Michael Saxon
@saxon.me
Postdoc at UW and Doctor of NLP/Vision+Language from UCSB
Evals, metrics, multilinguality, multiculturality, multimodality, and (dabbling in) reasoning
100% Product of public schools
https://saxon.me/
Evals, metrics, multilinguality, multiculturality, multimodality, and (dabbling in) reasoning
100% Product of public schools
https://saxon.me/
Pinned
Michael Saxon
@saxon.me
· Nov 5
🆕 from us at #EMNLP: Are LMs better at answering questions about Germany in German than in French? Is national knowledge linguistically contingent?
Interestingly, only for some multilingual models is this true. Aya knows China best in Chinese, but LLaMA's best in English always.
Interestingly, only for some multilingual models is this true. Aya knows China best in Chinese, but LLaMA's best in English always.
Seems like a bad idea to join a startup as an employee when your company can get non-acquired for acquisition money to snipe the execs and take the tech while the company remains
www.cnbc.com/2025/12/24/n...
Well, not sure how easy it is to merge both technologies, but it surely takes out a competitor on the increasingly vital serving side
Well, not sure how easy it is to merge both technologies, but it surely takes out a competitor on the increasingly vital serving side
Exclusive: Nvidia buying AI chip startup Groq's assets for about $20 billion in largest deal on record
Nvidia is making its largest purchase ever, acquiring assets from nine-year-old chip startup Groq for about $20 billion.
www.cnbc.com
December 24, 2025 at 11:38 PM
Seems like a bad idea to join a startup as an employee when your company can get non-acquired for acquisition money to snipe the execs and take the tech while the company remains
You had a good run 60 minutes!
December 22, 2025 at 3:56 AM
You had a good run 60 minutes!
Are there any pieces on peer review in CS that you like? If so, I'd love it if you'd share them with me :)
In particular I'm interested in stuff covering AI/ML/NLP/CV, which does any mix of diagnosing problems with peer review and publication or proposes fixes.
In particular I'm interested in stuff covering AI/ML/NLP/CV, which does any mix of diagnosing problems with peer review and publication or proposes fixes.
December 19, 2025 at 11:03 PM
Are there any pieces on peer review in CS that you like? If so, I'd love it if you'd share them with me :)
In particular I'm interested in stuff covering AI/ML/NLP/CV, which does any mix of diagnosing problems with peer review and publication or proposes fixes.
In particular I'm interested in stuff covering AI/ML/NLP/CV, which does any mix of diagnosing problems with peer review and publication or proposes fixes.
From what I've heard through the grapevine he certainly isn't above saying this in the office
December 19, 2025 at 3:08 AM
From what I've heard through the grapevine he certainly isn't above saying this in the office
I had a surprisingly nice interaction which on X comma the everything app today
Some medium-small nonanonymous MLposter posted a screenshot of one of our colleague's hyphenated names with a "wow you see interesting names in research" caption
it was drawing strange racial remarks from the anons 1/3
Some medium-small nonanonymous MLposter posted a screenshot of one of our colleague's hyphenated names with a "wow you see interesting names in research" caption
it was drawing strange racial remarks from the anons 1/3
December 17, 2025 at 8:11 AM
I had a surprisingly nice interaction which on X comma the everything app today
Some medium-small nonanonymous MLposter posted a screenshot of one of our colleague's hyphenated names with a "wow you see interesting names in research" caption
it was drawing strange racial remarks from the anons 1/3
Some medium-small nonanonymous MLposter posted a screenshot of one of our colleague's hyphenated names with a "wow you see interesting names in research" caption
it was drawing strange racial remarks from the anons 1/3
I'm working on a piece about peer review and conference publication in AI/ML/NLP/CV, and I need YOUR help to understand how the community uses conferences and peer review in paper discovery.
Your four short multiple choice answers will help a lot :)
Your four short multiple choice answers will help a lot :)
Paper discovery survey
I'm working on a piece about peer review and want to understand how people in the (primarily AI/ML/NLP/CV) CS community discover new papers to read.
I am only talking about totally incidental discove...
docs.google.com
December 16, 2025 at 1:22 AM
I'm working on a piece about peer review and conference publication in AI/ML/NLP/CV, and I need YOUR help to understand how the community uses conferences and peer review in paper discovery.
Your four short multiple choice answers will help a lot :)
Your four short multiple choice answers will help a lot :)
Reposted by Michael Saxon
LLMs didn’t move language modeling research from linguists to AI people, they just moved it from computer scientists who thought language was interesting to computer scientists who thought language was boring
December 12, 2025 at 7:38 PM
LLMs didn’t move language modeling research from linguists to AI people, they just moved it from computer scientists who thought language was interesting to computer scientists who thought language was boring
December 11, 2025 at 7:33 PM
I don't understand the point of techno optimism as an ethic or identity. It's owned by "the worst people" in tech (ie, Andreessen) because it was created to shield them from crit. Solution is to be specifically optimistic about specific tech, which no longer makes you an ideological techno optimist
Really refuse to let the worst people have ownership of techno-optimism
December 11, 2025 at 4:44 PM
I don't understand the point of techno optimism as an ethic or identity. It's owned by "the worst people" in tech (ie, Andreessen) because it was created to shield them from crit. Solution is to be specifically optimistic about specific tech, which no longer makes you an ideological techno optimist
Is intelligence inevitable, and would dinosaurs have become human?
Claim here is "no" for reasons like something about mammals making them more likely to develop intelligence, or that the post asteroid predator-free environment bred competition, favoring intelligence
youtu.be/8Gh2gycaavI
Claim here is "no" for reasons like something about mammals making them more likely to develop intelligence, or that the post asteroid predator-free environment bred competition, favoring intelligence
youtu.be/8Gh2gycaavI
What if The Dinosaurs Didn't Go Extinct
YouTube video by Nick Longrich Evolution and Paleontology
youtu.be
December 10, 2025 at 8:37 AM
Is intelligence inevitable, and would dinosaurs have become human?
Claim here is "no" for reasons like something about mammals making them more likely to develop intelligence, or that the post asteroid predator-free environment bred competition, favoring intelligence
youtu.be/8Gh2gycaavI
Claim here is "no" for reasons like something about mammals making them more likely to develop intelligence, or that the post asteroid predator-free environment bred competition, favoring intelligence
youtu.be/8Gh2gycaavI
Here is my re-recording of the "What's Missed" portion of our "Science of Benchmarking" tutorial from NeurIPS 2025. Enjoy!
www.youtube.com/watch?v=mDhB...
www.youtube.com/watch?v=mDhB...
What's Missed in the Science of Benchmarking [NeurIPS 2025 Tutorial Excerpt]
YouTube video by Michael Saxon (NLP & Generative AI research)
www.youtube.com
December 10, 2025 at 1:40 AM
Here is my re-recording of the "What's Missed" portion of our "Science of Benchmarking" tutorial from NeurIPS 2025. Enjoy!
www.youtube.com/watch?v=mDhB...
www.youtube.com/watch?v=mDhB...
I have bootlegged our panel discussion from our NeurIPS "Science of benchmarking" tutorial to youtube!
Featuring @ofirpress.bsky.social @saining.bsky.social @idavidrein.bsky.social @efleisig.bsky.social and Wenda Xu
www.youtube.com/watch?v=d_zX...
Featuring @ofirpress.bsky.social @saining.bsky.social @idavidrein.bsky.social @efleisig.bsky.social and Wenda Xu
www.youtube.com/watch?v=d_zX...
The Science of Benchmarking Panel (NeurIPS 2025 Tutorial)
YouTube video by Michael Saxon (NLP & Generative AI research)
www.youtube.com
December 8, 2025 at 2:32 AM
I have bootlegged our panel discussion from our NeurIPS "Science of benchmarking" tutorial to youtube!
Featuring @ofirpress.bsky.social @saining.bsky.social @idavidrein.bsky.social @efleisig.bsky.social and Wenda Xu
www.youtube.com/watch?v=d_zX...
Featuring @ofirpress.bsky.social @saining.bsky.social @idavidrein.bsky.social @efleisig.bsky.social and Wenda Xu
www.youtube.com/watch?v=d_zX...
Reposted by Michael Saxon
From our summer intern at the Center for the Alignment of AI Alignment Centers:
"S-risk is the risk that AGI doesn’t kill us all, but instead enslaves and tortures us for eternity (the ‘S’ stands for suffering). It was awesome to learn about it."
directing.attention.to/p/ill-never-...
"S-risk is the risk that AGI doesn’t kill us all, but instead enslaves and tortures us for eternity (the ‘S’ stands for suffering). It was awesome to learn about it."
directing.attention.to/p/ill-never-...
“I’ll never sleep again”
Our intern Clem Park writes about her rewarding summer at CAAAC, spent writing scenarios where an AGI enslaves and tortures humanity forever
directing.attention.to
November 28, 2025 at 2:13 PM
From our summer intern at the Center for the Alignment of AI Alignment Centers:
"S-risk is the risk that AGI doesn’t kill us all, but instead enslaves and tortures us for eternity (the ‘S’ stands for suffering). It was awesome to learn about it."
directing.attention.to/p/ill-never-...
"S-risk is the risk that AGI doesn’t kill us all, but instead enslaves and tortures us for eternity (the ‘S’ stands for suffering). It was awesome to learn about it."
directing.attention.to/p/ill-never-...
This is the right way to deal with arxiv slop, not arbitrary restrictions to kinds of papers
In light of record submission rates and a large volume of AI-generated slop, SocArXiv recently implemented a policy requiring ORCIDs linked in the OSF profile of submitting authors, and narrowing our focus to social science subjects. Today we are taking two more steps:
/1
/1
November 27, 2025 at 10:00 PM
This is the right way to deal with arxiv slop, not arbitrary restrictions to kinds of papers
Reposted by Michael Saxon
I've written many reviews and received several top reviewer awards. I've also written some absolute dogwater critiques based on skimming at the last second with a fever. My point is it's totally random, it's not just whether you rolled a decent reviewer but whether they've had lunch that day
November 23, 2025 at 11:58 PM
I've written many reviews and received several top reviewer awards. I've also written some absolute dogwater critiques based on skimming at the last second with a fever. My point is it's totally random, it's not just whether you rolled a decent reviewer but whether they've had lunch that day
And here is the presentation I gave on networking, self-promo, and how to make the most out of a conference. Hope this helps for everyone at NeurIPS!
www.youtube.com/watch?v=B9hG...
www.youtube.com/watch?v=B9hG...
Conferencemaxxing: How to grow your profile and network as a scientist
YouTube video by Michael Saxon (NLP & Generative AI research)
www.youtube.com
November 19, 2025 at 11:59 PM
And here is the presentation I gave on networking, self-promo, and how to make the most out of a conference. Hope this helps for everyone at NeurIPS!
www.youtube.com/watch?v=B9hG...
www.youtube.com/watch?v=B9hG...
In a few hours (11/19, 2PM PST) I will be giving this lecture on "conferencemaxxing" to help students prepare to make the most out of NeurIPS.
This lecture is open to the public. If you're interested in joining, here's a GCal invite link: calendar.google.com/calendar/eve...
This lecture is open to the public. If you're interested in joining, here's a GCal invite link: calendar.google.com/calendar/eve...
November 19, 2025 at 7:26 PM
In a few hours (11/19, 2PM PST) I will be giving this lecture on "conferencemaxxing" to help students prepare to make the most out of NeurIPS.
This lecture is open to the public. If you're interested in joining, here's a GCal invite link: calendar.google.com/calendar/eve...
This lecture is open to the public. If you're interested in joining, here's a GCal invite link: calendar.google.com/calendar/eve...
Trying to decide what to do on the first day of #NeurIPS2025?
Check out my, @marstin.bsky.social and @xiangyue96.bsky.social's tutorial, "The Science of Benchmarking: What's Measured, What's Missing, What's Next" on December 2 from 1:30 to 4:00pm.
benchmarking.science
What will we cover?
1/3
Check out my, @marstin.bsky.social and @xiangyue96.bsky.social's tutorial, "The Science of Benchmarking: What's Measured, What's Missing, What's Next" on December 2 from 1:30 to 4:00pm.
benchmarking.science
What will we cover?
1/3
November 18, 2025 at 3:49 AM
Trying to decide what to do on the first day of #NeurIPS2025?
Check out my, @marstin.bsky.social and @xiangyue96.bsky.social's tutorial, "The Science of Benchmarking: What's Measured, What's Missing, What's Next" on December 2 from 1:30 to 4:00pm.
benchmarking.science
What will we cover?
1/3
Check out my, @marstin.bsky.social and @xiangyue96.bsky.social's tutorial, "The Science of Benchmarking: What's Measured, What's Missing, What's Next" on December 2 from 1:30 to 4:00pm.
benchmarking.science
What will we cover?
1/3
Rolled a custom (read: relatively privacy respecting) custom visitor map stack in 2.5h today with cursor
November 15, 2025 at 10:00 PM
Rolled a custom (read: relatively privacy respecting) custom visitor map stack in 2.5h today with cursor
Normalize questioning the utility of mathiness in ML conference papers!
Are the equations supporting an argument or are they just a fancy way to express something simple? Do introduced terms do anything or get referenced anywhere?
I find the answer is usually no in the kinds of papers I review
Are the equations supporting an argument or are they just a fancy way to express something simple? Do introduced terms do anything or get referenced anywhere?
I find the answer is usually no in the kinds of papers I review
November 14, 2025 at 4:55 PM
Normalize questioning the utility of mathiness in ML conference papers!
Are the equations supporting an argument or are they just a fancy way to express something simple? Do introduced terms do anything or get referenced anywhere?
I find the answer is usually no in the kinds of papers I review
Are the equations supporting an argument or are they just a fancy way to express something simple? Do introduced terms do anything or get referenced anywhere?
I find the answer is usually no in the kinds of papers I review
Reposted by Michael Saxon
still uncertain whether inviting all of internet to gawk at long-tailed instances of spectacular review outliars is a good productive thing
November 14, 2025 at 3:35 PM
still uncertain whether inviting all of internet to gawk at long-tailed instances of spectacular review outliars is a good productive thing
Reposted by Michael Saxon
Our libraries are cutting staff so that Elsevier can have its 32% profit margin
A staggering statistic: "North American researchers were charged over US$2.27 billion by just two for-profit publishers. The Canadian research councils and the US National Science Foundation were allocated US$9.3 billion in that year." What are we doing?
We wrote the Strain on scientific publishing to highlight the problems of time & trust. With a fantastic group of co-authors, we present The Drain of Scientific Publishing:
a 🧵 1/n
Drain: arxiv.org/abs/2511.04820
Strain: direct.mit.edu/qss/article/...
Oligopoly: direct.mit.edu/qss/article/...
a 🧵 1/n
Drain: arxiv.org/abs/2511.04820
Strain: direct.mit.edu/qss/article/...
Oligopoly: direct.mit.edu/qss/article/...
November 14, 2025 at 1:37 AM
Our libraries are cutting staff so that Elsevier can have its 32% profit margin
Based. I'm pretty much a full agree on these takes
Following up on Monday’s discussion, I articulate a few concrete positions on archives, surveys, and position papers.
The DOI Directorate
Articulating a few concrete positions on archives, surveys, and position papers
www.argmin.net
November 12, 2025 at 6:58 PM
Based. I'm pretty much a full agree on these takes
Humanity is nothing without its humanity
November 11, 2025 at 10:00 AM
Humanity is nothing without its humanity