao? Cheers, SunloungerFrog (talk) 10:32, 16 January 2026 (UTC)[reply]

@SunloungerFrog you can see that the two sources you mentioned are attributed to the entire paragraph called "Mamba and state space sequence models", sources all peer reviewed, cite Dao, and explain what Mamba is and how it is used for. If you read the source "Bergmann, Dave (2025-07-07). "What Is A Mamba Model? | IBM". IBM Research", you will find that it says : "Some, such as Mistral AI’s Codestral Mamba, are pure Mamba models", and the source itself is from IBM Research, which is all factually correct, and more sources can be found via a simple google search regarding this specific paragraph.

The source "Soares, Eduardo Almeida; Brazil, Emilio Ashton Vital; Shirasuna, Victor Yukio; Zubarev, Dmitry; Cerqueira, Renato; Schmidt, Kristin (2025-06-09). "A Mamba-Based Foundation Model for Materials". npj Artificial Intelligence. 1 8. doi:10.1038/s44387-025-00009-7." focuses entirely on chemical and molecular applications of Mamba-based SSMs, in relation to the genomics mentioned in the paragraph.

Then moving on into visual Mamba. In May 2024, Dao published follow-up work titled Mamba-2 (SSD), which refined the Mamba architecture to enable subsequent adaptations of Mamba models to computer vision by independent research groups, including work known as Vision Mamba. So I am not attribution anything to anyone out of my head, and explaining in one sentence how the project is evolving, the original published by the peer review conference can be added - https://proceedings.mlr.press/v235/dao24a.html Again, I condensed the sources, to make the text more clear. WestwoodHights573 (talk) 18:07, 16 January 2026 (UTC)[reply]

No. you can see that the two sources you mentioned are attributed to the entire paragraph called "Mamba and state space sequence models" No. It is a section with two paragraphs, the first one of which is entirely unsourced, and the second of which contained those two sources which said nothing about Mamba Vision! I suggest that you re-read my post on your talk page about verification and WP:BLP. You must not be sloppy with sources. Maybe you should also check out WP:CITE to help you put online citations to reliable sources in the right place. Cheers, SunloungerFrog (talk) 18:54, 16 January 2026 (UTC)[reply]

@SunloungerFrog No, it does not look like that to me. I selected the same sources again via the re-use and the citation are shown in the same way, at the end of the section. I did not put them after every sentence. Both of these sources are published by reliable , peer reviewed and independent places such as IBM and Nature journal. And these 2 sources are absolutely relevant to the section Mamba and state space sequence models". I also did not say that those two sources say anything about Mamba vision and shared which source does. I added that source in my reply to you here and in the text as well. WestwoodHights573 (talk) 19:14, 16 January 2026 (UTC)[reply]

@WestwoodHights573, I'd like to ask about this section of the draft.

Dao has served as an organizer of workshops on efficient systems for foundation models at the International Conference on Machine Learning and as an area chair for conferences including COLM, ICLR, and ICML. He has also acted as a reviewer for primary machine learning conferences and journals, including NeurIPS, ICML, ICLR, and AISTATS.

For which you have cited one source:

• Poli, Michael; Massaroli, Stefano; Nguyen, Eric; Fu, Daniel Y.; Dao, Tri; Baccus, Stephen; Bengio, Yoshua; Ermon, Stefano; Re, Christopher (2023-07-03). "Hyena Hierarchy: Towards Larger Convolutional Language Models". Proceedings of the 40th International Conference on Machine Learning. PMLR: 28043–28078.

I am at a loss as to how that source backs up the preceding claims. Could you clarify please? Cheers, SunloungerFrog (talk) 10:40, 16 January 2026 (UTC)[reply]

@SunloungerFrog You can see an older draft where I had multiple links from the procedures at these conferences, which would be too reference bombed, so I left one of them instead- the ICML. I do not mind putting it back. Is that necessary to have 7-10 citations for a paragraph? WestwoodHights573 (talk) 17:26, 16 January 2026 (UTC)[reply]

You need a reference for each claim. I would put them after each of the conferences/journals in question. And I'd advise against using the word "including", because then you are implying, without a source, that there are others. I might further suggest that this is run-of-the-mill stuff that many academics do and I would not necessarily include it in an encyclopedia entry myself. Cheers, SunloungerFrog (talk) 17:41, 16 January 2026 (UTC)[reply]

@SunloungerFrog I actually agree about this specific part. That is why I removed some of the conferences proceeding already before submitting the draft, because I thought that it is just too much adding 5+ or so sources for every conference separately to a very small paragraph. WestwoodHights573 (talk) 18:13, 16 January 2026 (UTC)[reply]

I also have a question about your comments to this part. @SunloungerFrog Why are you saying this is promotional or that we need secondary sources? This is a known fact, a person developed something important that is widely used by the biggest large language models, which is an important achievement. It is not the main fact of the draft, so I think primary sources are fine.

For the statement that says Open AI is using FlashAttention", the most reliable source is Open AI confirming it, nobody but Open AI know better what their own researches are using. I attached the research published by Open AI and others about this. I think primary sources for this paragraph are more that fine.

"It has also been integrated into widely used machine learning frameworks and has influenced subsequent research on memory efficient attention mechanisms such as PyTorch, Jax, Huggingface Transformers, and Microsoft DeepSpeed, and used to accelerate training and inference of large language models by organizations including Meta, Microsoft, OpenAI, and Google." WestwoodHights573 (talk) 18:30, 16 January 2026 (UTC)[reply]

Because the original single source [1] is written by the maintainers themselves, and it's probably overegging the pudding. Statements like this We've been very happy to see FlashAttention being adopted by many organizations and research labs to speed up their training / inference (within 6 months after FlashAttention's release, at the time of writing). This page contains a partial list of places where FlashAttention is being used. If you'd like to add links to your organization / product / codebase, please open a PR or email us. We'd very much like to hear from you! are clearly not a neutral point of view, and quite promotional. Cheers, SunloungerFrog (talk) 18:42, 16 January 2026 (UTC)[reply]

@SunloungerFrog Yeah, I saw that. I added that source because it combined all these companies, instead of addict multiple. It works the same way as GitHub, it allows the developers to see who is working on what, they pull data from public GitHub sources. Nonetheless, I added primary sources from Open AI, Meta and others instead, per your suggestions. WestwoodHights573 (talk) 19:04, 16 January 2026 (UTC)[reply]

@Theroadislong: Your comment I am minded to decline this draft, but this should probably be reviewed by an admin so that the version that was at WP:AFD can be compared. Tri Dao was redirected not deleted so the last version of the page before it was 'deleted' is available from the page history. Comparing the last version of Tri Dao before redirection with the creating edit of Draft:Tri Dao gives a good match. See Earwig's Copyvio Detector.Hope this helps. Regards. John B123 (talk) 21:14, 17 January 2026 (UTC)[reply]

I am here from a group chat after seeing a message from a fellow student, @WestwoodHights573. He shared that he was relentlessly harassed and prohibited from commenting on reviews that contained incorrect information. There are more than 130 students in this group chat who study computer science and are fully capable of defending this page against misleading claims. I was very surprised to read all of this.

Below, I am sharing Google Doc comments that @ was prohibited from posting. It seems biased to not allow someone to write or share them. With your permission, I am adding these comments here. Regardless, I will also write my own comments and encourage other students from the group chat to participate, because Tri Dao is an incredibly well-known professor, and many students are on waitlists to take a class with him. I want to to address comment by @Timtrent, who said that Tri Dao is not an academic, and I agree with every word written by Westwood.

WestwoodHights573: "I’d like to address comments made by “meet any of the eight academic-specific criteria”. Would you please provide your argument as to on what rationale does Dao not meet Criterion 1?

Criterion 1 states as follows “1. The person's research has made significant impact in their scholarly discipline, broadly construed, as demonstrated by independent reliable sources. See also notes to Criterion 2, some of which apply to Criterion 1 as well. The most typical way of satisfying Criterion 1 is to show that the academic has been an author of highly cited academic work – either several extremely highly cited scholarly publications or a substantial number of scholarly publications with significant citation rates. Reviews of the person's work, published in selective academic publications, can be considered together with ordinary citations here.”

I am adding a comment left by another user “I don't think that a 2025 collaboration [2] with Dao makes the March 2024 publication by IBM non-independent of Dao. Indeed, the 2024 publication by IBM was released almost contemporaneously with the MAMBA paper. Katzrockso (talk) 04:19, 8 January 2026 (UTC)”[reply]

I also included multiple citations from several extremely highly cited scholarly publications and a substantial number of scholarly publications with significant citation rates, such as Nature, and others. Do you have a reasonable justification on why those publications are not notable, not reliable, or maybe not peer reviewed for some reason and as a result of that should be dismissed?

Or perhaps why we should dismiss the fact that notable contributions created by Dao, such as Flash attention and Mamba, both deemed notable by wikipedia, both part of the scholarly discipline such as machine learning/ai space, or computer science, widely used by all exiting major AI research labs, and models such Chat GPT, Antropic, llama are not “significant impact” for some reason?

Someone else also mentioned “ It does look like he passes WP:NPROF#C1 with 6 papers with > 1K cites and one with > 7K. These indicate that his peers consider his work notable”. Do you have a reasonable justification to dismiss this argument?

In AfD about this subject you also wrote, “An assistant professor does not pass WP:NACADEMIC for their role. They must go above and beyond. This one has not. Thus we return to WP:BIO and WP:GNG neither of which they pass. They are a WP:ROTM person doing their job”.

Is that your personal opinion? While I appreciate your note, are you able to justify this statement with a policy based argument? WP:NACADEMIC policy says nothing about an “assistant professor” to not be considered an academic. I also cannot find one source on wikipedia or outside that would say so. https://teknopedia.ac.id/wiki/Associate_professor WP:ROTM used by you also says nothing about an associate professor at an academic institution. It does not look objective to me, and statements like “associate professor is not an academic” are factually wrong. If it is a personal opinion, I’d suggest another editor to review the page."

RaDiumMM (talk) 17:27, 24 January 2026 (UTC)[reply]

Can I hear quacking? John B123 (talk) 17:38, 24 January 2026 (UTC)[reply]

This is clear WP:MEAT and should be hatted. WestwoodHights is not permitted to use proxies to post comments while blocked. Dclemens1971 (talk) 17:49, 24 January 2026 (UTC)[reply]

I agree with the points raised in the group chat and shared by a fellow student, and I am here to discuss those points. I do not see any reason other than something personal for someone to ignore this and claim that Tri Dao is not an academic. I am a graduate student and took a class with Professor Tri Dao last spring. I had no idea that he was not a professor at an academic institution. RaDiumMM (talk) 17:57, 24 January 2026 (UTC)[reply]

I am more surprised by how many hostile comments students have received on Wikipedia. I personally received 4 hostile comments for making a reasonable, policy based argument. It is astonishing to see something like this. I would appreciate the opportunity to speak with real editors who are here to edit and to participate in a genuine discussion. If you @Dclemens1971 are not interested in discussion, it may be better to step away from the conversation. I took a class with Dao at Stanford, and Stanford university has a fake academic Tri Dao teaching courses. It is certainly something new and unheard of. RaDiumMM (talk) 18:13, 24 January 2026 (UTC)[reply]

I actually have an idea. @ Timtrent, do you have any connection to Berkeley by any chance? I am fairly certain that only someone from Berkeley would say that our professor is not an academic. Technically, it is not Stanford’s fault that Berkeley loses to us so all the time. Just a bit of healthy competition. RaDiumMM (talk) 18:19, 24 January 2026 (UTC)[reply]

@RaDiumMM No-one has suggested they are not an academic. They simply have not been shown to pass WP:NACADEMIC.

Do not impute motivation to me. You are required to be civil at all times, and not to attack other editors, however aggrieved you may feel about something. 🇵🇸‍🇺🇦 Fiddle^Timtrent Faddle^Talk to me 🇺🇦‍🇵🇸 18:36, 24 January 2026 (UTC)[reply]

@Timtrent It’s just an observation. It was someone on Wikipedia who blocked the Wi-Fi of actual student housing 😂 No jokes, trying to be very civil. I’m serious, trying to keep this civil, who does that? That kind of behavior screams pity.

This is what you wrote: An assistant professor does not pass WP:NACADEMIC for their role. They must go above and beyond. This one has not. Thus we return to WP:BIO and WP:GNG neither of which they pass. They are a WP:ROTM person doing their job. 🇵🇸‍🇺🇦 Fiddle^Timtrent Faddle^{Talk to me} 🇺🇦‍🇵🇸 23:59, 14 January 2026 (UTC).[reply]

Looks like the links didn’t get copied in my response above. You still haven’t addressed the point I made. Now you’re just saying he hasn’t been shown to pass #academic and ignoring the actual rebuttal. RaDiumMM (talk) 18:45, 24 January 2026 (UTC)[reply]

Like, why ban students in student dorms from editing Wikipedia about their own professor? I was forced to make an account and required to add my email just to make a comment. It’s ridiculous and not civil. I am here for discussion. It would be great to get a real rebuttal from you addressing the actual arguments why all those citations, papers and rest should be ignored. RaDiumMM (talk) 18:51, 24 January 2026 (UTC)[reply]

@RaDiumMM Until you start being civil I have absolutely no interest in any further conversation with you. The item you quotes above is, with precision, the case with this draft. Prove Tri Dao passes WP:NACADEMIC and this can go forwards. Fail and it will not.

I have no interest in whether it is a pass or a fail. Wikipedia is improved either way. 🇵🇸‍🇺🇦 Fiddle^Timtrent Faddle^Talk to me 🇺🇦‍🇵🇸 18:51, 24 January 2026 (UTC)[reply]

@Timtrent I am being civil. I will copy the reason why. Very civil. Policy quote: WP:NACADEMIC

Criterion 1. The person's research has made significant impact in their scholarly discipline, broadly construed, as demonstrated by independent reliable sources. See also notes to Criterion 2, some of which apply to Criterion 1 as well. The most typical way of satisfying Criterion 1 is to show that the academic has been an author of highly cited academic work – either several extremely highly cited scholarly publications or a substantial number of scholarly publications with significant citation rates. Reviews of the person's work, published in selective academic publications, can be considered together with ordinary citations here.

Can you make an argument why Dao not meet Criterion 1? We got student folks to fix the problem, very civil. It is written above already, and I repeat: Do you have a reasonable justification on why Nature and those publications are not notable, not reliable, or maybe not peer reviewed for some reason and as a result of that should be dismissed?

Nature, IBM is not reliable, great, will remove it. Otherwise, it does not seem civil, it screams sus that out of 5 people, only I was able to make a Wikipedia account to ask these questions in the first place.

This seems to be notable Jonathan Shewchuk, and for some weird reason pass #academic RaDiumMM (talk) 19:09, 24 January 2026 (UTC)[reply]

@RaDiumMM Then sort it out in the draft and resubmit it. Arguing solves nothing. I have no interest either way. 🇵🇸‍🇺🇦 Fiddle^Timtrent Faddle^Talk to me 🇺🇦‍🇵🇸 19:17, 24 January 2026 (UTC)[reply]

@Timtrent I am sorting this out here and being very civil. Do you have a reasonable justification for why Nature and those other publications are not notable, not reliable, or maybe not peer-reviewed, and as a result should be dismissed?

This is a dialogue, right? A discussion. #academic is a policy. So reviewers here are just putting someone’s personal opinion above the policy? I’d really like to know. I won’t have anything to discuss if your stance is that Nature, IBM, or the rest aren’t significant and criterion 1 should be ignored, just based on your personal opinion. No offense, I’m saying this very respectfully to make that clear. Are we really ignoring publications on the basis of “I think so and I don’t like this professor, he’s not an academic”? Cause this is how your answer reads like. RaDiumMM (talk) 19:26, 24 January 2026 (UTC)[reply]

@Timtrent In case you have a personal connection to Berkeley and don’t want to get into a discussion, not a big deal and hard feelings. I’m submitting the page and ask the editors to have a conversation and a civil discussion on the Talk page about Criterion 1 and why Dao is not an academic. I want to clear your position first.

Let me know if we need to invite folks working with Dao to personally comment on the talk here about their scholarly publications citing Dao, his notable research creating mamba and Flash attention, in peer reviewed journals, because Nature and others do not satisfy criterion 1 as being highly reputable scholarly publications, for some undisclosed reason. RaDiumMM (talk) 20:24, 24 January 2026 (UTC)[reply]

I have no idea why Westwood was prohibited to address a comments which is absolutely false. I had a wow reaction. If @Northernhenge thinks his comment is not false, it is not objective at all. I'd hope it was not deliberate. I and other students can address every point made my Westwood.

Westwood: “The recent additions have all been about the software products, not about the subject of the article, and not about the subject's notability. Northernhenge (talk) 23:46, 19 January 2026 (UTC)[reply]

Can you please clarify what software products you are talking about? We do not have any software products in the article at all, and the person did not create any software products.

FlashAttention is an algorithm, developed in research, a mechanism and not a product. Sentence 1 says “[In 2022, Dao introduced FlashAttention…]. Sentence 2 says “Dao's research addresses the computational and memory challenges of large-scale machine learning models]. Sentence 3 says [The Generation AI podcast said that Dao's work on Flash Attention  enabled] a fast and efficient implementation of the Attention mechanism in Transformers ..]. Then it is followed by explanation on how he and his research has made a significant contribution to the scholarly field [Alex Dremov, an AI engineer, described FlashAttention as "a revolutionary technique that dramatically accelerates the attention mechanism..]. Which is a requirement for the criterion 1.

Then it is followed by how exactly his contribution is used in the research labs [It has also been integrated into widely used machine learning frameworks and has influenced subsequent research on memory efficient attention mechanisms]], as citations are used sources from Microsoft Lab, Facebook lab, OpenAI Lab - which are all research places.

Next section Mamba, added by me. Mamba is a neural sequence model (read the hyperlink) with a specific architectural design.  Calling Mamba a “software product” is factually wrong in a technical or academic context. Sentence 1 [Dao created Mamba in 2023…], sentence 2 is how it is used. Sentence 3 [In May 2024, Dao published follow-up work titled Mamba-2 (SSD)..]. Sentence 4 explains why it is significant  [has been primarily used in Large Language Models, as well as other sequence modeling applications such as audio, processing, genomics and time series analysis by Mistral AI, AI21 Labs, IBM Research, academic labs and open source]].

To sum up my point, there are zero additions about software products. It is the description of his research work and its role in the scholarly discipline such as artificial intelligence and machine learning, or more broadly computer science. I think you should either remove your comment or edit it, because it is factually wrong." RaDiumMM (talk) 17:37, 24 January 2026 (UTC)[reply]

See WP:MEATPUPPET. Theroadislong (talk) 17:38, 24 January 2026 (UTC)[reply]

@Theroadislong I can rewrite it myself if that is your problem. I openly said I read this in group chat and it seems crazy to me. Would you please help explain what software @Northernhenge is talking about? There is no software mentioned anywhere in the article. Wikipedia is an open source for millions of people, and I have the right as a user to ask this questions. I, as a computer science student, am happy to explain what software is. RaDiumMM (talk) 18:03, 24 January 2026 (UTC)[reply]

@RaDiumMM if Tri Dao has been your instructor, as you've said, the fact that you know the subject personally could give rise to a conflict of interest. Please read WP:COI carefully before editing on this subject. Your attitude toward the other editors here (which, perhaps coincidentally, is identical to WestwoodHights' before they were blocked) will be an impediment to your participation in this collaborative project, through. Dclemens1971 (talk) 19:03, 24 January 2026 (UTC)[reply]

You actually have no "rights" at all here! Theroadislong (talk) 19:04, 24 January 2026 (UTC)[reply]

@Dclemens1971 I think I was very, very transparent and honestly stunned, and I came here from the group chat. I don’t see what form of conflict of interest I have if I’m not the one making this page. I’m fine with the page as it is. I have nothing personal to add. I don’t think students have a deep personal relationship with the 30 instructors they might have in college. In fact, random people seem to have a weirdly unjustified attitude towards Dao here, and I got a bunch of comments that read hostile. Including this one below:

“You actually have no 'rights' at all here! Theroadislong (talk) 19:04, 24 January 2026 (UTC)”

I came here with a real argument, saying there is a wrong comment, where is software, and someone writes me that I have no rights. It reads very civil. Then what is "civil" in here? RaDiumMM (talk) 19:17, 24 January 2026 (UTC)[reply]

The only reviewer this topic is addressed to is @Northernhenge. I hope he can sort this out and edit the comment, and have a civil discussion. Maybe he wrote that by mistake and had another article in mind, not Dao. Right now the comment is false and does not reflect what the article is about. I am here to explain what the software is in case someone is not familiar with it. RaDiumMM (talk) 19:35, 24 January 2026 (UTC)[reply]

I was referring to FlashAttention and Mamba. The material in this article about those subjects would be better in their own pages, so it would be better to put the Mamba material into the Mamba article. I don’t know if FlashAttention and Mamba should be called algorithms or models or software or something else, but my point was that saying more about them here didn’t tell me anything new about Dao’s notability. --Northernhenge (talk) 20:55, 24 January 2026 (UTC)[reply]

@Northernhenge got it. It was explained that FlashAttention and Mamba are absolutely not software products. A fellow student, the creator of this article, provided explanations for everything they wrote. I think the explanations are solid and clear up any possible confusion.

From what I can see, both of them - mamba/ flash attention already have separate pages explaining in detail what they are. The information in this article is limited to Dao’s involvement in creating them. I am here to submit the draft. Do you want me to remove your comment? It is wrong now.

Do you think Dao’s contribution should be removed entirely? That is his key academic work. It does not make sense to remove it. Maybe it could be trimmed, but not removed. I will see what to remove now. RaDiumMM (talk) 21:07, 24 January 2026 (UTC)[reply]