Tootfinder

How Large Language Models are Designed to Hallucinate
Richard Ackermann, Simeon Emanuilov
https://arxiv.org/abs/2509.16297 https://arxiv.org/pdf/2509.16297…

How Large Language Models are Designed to Hallucinate
Large language models (LLMs) achieve remarkable fluency across linguistic and reasoning tasks but remain systematically prone to hallucination. Prevailing accounts attribute hallucinations to data gaps, limited context, or optimization errors. We argue instead that hallucination is a structural outcome of the transformer architecture. As coherence engines, transformers are compelled to produce fluent continuations, with self-attention simulating the relational structure of meaning but lacking t…

Turns out that what hit flight #UA1093 was most likely a balloon from https://windbornesystems.com - this was brought up first in the thread https://x.com/vk5qi/status/1980386581497409891 and now the company confirms that it's a very strong possibility: https://x.com/johndeanl/status/1980462264974209292. See also https://www.youtube.com/watch?v=YZzbS30xdjM for a short video about the company and how it's filling the atmosphere with long-duration balloons carrying weather sensors. They had just been hailed as one of the best inventions of 2025: https://time.com/collections/best-inventions-2025/7318250/windborne-systems-atlas/ ...

REAL ENEMIES is an evening-length, multimedia, jazz-fueled exploration of American paranoia. This innovative marriage of music-theater and hybrid nonfiction marks the first collaboration between composer Darcy James Argue, filmmaker Peter Nigrini and writer Isaac Butler, who together chronicle a shadow history of post-war America that may—or may not—be true.
https://www.youtube.com/watch?v=GpXE-aZ-xhA

CAD model of the #tricycle subframe. The chain ring is very nearly as big as the front wheel (80 teeth!). It drives an epicyclic gearbox mounted in the nose of the vehicle, which also has a disc brake. The epicyclic will drive a secondary chain via a sprocket mounted to the left hand spoke flange, but this is not yet shown. There will be no ratchet on the secondary chain, so the disk brake on the…

By failing to sanction Israel, EU leaders are complicit in its crimes. They must act now. https://www.theguardian.com/commentisfree/2025/aug/01/israel-failure-eu-leaders-complicit-crime-genocide?CMP=Share_iO…

The Los Angeles field office director for the Department of Homeland Security
testified on Monday that Immigration and Customs Enforcement officers
desperately needed the help of military personnel in carrying out arrests.
The question is whether Donald Trump‘s deployment of armed forces goes against U.S. law that generally prohibits the president from using the military to police domestic affairs.
Ernesto Santacruz Jr. testified at the start of a three-day trial in S…

Trial to start over National Guard deployment to Los Angeles
A federal judge in San Francisco will hear arguments on whether military troops deployed this summer by the Trump administration to Los Angeles violated a federal law that bars troops from conducting law enforcement duties within the country. Judge Charles Breyer will decide whether the Posse Comitatus Act, an 1878 law that prevents the president from using the military as a domestic police force, was violated by the 4,700 California National Guard members and Marines sent following protests. T…

Factuality Beyond Coherence: Evaluating LLM Watermarking Methods for Medical Texts
Rochana Prih Hastuti, Rian Adam Rajagede, Mansour Al Ghanim, Mengxin Zheng, Qian Lou
https://arxiv.org/abs/2509.07755 …

Factuality Beyond Coherence: Evaluating LLM Watermarking Methods for Medical Texts
As large language models (LLMs) adapted to sensitive domains such as medicine, their fluency raises safety risks, particularly regarding provenance and accountability. Watermarking embeds detectable patterns to mitigate these risks, yet its reliability in medical contexts remains untested. Existing benchmarks focus on detection-quality tradeoffs, overlooking factual risks under low-entropy settings often exploited by watermarking's reweighting strategy. We propose a medical-focused evaluation w…

The ASPIICS solar coronagraph aboard the Proba-3 formation flying mission. Scientific objectives and instrument design
A. N. Zhukov, C. Thizy, D. Galano, B. Bourgoignie, L. Dolla, C. Jean, B. Nicula, S. Shestov, C. Galy, R. Rougeot, J. Versluys, J. Zender, P. Lamy, S. Fineschi, S. Gunar, B. Inhester, M. Mierla, P. Rudawy, K. Tsinganos, S. Koutchmy, R. Howard, H. Peter, S. Vives, L. Abbo, C. Aime, K. Aleksiejuk, J. Baran, U. Bak-Steslicka, A. Bemporad, D. Berghmans, D. Besliu-Ionescu, S…

The ASPIICS solar coronagraph aboard the Proba-3 formation flying mission. Scientific objectives and instrument design
We describe the scientific objectives and instrument design of the ASPIICS coronagraph launched aboard the Proba-3 mission of the European Space Agency (ESA) on 5 December 2024. Proba-3 consists of two spacecraft in a highly elliptical orbit around the Earth. One spacecraft carries the telescope, and the external occulter is mounted on the second spacecraft. The two spacecraft fly in a precise formation during 6 hours out of 19.63 hour orbit, together forming a giant solar coronagraph called AS…

Bidirectional Representations Augmented Autoregressive Biological Sequence Generation:Application in De Novo Peptide Sequencing
Xiang Zhang, Jiaqi Wei, Zijie Qiu, Sheng Xu, Zhi Jin, ZhiQiang Gao, Nanqing Dong, Siqi Sun
https://arxiv.org/abs/2510.08169

Bidirectional Representations Augmented Autoregressive Biological Sequence Generation:Application in De Novo Peptide Sequencing
Autoregressive (AR) models, common in sequence generation, are limited in many biological tasks such as de novo peptide sequencing and protein modeling by their unidirectional nature, failing to capture crucial global bidirectional token dependencies. Non-Autoregressive (NAR) models offer holistic, bidirectional representations but face challenges with generative coherence and scalability. To transcend this, we propose a hybrid framework enhancing AR generation by dynamically integrating rich c…

Just saw this:
#AI can mean a lot of things these days, but lots of the popular meanings imply a bevy of harms that I definitely wouldn't feel are worth a cute fish game. In fact, these harms are so acute that even "just" playing into the AI hype becomes its own kind of harm (it's similar to blockchain in that way).
@… noticed that the authors claim the code base is 80% AI generated, which is a red flag because people with sound moral compasses wouldn't be using AI to "help" write code in the first place. The authors aren't by some miracle people who couldn't build this app without help, in case that influences your thinking about it: they have the skills to write the code themselves, although it likely would have taken longer (but also been better).
I was more interested in the fish-classification AI, and how much it might be dependent on datacenters. Thankfully, a quick glance at the code confirms they're using ONNX and running a self-trained neural network on your device. While the exponentially-increasing energy & water demands of datacenters to support billion-parameter models are a real concern, this is not that. Even a non-AI game can burn a lot of cycles on someone's phone, and I don't think there's anything to complain about energy-wise if we're just using cycles on the end user's device as long as we're not having them keep it on for hours crunching numbers like blockchain stuff does. Running whatever stuff locally while the user is playing a game is a negligible environmental concern, unlike, say, calling out to ChatGPT where you're directly feeding datacenter demand. Since they claimed to have trained the network themselves, and since it's actually totally reasonable to make your own dataset for this and get good-enough-for-a-silly-game results with just a few hundred examples, I don't have any ethical objections to the data sourcing or training processes either. Hooray! This is finally an example of "ethical use of neutral networks" that I can hold up as an example of what people should be doing instead of the BS they are doing.
But wait... Remember what I said about feeding the AI hype being its own form of harm? Yeah, between using AI tools for coding and calling their classifier "AI" in a way that makes it seem like the same kind of thing as ChatGPT et al., they're leaning into the hype rather than helping restrain it. And that means they're causing harm. Big AI companies can point to them and say "look AI enables cute things you like" when AI didn't actually enable it. So I'm feeling meh about this cute game and won't be sharing it aside from this post. If you love the cute fish, you don't really have to feel bad for playing with it, but I'd feel bad for advertising it without a disclaimer.