arXiv:2601.20005v1 Announce Type: new Abstract: The urgent need for building decarbonization calls for a paradigm shift in future autonomous building energy operation, from human-intensive engineering workflows toward intelligent agents that interact with physics-grounded digital environments. This study proposes an end-to-end agentic AI-enabled Physics-Informed Machine Learning (PIML) environment for scalable building energy modeling, simulation, control, and automation. The framework consists of (1) a modular and physics-consistent PIML digital environment spanning building thermal dynamics, Heating, Ventilation, and Air Conditioning (HVAC), […]

Ver mais

We just released a new paper benchmarking reasoning models (CoT as well as actual reasoning models) on Theory of Mind tests. These tests originally developed for human test persons, tests whether the person/models behaves as if it can understand mental states (intentions, emotions etc) (with our emphasis on as-if). Reasoning models perform well on these tasks, what does this say? That these tests are not always valid, that these models have improved ToM abilities compare to non-reasoning models, […]

Ver mais

We present Tape, a controlled reinforcement-learning benchmark designed to isolate out-of-distribution (OOD) failure under latent rule shifts.Tape is derived from one-dimensional cellular automata, enabling precise train/test splits where observation and action spaces are held fixed while transition rules change. Using a reproducible evaluation pipeline, we compare model-free baselines, model-based planning with learned world models, and task-inference (meta-RL) methods. A consistent pattern emerges: methods that are strong in-distribution (ID) can collapse under heldout-rule OOD, and high-variance OOD evaluation can […]

Ver mais

arXiv:2601.14991v1 Announce Type: cross Abstract: We study various types of consistency of honest decision trees and random forests in the regression setting. In contrast to related literature, our proofs are elementary and follow the classical arguments used for smoothing methods. Under mild regularity conditions on the regression function and data distribution, we establish weak and almost sure convergence of honest trees and honest forest averages to the true regression function, and moreover we obtain uniform convergence over compact […]

Ver mais

Recent advances in logistics optimization have enabled AI-driven solutions for demand forecasting, route planning, and supply chain management, yet most existing systems rely on fixed feature engineering and centralized decision-making, limiting adaptability, scalability, and effective human–AI collaboration in dynamic environments. To address these challenges, we propose Adaptive Multi-Agent Logistics Optimization with Human–AI Coordination Learning, a three-stage framework that integrates context-aware feature generation, hybrid multi-agent decision-making with intelligent caching, and adaptive coordination learning. The proposed system dynamically adjusts feature […]

Ver mais

arXiv:2601.01594v3 Announce Type: replace Abstract: We study variance reduction for score estimation and diffusion-based sampling in settings where the clean (target) score is available or can be approximated. Starting from the Target Score Identity (TSI), which expresses the noisy marginal score as a conditional expectation of the target score under the forward diffusion, we develop: (i) a plug-and-play nonparametric self-normalized importance sampling estimator compatible with standard reverse-time solvers, (ii) a variance-minimizing emph{state- and time-dependent} blending rule between Tweedie-type […]

Ver mais

If you’ve been listening to your favorite influencer, you’ve probably heard the gospel: “taste” will be the biggest differentiator in the AI era. LLMs can now write, draw, and compose with minimal prompting, but what they produce often feels lifeless. “Slop,” as the internet calls it. Taste, supposedly, is the antidote. The general sentiment is directionally correct. In a world where everyone has outsourced thinking and production to machines, those who take back the means of production and […]

Ver mais

If you are active on X (aka Twitter), you must be aware that Elon Musk keeps posting things from his personal X handle. He is the richest person on the planet, so what he thinks about the future, technology, and AI immediately becomes a headline. Besides courtroom rivalry with OpenAI and Microsoft, Musk’s favorite topic these days appears to be robotaxis. Or maybe he is talking about them because his automotive company, Tesla, plans to roll out a […]

Ver mais

arXiv:2601.21301v1 Announce Type: cross Abstract: We present the convergence rates of synchronous and asynchronous Q-learning for average-reward Markov decision processes, where the absence of contraction poses a fundamental challenge. Existing non-asymptotic results overcome this challenge by either imposing strong assumptions to enforce seminorm contraction or relying on discounted or episodic Markov decision processes as successive approximations, which either require unknown parameters or result in suboptimal sample complexity. In this work, under a reachability assumption, we establish optimal $widetilde{O}(varepsilon^{-2})$ […]

Ver mais

Contemporary artificial intelligence systems increasingly rely on recursive feedback processes, including self-training, preference optimization, and algorithmic governance loops. Across these settings, practitioners report a recurring failure mode in which system behavior contracts, dominant patterns emerge, and corrective interventions lose effectiveness. While such phenomena are often attributed to convergence or over-optimization, there remains no general empirical criterion for declaring when a system has entered an irrecoverable state. This paper introduces a fully empirical system for detecting entropy collapse in […]

Ver mais