iiinova — where your digital self keeps evolving

Reposted by Alexander Hoyle

Complexity Science Hub @csh.ac.at · 07/10/2026

⏳ 𝗟𝗮𝘀𝘁 𝗰𝗵𝗮𝗻𝗰𝗲! Want to join us for your #PhD and dive into #Complexity Science? Applications for the 𝗚𝗿𝗮𝗱𝘂𝗮𝘁𝗲 𝗣𝗿𝗼𝗴𝗿𝗮𝗺 🎓 𝗶𝗻 𝗖𝗼𝗺𝗽𝗹𝗲𝘅𝗶𝘁𝘆 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 at the Complexity Science Hub are open until 📅 𝟭𝟱 𝗢𝗰𝘁𝗼𝗯𝗲𝗿 𝟮𝟬𝟮𝟲. #AcademicSky 👉 More info: csh.ac.at/education/gr...

033

Reposted by Alexander Hoyle

Vilém Zouhar @zouhar.bsky.social · 29/09/2026

I'm on the faculty job market for Fall 2027 assistant professorship. I work on the science of AI/NLP evaluation, which is currently undergoing a crisis. Get in touch! My research statement is public: vilda.net

1144

Alexander Hoyle @alexanderhoyle.bsky.social · 25/09/2026

gonna have to seek in-patient treatment for fomo

020

Alexander Hoyle @alexanderhoyle.bsky.social · 30/08/2026

My personal website failed but an institutional page seems to have worked

120

Reposted by Alexander Hoyle

Jordan Boyd-Graber @boydgraber.bsky.social · 28/08/2026

I'm moving to Nanyang Technological University in Singapore to start a new lab! We'll still be dedicated to adversarial QA, human-computer collaboration, probabilistic modeling, and the other fun hijinx my students and I have been up to, but now (hopefully) bigger and better.

Jordan and his wife in front of Singapore Strait

Jordan and his daughters (faces blured) in front of Nanyang gate in Yunan garden.

3283

Alexander Hoyle @alexanderhoyle.bsky.social · 28/08/2026

Congratulations!! Please eat all the food I cannot

000

Alexander Hoyle @alexanderhoyle.bsky.social · 25/08/2026

Hand-written Python! Are you planning to do any OCRing or grade it all by hand?

110

Alexander Hoyle @alexanderhoyle.bsky.social · 25/08/2026

Would you mind adding me? I’m a researcher in the field (and appreciate your posts!)

110

Alexander Hoyle @alexanderhoyle.bsky.social · 24/08/2026

I’ve been thinking vaguely about some kind of public token system—you get say N a year to assign to other papers you like Could also have an author version, where you mark the papers you really vouch for (like how people list “selected work” on their homepage)

000

Alexander Hoyle @alexanderhoyle.bsky.social · 23/08/2026

perhaps venues/arxiv could host automated high-precision certifiers/desk rejectors (for ML: are citations real? are there baselines? does it reproduce?) that gate submissions and can only be run once per cycle & paper. Include a simple appeal process (with a cycle-long ban if abused)

000

Alexander Hoyle @alexanderhoyle.bsky.social · 21/08/2026

Ah, I would have maybe compared only the main track to other conferences (in the past, findings+main was even higher, ~35-37%) Per this analysis, kinit.sk/quality-of-a..., "Findings" is comparable to CORE A and main *ACL conferences are CORE A* (not that I've ever really looked into these rankings)

112

Alexander Hoyle @alexanderhoyle.bsky.social · 21/08/2026

To elaborate, "Findings" was introduced to help handle the large number of submissions to NLP venues; they are scientifically sound papers deemed worthy of publication but that did not make it to the main conference (e.g., they may be percieved as more specialized).

111

Alexander Hoyle @alexanderhoyle.bsky.social · 21/08/2026

! EMNLP has a 15% acceptance rate this year, down from the usual 20-22% Seems like an appropriate response to the wave of slop Of course, the real problem is contending with the submissions, and I still feel we‘re being too conservative in our thinking there

This year, we received an unprecedented 17669 submissions. We accepted 2719 as Main Conference papers and 2533 to Findings of the ACL. This represents an acceptance rate of 15.4% for Main Conference papers and 14.3% for Findings papers.

1234

Reposted by Alexander Hoyle

Lucy Li @lucy3.bsky.social · 07/08/2026

Spent a good amount of this summer digging through a 20+ person team of teachers' free-text annotations, learning what "scaffolding" and "push for rigor" means, and iterating on this pipeline. We find that AI tutors, by default, frequently over-scaffold and rarely push for rigor.

27412

Alexander Hoyle @alexanderhoyle.bsky.social · 29/07/2026

A (very small) silver lining of writing a metareview with dozens of spammed LLM rebuttals is that occasionally the authors' unproofed generated response gives the game away by saying something like "Your criticism is spot on—it points to a serious oversight that requires a major revision."

3221

Alexander Hoyle @alexanderhoyle.bsky.social · 29/07/2026

Claude loves to invent its own project-specific jargon/shorthand while still being too verbose. It's information dense in the wrong places. Play-by-play project updates or esoteric technical conversations are a also genre that has no real human precedent, so it must be largely out-of-distribution

030

Alexander Hoyle @alexanderhoyle.bsky.social · 28/07/2026

ok but what are the other cities on this list. I'm now dying to know

000

Alexander Hoyle @alexanderhoyle.bsky.social · 28/07/2026

Whereas, in my understanding, a political science paper presented at APSA is a work-in-progress, and the conference's purpose is to provide feedback that will strengthen a future submission that will go to a journal (and the APSA presentation itself is not meaningful to a hiring/tenure committee)

220

Alexander Hoyle @alexanderhoyle.bsky.social · 28/07/2026

A place where the canonical/final version of the work is published (and that has value as a line-item in a CV for the bean counters) In comp sci, conference publications have been (historically) treated as ~equivalent to journal articles, so once work is accepted there it is considered complete

110

Alexander Hoyle @alexanderhoyle.bsky.social · 28/07/2026

DH people: a student is working on an interdisciplinary art history / ML project. With #DH2026 going on, do you view it as a “terminal” venue, like in CS, or is it more like IC2S2 where it’s a stepping stone to other venues? If so, what are they?

242

Alexander Hoyle @alexanderhoyle.bsky.social · 20/07/2026

are the closed embedding models that much better? or are you saying that there’s nothing end-to-end open like an OLMo

120

Alexander Hoyle @alexanderhoyle.bsky.social · 14/07/2026

yea, if we assume there have been several bad cycles, then being a recently-qualified “insider“ may not be a meaningful designation?

030

Alexander Hoyle @alexanderhoyle.bsky.social · 08/07/2026

right?? it's spectacular. I love mountains/elevation in a city. There's also good architectural variation (in parts), lots of small streets and alleyways. A gem

020

Alexander Hoyle @alexanderhoyle.bsky.social · 07/07/2026

I've talked to several people on hiring committees (and heard this second-hand about industry roles) that they look unfavorably on too many publications, as it likely indicates a lack of rigor/personal investment. Now that I'm hiring my own PhD students, I find myself doing the same thing

030

Alexander Hoyle @alexanderhoyle.bsky.social · 07/07/2026

Of course we also need to reduce submissions. Part of that is actually impressing upon students that low-effort papers on their CV *will hurt them* (advice I used to believe was out-of-touch/cope, but have come to learn is true) On the conference side, we need better filtering mechanisms

120

Alexander Hoyle @alexanderhoyle.bsky.social · 07/07/2026

Under the current policy of fixed acceptance rates, the signal of a published paper is going to go to ~zero (negative for slop) I think we need to take a longer view: when the cost of producing a paper is so low, what do we want a publication to mean, and how do we encourage that meaning?

380

Alexander Hoyle @alexanderhoyle.bsky.social · 07/07/2026

Per some back-of-the-envelope math, I think this doesn't pan out. Even if we capped submissions at max one per author (which no one is suggesting), we'd still have at least ~10k, or the total number of submissions for ACL and EMNLP in 2024

120

Alexander Hoyle @alexanderhoyle.bsky.social · 07/07/2026

desk rejecting low effort “interdisciplinary“ work may be a start (although I guess @davidjurgens.bsky.social said this is already happening) bsky.app/profile/yoav...

120

Alexander Hoyle @alexanderhoyle.bsky.social · 04/07/2026

Your talk at ETH was great! I've referenced this paper a few times

110

Reposted by Alexander Hoyle

daniel holmgren 🫠 @dholms.at · 30/06/2026

one benefit of the rise of ai writing is that i feel more personal liberty to let my voice come through in my writing. colloquialisms, asides, jokes, mixing of registers, etc

413316

Alexander Hoyle @alexanderhoyle.bsky.social · 30/06/2026

I increasingly see this type of work while reviewing. If we continue to let it slip in, then *ACL will become a dumping ground for half-baked LLM industry applications with little scientific merit. Even when it’s very good, it sends a signal about the conference that maybe we don’t want

080

Alexander Hoyle @alexanderhoyle.bsky.social · 25/06/2026

thank you for fighting the good fight. I have tried making the argument that generative models far predate LLMs and the term generative AI is over-inclusive but was insufficiently convincing 🫡

020

Alexander Hoyle @alexanderhoyle.bsky.social · 24/06/2026

Isn't that mistake pre-LLM code generation? I guess to me that's evidence that it's the same as it ever was---there could always be some bug in the big stack of abstractions

100

Alexander Hoyle @alexanderhoyle.bsky.social · 23/06/2026

Fair points, but what is really lost when generating basic matplotlib or a one-off annotation UI? They aren’t what I consider my core competencies, nor even skills I want my students to learn, and the output is faster to check than produce by at least 10:1

110

Alexander Hoyle @alexanderhoyle.bsky.social · 22/06/2026

Now that's a provocative idea! I think this already happens with some workshops and regional/smaller conferences: the signal of a paper there is minimal, so there's less incentive to submit slop An issue with your suggestion is that high-quality work may go to other venues in search of recognition

100

Alexander Hoyle @alexanderhoyle.bsky.social · 16/06/2026

I am biased, but I think the Apertus team at ETH/EPFL are doing what you describe: an ongoing effort with planned regular releases and accumulating expertise. You should visit!

010

Alexander Hoyle @alexanderhoyle.bsky.social · 01/06/2026

The one problem is it's way too hot and the pool can end up like bathwater. Still fun though

000

Alexander Hoyle @alexanderhoyle.bsky.social · 01/06/2026

The one summer I stayed at UMD, I was writing my dissertation and everyone in my lab was away on internships. For my lunch breaks, I'd bike to the nearly empty gym, work out, swim at the outdoor pool, and end with a big meal at the cafeteria. It was like a summer camp for one

110

Alexander Hoyle @alexanderhoyle.bsky.social · 27/05/2026

Yeah--collecting demographic data also would have triggered a ”human subjects research“ designation by the UMD IRB, whereas anonymous annotations didn’t, and I assume the same is true of other institutions

100

Alexander Hoyle @alexanderhoyle.bsky.social · 27/05/2026

Thanks---and sorry to have missed you!

010

Alexander Hoyle @alexanderhoyle.bsky.social · 27/05/2026

If you were considering schools in the US but are having visa issues/concerns about relocating there, please consider applying—of course Americans looking for outside opportunities are also welcome!

020

Alexander Hoyle @alexanderhoyle.bsky.social · 27/05/2026

Why join? Vienna has a vibrant, rapidly-growing AI scene and a terrific quality of life; TU Wien has an excellent CS department; the Complexity Science Hub is a home to ambitious interdisciplinary work. If you are curious, opinionated, driven: come!

160

Alexander Hoyle @alexanderhoyle.bsky.social · 27/05/2026

There is also the doctoral program offered by the Complexity Science Hub opening in August with a December deadline. You can name me as an advisor. csh.ac.at/education/gr...

140