Tootfinder

Can You Hear Me Now? A Benchmark for Long-Range Graph Propagation
Luca Miglior, Matteo Tolloso, Alessio Gravina, Davide Bacciu
https://arxiv.org/abs/2512.17762 https://arxiv.org/pdf/2512.17762 https://arxiv.org/html/2512.17762
arXiv:2512.17762v1 Announce Type: new
Abstract: Effectively capturing long-range interactions remains a fundamental yet unresolved challenge in graph neural network (GNN) research, critical for applications across diverse fields of science. To systematically address this, we introduce ECHO (Evaluating Communication over long HOps), a novel benchmark specifically designed to rigorously assess the capabilities of GNNs in handling very long-range graph propagation. ECHO includes three synthetic graph tasks, namely single-source shortest paths, node eccentricity, and graph diameter, each constructed over diverse and structurally challenging topologies intentionally designed to introduce significant information bottlenecks. ECHO also includes two real-world datasets, ECHO-Charge and ECHO-Energy, which define chemically grounded benchmarks for predicting atomic partial charges and molecular total energies, respectively, with reference computations obtained at the density functional theory (DFT) level. Both tasks inherently depend on capturing complex long-range molecular interactions. Our extensive benchmarking of popular GNN architectures reveals clear performance gaps, emphasizing the difficulty of true long-range propagation and highlighting design choices capable of overcoming inherent limitations. ECHO thereby sets a new standard for evaluating long-range information propagation, also providing a compelling example for its need in AI for science.
toXiv_bot_toot

as_skitter: Skitter IP graph (2005)
An aggregate snapshot of the Internet Protocol (IP) graph, as measured by the traceroute tool on CAIDA's skitter infrastructure, in 2005.
This network has 1696415 nodes and 11095298 edges.
Tags: Technological, Communication, Unweighted
https://networks.skewed.de/net/as_s…

Series D, Episode 06 - Headhunter
AVON: Where is Orac?
SOOLIN: Behind a panel in section four.
AVON: We have to get Orac outside.
TARRANT: Why risk it?
https://blake.torpidity.net/m/406/498 B7B5

Part of why #Trump has always been so hard to pin down politically is that he was always representing highly conflicting interests. Now, as that eats him alive, the GOP is fracturing in to two main groups: the Pinochet/Franco wing and the Hitler wing.
The Pinochet/Franco wing (let's call them PF) are lead by Vance. PF are also a coalition with some competing interests, but basically it's evangelical leaders, Opus Dei (fascist catholics), tech fascists (Yarvinites), pharma, and the other normal big republican donors. They support Israel, some because apartheid is extremely profitable and some because they support the genocide of Palestinian in order to bring the end of the world. They are split between extremely antisemitic evangelicals and Zionists, wanting similar things for completely different reasons. PF wants strong immigration enforcement because it lets them exploit immigrants, they don't want actual ethnic cleansing (just the constant threat). They want H1B visas because they want to a precarious tech work force. They want to end tariffs because they support free trade and don't actually care about things being made here.
The Hitler wing are lead by Nick Fuentes. I think they're a more unified group, but they're going to try to pull together a coalition that I don't think can really work. They're against Israel because they believe in some bat shit antisemitic conspiracy theory (which they are trying to inject along side legitimate criticism of Israel). They are focused on release of the #EpsteinFiles because they believe that it shows that Epstein worked for Mossad. They don't think that the ICE raids are going far enough, they oppose H1Bs because they are racists. They want a full ethnic cleansing of the US where everyone who isn't "white" is either enslaved for menial labor, deported, or dead. But they're also critical of big business (partially because of conspiracy theories but also) because they think their best option is to push for a white socialism (red/brown alliance).
Both of them want to sink Trump because they see him as standing in the way of their objectives. Both see #Epstein as an opportunity. Both of them have absolutely terrifying visions of authoritarian dictatorships, but they're different dictatorships.with opposing interests. Even within these there may be opportunities to fracture these more.
While these fractures decrease the likelihood of either group getting enough people together, their vision is more clear and thus more likely to succeed if they can make that happen. Now is absolutely *not* the time to just enjoy the collapse, we need to keep up or accelerate anti-fascist efforts to avoid repeating some of the mistakes of history.
Edit:
I should not that this isn't *totally* original analysis. I'll link a video later when I have time to find it.
Here it is:
#USPol

Spatially-informed transformers: Injecting geostatistical covariance biases into self-attention for spatio-temporal forecasting
Yuri Calleo
https://arxiv.org/abs/2512.17696 https://arxiv.org/pdf/2512.17696 https://arxiv.org/html/2512.17696
arXiv:2512.17696v1 Announce Type: new
Abstract: The modeling of high-dimensional spatio-temporal processes presents a fundamental dichotomy between the probabilistic rigor of classical geostatistics and the flexible, high-capacity representations of deep learning. While Gaussian processes offer theoretical consistency and exact uncertainty quantification, their prohibitive computational scaling renders them impractical for massive sensor networks. Conversely, modern transformer architectures excel at sequence modeling but inherently lack a geometric inductive bias, treating spatial sensors as permutation-invariant tokens without a native understanding of distance. In this work, we propose a spatially-informed transformer, a hybrid architecture that injects a geostatistical inductive bias directly into the self-attention mechanism via a learnable covariance kernel. By formally decomposing the attention structure into a stationary physical prior and a non-stationary data-driven residual, we impose a soft topological constraint that favors spatially proximal interactions while retaining the capacity to model complex dynamics. We demonstrate the phenomenon of ``Deep Variography'', where the network successfully recovers the true spatial decay parameters of the underlying process end-to-end via backpropagation. Extensive experiments on synthetic Gaussian random fields and real-world traffic benchmarks confirm that our method outperforms state-of-the-art graph neural networks. Furthermore, rigorous statistical validation confirms that the proposed method delivers not only superior predictive accuracy but also well-calibrated probabilistic forecasts, effectively bridging the gap between physics-aware modeling and data-driven learning.
toXiv_bot_toot

Interesting graph from @….
There was a •fascinating• 99pi about the push to create an official Category 6 — and the surprisingly compelling arguments against doing that (more fully developed in the story audio than in the web text summary):
https://99percentinvisible.org/episode/category-6/
One thing that’s a clear point of consensus in that debate is that the existing Saffir-Simpson hurricane scale isn’t up to the task of describing what we face now. https://fediscience.org/@rahmstorf/115481161250561175

Series D, Episode 01 - Rescue
[Interior Xenon base lounge. Soolin sets out glasses on a tray. She stops, regards the tray and then takes another glass from the shelf. She begins to pour wine into the glasses. Dorian leads the group into the lounge. Dorian grabs her and kisses her.]
SOOLIN: Are you all right?
https://blake.torpi…

#Shoutout geh raus an alle KI-begeisterten polnisch Sprechenden:
#Polnisch erzielt bei der Interaktion mit #KI die besten Ergebnisse unter 26 getesteten Sprachen.
Eine Analyse zeigte, dass groß…

Polish is the most effective language for prompting AI, study reveals
A group of researchers studied which language is best understood by artificial intelligence, with Polish in the lead and English only ranking at sixth place.

Series B, Episode 05 - Pressure Point
AVON: Frankly, I don't see how you can do it without me. Your strongest enemies are going to be the defence computers. I am the only one qualified to tackle them.
BLAKE: Yes, that had occurred to me. You better start getting kitted up. Do you want to tell me why?
https://blake.torpidit…