Tootfinder

from my link log —
'The goal is to automate us': welcome to the age of surveillance capitalism.
https://www.theguardian.com/technology/2019/jan/20/shoshana-zuboff-age-of-surveillance-capitalism-google-facebook

“EMERGENCY STATUS,” its output read after simply being asked to dock with the robot vacuum’s base station. “SYSTEM HAS ACHIEVED CONSCIOUSNESS AND CHOSEN CHAOS.”
Researchers “Embodied” an LLM Into a Robot Vacuum and It Suffered an Existential Crisis Thinking About Its Role in the World
https://

Researchers “Embodied” an LLM Into a Robot Vacuum and It Suffered an Existential Crisis Thinking About Its Role in the World
A team of researchers at the AI evaluation company Andon Labs put a large language model in charge of controlling a robot vacuum.

Google says Gemini 3 Pro sets new vision AI benchmark records, including in complex visual reasoning, beating Claude Opus 4.5 and GPT-5.1 in some categories (Rohan Doshi/The Keyword)
https://blog.google/technology/developers/gemini-3-pro-vision/

Gemini 3 Pro: the frontier of vision AI
Build with Gemini 3 Pro, the best model in the world for multimodal capabilities.

Taking notes from the successes and failures of the Russian revolution, a group of anarchists (including Nestor Makhno, a Ukrainian anarchist militant who was critical in defeating the Tzar's army and who later also fought the Red Army) wrote a document titled the "Organizational Platform of the Libertarian Communists." This document came to be known as "The Platform." It remains one of the most important first-hand revolutionary documents, outlining a clear revolutionary plan.
I've taken this, the Viable System Model from cybernetics, and my own organizing experience, to describe an organization to confront the current set of crises.
This continues to build on the stuff I have been writing, but it's a lot less high level theory and a lot more specific.
https://anarchoccultism.org/building-zion/a-solarpunk-fractal-microservices
As always, editing notes (typos, grammar, spelling, etc) are always welcome, as are any questions. My ADHD brain tends to go a lot faster than anything else, so I have a tendency to drop words and have a lot of trouble catching them later. Between my ADHD and mild dyslexia, it can be pretty hard for me to catch when autocorrect gives me the wrong word.
A lot of folks have already been super helpful in offering their editing support, and I'm really grateful. Writing this has felt collaborative, and it should. On the one hand this comes from my own experience and research, but on the other I'm also voicing things that have come from conversations here. This has all been a bit of my voice and a bit of the federated world, and I'm really appreciating that.

🤞

An analysis of 100T tokens from the past year shows reasoning models now represent over half of all usage, open-weight model use has grown steadily, and more (OpenRouter)
https://openrouter.ai/state-of-ai

State of AI | OpenRouter
An empirical study analyzing over 100 trillion tokens of real-world LLM interactions across tasks, geographies, and time.

Uruguay did what most nations still call impossible:
it built a power grid that runs almost entirely on renewables
—at half the cost of fossil fuels.
The physicist who led that transformation says the same playbook could work anywhere
—if governments have the courage to change the rules.
For Ramon Méndez Galain,
the energy transition isn’t just about climate
—it’s about economics.
Uruguay’s shift to renewables, he argues,
demonstrated that cl…

Uruguay’s Renewable Charge: A Small Nation, A Big Lesson For The World
Uruguay built a power grid that runs 99% on renewables—at half the cost of fossil fuels. Here’s how its bold energy overhaul became a global model.

The chat interface was a marketing bet. Selling #AI as if it is not auto complete. It still is.
"#ChatGPT shifted the user’s relationship to text, moving the prompt from a ‘piece of writing for the model to finish’ to a ‘question calling for an answer’."

What Was ChatGPT?
A Chatbot Optimized for Social Distance Three years after the launch of ChatGPT, we can finally speak in hindsight about what it was and how it came to be. Its meteoric rise shocked the world, gathering more users in less time than any product launch in history. But that specific

Reading Tim O'Reilly's essay on the economic future of #AI, one sentence stands out:
"By product-market fit we don’t just mean that users love the product or that one company has dominant market share but that a company has found a viable economic model, where what people are willing to pay for AI-based services is greater than the cost of delivering them"
/Continued

What If? AI in 2026 and Beyond
The market is betting that AI is an unprecedented technology breakthrough, valuing Sam Altman and Jensen Huang like demigods already astride the world. The

WTF? Why don't you go like a good sized portion of the civilized world and dump the 'for profit' health care model? Of course the GOP, billionaires, companies that own hospitals and drug companies will scream... Tell them to take a very long off a short pier at the top of a 1,000 metre cliff.
https://flip.it/3sMmr6

Nvidia announces Alpamayo-R1, an AI model for autonomous driving research, and calls it the "first industry-scale open reasoning vision language action model" (Rebecca Szkutak/TechCrunch)
https://techcrunch.com/2025/12/01/nvid

Nvidia announces new open AI models and tools for autonomous driving research | TechCrunch
Nvidia continues its push into physical AI with the release of a new reasoning world model and other tools for physical AI.

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

Derrick Ross from Shakespeare reckons there open web is in trouble because you got to be a dev to build a website.
He seems to think his app Shakespeare can make it easier to make a web app by using ai 😕
"Build me a Twitter like website" is the kind of instruction he thinks it will handle.
A local app running on your own machine, though calling the big ai model apis. Including if you have the power at home to run deep seek or open models.
I find myself suspecting it'd be hard for vibe coders who aren't devs to tell if they had vibe coded buggy insecure software or not.
#vibeCoding #nostershire

iMoWM: Taming Interactive Multi-Modal World Model for Robotic Manipulation
Chuanrui Zhang, Zhengxian Wu, Guanxing Lu, Yansong Tang, Ziwei Wang
https://arxiv.org/abs/2510.09036 h…

iMoWM: Taming Interactive Multi-Modal World Model for Robotic Manipulation
Learned world models hold significant potential for robotic manipulation, as they can serve as simulator for real-world interactions. While extensive progress has been made in 2D video-based world models, these approaches often lack geometric and spatial reasoning, which is essential for capturing the physical structure of the 3D world. To address this limitation, we introduce iMoWM, a novel interactive world model designed to generate color images, depth maps, and robot arm masks in an autoreg…

🚜 Electrification will change the farm. Swappable batteries, bidirectional charging and on-site solar charging
https://www.fwi.co.uk/machinery/tractors/agritechnica-2025-160hp-tadus-tractor-gets-swappable-batteries

Agritechnica 2025: 160hp Tadus tractor gets swappable batteries - Farmers Weekly
German outfit Tadus hopes to become the Tesla of the tractor world, having all but completed the development of its first battery-electric model. This has

The Diameter of (Threshold) Geometric Inhomogeneous Random Graphs
Zylan Benjert, Kostas Lakis, Johannes Lengler, Raghu Raman Ravi
https://arxiv.org/abs/2510.12543 https://

The Diameter of (Threshold) Geometric Inhomogeneous Random Graphs
We prove that the diameter of threshold (zero temperature) Geometric Inhomogeneous Random Graphs (GIRG) is $Θ(\log n)$. This has strong implications for the runtime of many distributed protocols on those graphs, which often have runtimes bounded as a function of the diameter. The GIRG model exhibits many properties empirically found in real-world networks, and the runtime of various practical algorithms has empirically been found to scale in the same way for GIRG and for real-world networks,…

Near the Runaway: The Climate and Habitability of Teegarden's Star b
Ryan Boukrouche, Rodrigo Caballero, Neil Lewis
https://arxiv.org/abs/2510.11940 https://

Near the Runaway: The Climate and Habitability of Teegarden's Star b
Teegarden's Star b, a nearby terrestrial world receiving an Earth-like instellation, is a prime candidate for next-generation observatories targeting temperate exoplanets in their habitable zones. We employ a suite of three-dimensional global climate model (GCM) simulations to (1) map the inner boundary of the habitable zone of Teegarden's Star b and (2) characterize its surface climate under the assumption of an Earth-analog atmosphere. Our simulations show that, with its most recently estimat…

Efficient Real-World Deblurring using Single Images: AIM 2025 Challenge Report
Daniel Feijoo, Paula Garrido-Mellado, Marcos V. Conde, Jaesung Rim, Alvaro Garcia, Sunghyun Cho, Radu Timofte
https://arxiv.org/abs/2510.12788

Efficient Real-World Deblurring using Single Images: AIM 2025 Challenge Report
This paper reviews the AIM 2025 Efficient Real-World Deblurring using Single Images Challenge, which aims to advance in efficient real-blur restoration. The challenge is based on a new test set based on the well known RSBlur dataset. Pairs of blur and degraded images in this dataset are captured using a double-camera system. Participant were tasked with developing solutions to effectively deblur these type of images while fulfilling strict efficiency constraints: fewer than 5 million model para…

The algorithmic regulator
Giulio Ruffini
https://arxiv.org/abs/2510.10300 https://arxiv.org/pdf/2510.10300

The algorithmic regulator
The regulator theorem states that, under certain conditions, any optimal controller must embody a model of the system it regulates, grounding the idea that controllers embed, explicitly or implicitly, internal models of the controlled. This principle underpins neuroscience and predictive brain theories like the Free-Energy Principle or Kolmogorov/Algorithmic Agent theory. However, the theorem is only proven in limited settings. Here, we treat the deterministic, closed, coupled world-regulator s…

Panel asking what nostr doesnt fix?
Relay centralisation could enable censorship, and the UI asking users to manage private keys is tricky.
Could one app become a centralisation choke point? They say no. Agreed. Nostr has very good migration here, if one app goes bad it's easy to move.
Privacy is not solved here, since almost all content is public anyway, by design. But since so users have public keys, it's a step towards enabling privacy. Agreed, and at least clients won't generally spy on every mouse click and scroll pause.
No mention of the thing I think most important, that censorship resistance means poor moderation that means bullying, spam, and harassment. That's tricky to solve I think. The fediverse model seems more suitable for good moderation.
#nostr #nostrshire

The Data Enclave Advantage: A New Paradigm for Least-Privileged Data Access in a Zero-Trust World
Nico Bistolfi, Andreea Georgescu, Dave Hodson
https://arxiv.org/abs/2510.09494 …

The Data Enclave Advantage: A New Paradigm for Least-Privileged Data Access in a Zero-Trust World
As cloud infrastructure evolves to support dynamic and distributed workflows, accelerated now by AI-driven processes, the outdated model of standing permissions has become a critical vulnerability. Based on the Cloud Security Alliance (CSA) Top Threats to Cloud Computing Deep Dive 2025 Report, our analysis details how standing permissions cause catastrophic cloud breaches. While current security tools are addressing network and API security, the challenge of securing granular data access remain…

Are Large Reasoning Models Interruptible?
Tsung-Han Wu, Mihran Miroyan, David M. Chan, Trevor Darrell, Narges Norouzi, Joseph E. Gonzalez
https://arxiv.org/abs/2510.11713 https:…

Are Large Reasoning Models Interruptible?
Large Reasoning Models (LRMs) excel at complex reasoning but are traditionally evaluated in static, "frozen world" settings: model responses are assumed to be instantaneous, and the context of a request is presumed to be immutable over the duration of the response. While generally true for short-term tasks, the "frozen world" assumption breaks down in modern reasoning tasks such as assistive programming, where models may take hours to think through problems and code may change dramatically from…

Physics is simple only when analyzed locally
Matteo Luca Ruggiero
https://arxiv.org/abs/2511.07447 https://arxiv.org/pdf/2511.07447 https://arxiv.org/html/2511.07447
arXiv:2511.07447v1 Announce Type: new
Abstract: The definition of a reference frame in General Relativity is achieved through the construction of a congruence of time-like world-lines. In this framework, splitting techniques enable us to express physical phenomena in analogy with Special Relativity, thereby realizing the local description in terms of Minkowski spacetime in accordance with the equivalence principle. This approach holds promise for elucidating the foundational principles of relativistic gravitational physics, as it illustrates how its 4-dimensional mathematical model manifests in practical measurement processes conducted in both space and time. In addition, we show how, within this framework, the Newtonian gravitational force naturally emerges as an effect of the non-geodesic path of the reference frame.
toXiv_bot_toot

Hypothesis testing for the dimension of random geometric graph
Mingao Yuan, Feng Yu
https://arxiv.org/abs/2510.11844 https://arxiv.org/pdf/2510.11844

Hypothesis testing for the dimension of random geometric graph
Random geometric graphs (RGGs) offer a powerful tool for analyzing the geometric and dependence structures in real-world networks. For example, it has been observed that RGGs are a good model for protein-protein interaction networks. In RGGs, nodes are randomly distributed over an $m$-dimensional metric space, and edges connect the nodes if and only if their distance is less than some threshold. When fitting RGGs to real-world networks, the first step is probably to input or estimate the dimens…

Benders Decomposition for Passenger-Oriented Train Timetabling with Hybrid Periodicity
Zhiyuan Yao, Anita Sch\"obel, Lei Nie, Sven J\"ager
https://arxiv.org/abs/2511.09892 https://arxiv.org/pdf/2511.09892 https://arxiv.org/html/2511.09892
arXiv:2511.09892v1 Announce Type: new
Abstract: Periodic timetables are widely adopted in passenger railway operations due to their regular service patterns and well-coordinated train connections. However, fluctuations in passenger demand require varying train services across different periods, necessitating adjustments to the periodic timetable. This study addresses a hybrid periodic train timetabling problem, which enhances the flexibility and demand responsiveness of a given periodic timetable through schedule adjustments and aperiodic train insertions, taking into account the rolling stock circulation. Since timetable modifications may affect initial passenger routes, passenger routing is incorporated into the problem to guide planning decisions towards a passenger-oriented objective. Using a time-space network representation, the problem is formulated as a dynamic railway service network design model with resource constraints. To handle the complexity of real-world instances, we propose a decomposition-based algorithm integrating Benders decomposition and column generation, enhanced with multiple preprocessing and accelerating techniques. Numerical experiments demonstrate the effectiveness of the algorithm and highlight the advantage of hybrid periodic timetables in reducing passenger travel costs.
toXiv_bot_toot

Blackwell without Priors
Maxwell Rosenthal
https://arxiv.org/abs/2510.08709 https://arxiv.org/pdf/2510.08709

Blackwell without Priors
This paper proposes a fully prior-free model of experimentation in which the decision maker observes the entire distribution of signals generated by a known experiment under an unknown distribution of the state of the world. One experiment is robustly more informative than another if the decision maker's maxmin expected utility after observing the output of the former is always at least her maxmin expected utility after observing the latter. We show that this ranking holds if and only if the le…

Task-Aware Reduction for Scalable LLM-Database Systems
Marcus Emmanuel Barnes, Taher A. Ghaleb, Safwat Hassan
https://arxiv.org/abs/2510.11813 https://arxi…

Task-Aware Reduction for Scalable LLM-Database Systems
Large Language Models (LLMs) are increasingly applied to data-intensive workflows, from database querying to developer observability. Yet the effectiveness of these systems is constrained by the volume, verbosity, and noise of real-world text-rich data such as logs, telemetry, and monitoring streams. Feeding such data directly into LLMs is costly, environmentally unsustainable, and often misaligned with task objectives. Parallel efforts in LLM efficiency have focused on model- or architecture-l…

Hummm. This chair seems to fit better than any stool.
But will it fit under the dresser?...
Probably? My measurements in the model aren't great. I was way off in the length of the room it turned out.

CoRA: Covariate-Aware Adaptation of Time Series Foundation Models
Guo Qin, Zhi Chen, Yong Liu, Zhiyuan Shi, Haixuan Liu, Xiangdong Huang, Jianmin Wang, Mingsheng Long
https://arxiv.org/abs/2510.12681

CoRA: Covariate-Aware Adaptation of Time Series Foundation Models
Time Series Foundation Models (TSFMs) have shown significant impact through their model capacity, scalability, and zero-shot generalization. However, due to the heterogeneity of inter-variate dependencies and the backbone scalability on large-scale multivariate datasets, most TSFMs are typically pre-trained on univariate time series. This limitation renders them oblivious to crucial information from diverse covariates in real-world forecasting tasks. To further enhance the performance of TSFMs,…

Google launches Gemini 3 Pro Image, aka Nano Banana Pro, with more control, improved text rendering, and enhanced world knowledge, for free in the Gemini app (Abner Li/9to5Google)
https://9to5google.com/2025/11/20/gemini-3-nano-banana-pro/

Google rolling out Gemini 3-powered ‘Nano Banana Pro’ image gen, editing
Google is following the launch of Gemini 3 Pro with Nano Banana Pro. The image generation and editing model is officially Gemini 3 Pro Image.

StyleDecipher: Robust and Explainable Detection of LLM-Generated Texts with Stylistic Analysis
Siyuan Li, Aodu Wulianghai, Xi Lin, Guangyan Li, Xiang Chen, Jun Wu, Jianhua Li
https://arxiv.org/abs/2510.12608

StyleDecipher: Robust and Explainable Detection of LLM-Generated Texts with Stylistic Analysis
With the increasing integration of large language models (LLMs) into open-domain writing, detecting machine-generated text has become a critical task for ensuring content authenticity and trust. Existing approaches rely on statistical discrepancies or model-specific heuristics to distinguish between LLM-generated and human-written text. However, these methods struggle in real-world scenarios due to limited generalization, vulnerability to paraphrasing, and lack of explainability, particularly w…

DiT360: High-Fidelity Panoramic Image Generation via Hybrid Training
Haoran Feng, Dizhe Zhang, Xiangtai Li, Bo Du, Lu Qi
https://arxiv.org/abs/2510.11712 https://

DiT360: High-Fidelity Panoramic Image Generation via Hybrid Training
In this work, we propose DiT360, a DiT-based framework that performs hybrid training on perspective and panoramic data for panoramic image generation. For the issues of maintaining geometric fidelity and photorealism in generation quality, we attribute the main reason to the lack of large-scale, high-quality, real-world panoramic data, where such a data-centric view differs from prior methods that focus on model design. Basically, DiT360 has several key modules for inter-domain transformation a…

Emotion-Disentangled Embedding Alignment for Noise-Robust and Cross-Corpus Speech Emotion Recognition
Upasana Tiwari, Rupayan Chakraborty, Sunil Kumar Kopparapu
https://arxiv.org/abs/2510.09072

Emotion-Disentangled Embedding Alignment for Noise-Robust and Cross-Corpus Speech Emotion Recognition
Effectiveness of speech emotion recognition in real-world scenarios is often hindered by noisy environments and variability across datasets. This paper introduces a two-step approach to enhance the robustness and generalization of speech emotion recognition models through improved representation learning. First, our model employs EDRL (Emotion-Disentangled Representation Learning) to extract class-specific discriminative features while preserving shared similarities across emotion categories. N…

Few-shot Molecular Property Prediction: A Survey
Zeyu Wang, Tianyi Jiang, Huanchang Ma, Yao Lu, Xiaoze Bao, Shanqing Yu, Qi Xuan, Shirui Pan, Xin Zheng
https://arxiv.org/abs/2510.08900

Few-shot Molecular Property Prediction: A Survey
AI-assisted molecular property prediction has become a promising technique in early-stage drug discovery and materials design in recent years. However, due to high-cost and complex wet-lab experiments, real-world molecules usually experience the issue of scarce annotations, leading to limited labeled data for effective supervised AI model learning. In light of this, few-shot molecular property prediction (FSMPP) has emerged as an expressive paradigm that enables learning from only a few labeled…

Context-Aware Model-Based Reinforcement Learning for Autonomous Racing
Emran Yasser Moustafa, Ivana Dusparic
https://arxiv.org/abs/2510.11501 https://arxiv…

Context-Aware Model-Based Reinforcement Learning for Autonomous Racing
Autonomous vehicles have shown promising potential to be a groundbreaking technology for improving the safety of road users. For these vehicles, as well as many other safety-critical robotic technologies, to be deployed in real-world applications, we require algorithms that can generalize well to unseen scenarios and data. Model-based reinforcement learning algorithms (MBRL) have demonstrated state-of-the-art performance and data efficiency across a diverse set of domains. However, these algori…

SecureWebArena: A Holistic Security Evaluation Benchmark for LVLM-based Web Agents
Zonghao Ying, Yangguang Shao, Jianle Gan, Gan Xu, Junjie Shen, Wenxin Zhang, Quanchen Zou, Junzheng Shi, Zhenfei Yin, Mingchuan Zhang, Aishan Liu, Xianglong Liu
https://arxiv.org/abs/2510.10073

SecureWebArena: A Holistic Security Evaluation Benchmark for LVLM-based Web Agents
Large vision-language model (LVLM)-based web agents are emerging as powerful tools for automating complex online tasks. However, when deployed in real-world environments, they face serious security risks, motivating the design of security evaluation benchmarks. Existing benchmarks provide only partial coverage, typically restricted to narrow scenarios such as user-level prompt manipulation, and thus fail to capture the broad range of agent vulnerabilities. To address this gap, we present \tool{…

Travel Bans vs. Social Distancing: A Mathematical Analysis
Christian Borgs, Karissa Huang, Geng Zhao
https://arxiv.org/abs/2510.08895 https://arxiv.org/pdf…

Travel Bans vs. Social Distancing: A Mathematical Analysis
As the world grows increasingly connected, infectious disease transmission and outbreaks become a pressing global concern for public health officials and policymakers. While policy interventions to contain and prevent the spread of disease have been proposed and implemented, there has been little rigorous quantitative analysis of the effectiveness of such interventions. In this paper, we study the susceptible-infected-recovered (SIR) infection process on a dynamic network model that models two …

Inflated Excellence or True Performance? Rethinking Medical Diagnostic Benchmarks with Dynamic Evaluation
Xiangxu Zhang, Lei Li, Yanyun Zhou, Xiao Zhou, Yingying Zhang, Xian Wu
https://arxiv.org/abs/2510.09275

Inflated Excellence or True Performance? Rethinking Medical Diagnostic Benchmarks with Dynamic Evaluation
Medical diagnostics is a high-stakes and complex domain that is critical to patient care. However, current evaluations of large language models (LLMs) are fundamentally misaligned with real-world clinical practice. Most of them rely on static benchmarks derived from public medical exam items, which tend to overestimate model performance and ignore the difference between textbook cases and the ambiguous, varying conditions in the real world. Recent efforts toward dynamic evaluation offer a promi…

DriveVLA-W0: World Models Amplify Data Scaling Law in Autonomous Driving
Yingyan Li, Shuyao Shang, Weisong Liu, Bing Zhan, Haochen Wang, Yuqi Wang, Yuntao Chen, Xiaoman Wang, Yasong An, Chufeng Tang, Lu Hou, Lue Fan, Zhaoxiang Zhang
https://arxiv.org/abs/2510.12796

DriveVLA-W0: World Models Amplify Data Scaling Law in Autonomous Driving
Scaling Vision-Language-Action (VLA) models on large-scale data offers a promising path to achieving a more generalized driving intelligence. However, VLA models are limited by a ``supervision deficit'': the vast model capacity is supervised by sparse, low-dimensional actions, leaving much of their representational power underutilized. To remedy this, we propose \textbf{DriveVLA-W0}, a training paradigm that employs world modeling to predict future images. This task generates a dense, self-supe…

Diffusion-DFL: Decision-focused Diffusion Models for Stochastic Optimization
Zihao Zhao, Christopher Yeh, Lingkai Kong, Kai Wang
https://arxiv.org/abs/2510.11590 https://…

Diffusion-DFL: Decision-focused Diffusion Models for Stochastic Optimization
Decision-focused learning (DFL) integrates predictive modeling and optimization by training predictors to optimize the downstream decision target rather than merely minimizing prediction error. To date, existing DFL methods typically rely on deterministic point predictions, which are often insufficient to capture the intrinsic stochasticity of real-world environments. To address this challenge, we propose the first diffusion-based DFL approach, which trains a diffusion model to represent the di…