arXiv:2604.09705v1 Announce Type: new Abstract: AI infrastructure has transitioned from a software-centric paradigm to a system tightly bound by physical and environmental limits. Energy availability, cooling capacity, and network connectivity now impose hard operational boundaries that cannot be relaxed through software optimization alone. This paper proposes a sustainability-constrained orchestration framework that treats carbon intensity, water usage, and power capacity as strict feasibility constraints rather than tunable penalties, and that jointly optimizes compute placement and optical network routing in […]

Ver mais

arXiv:2602.17675v1 Announce Type: new Abstract: Enterprise conversational UIs increasingly need to orchestrate heterogeneous backend agents and tools across project and account boundaries in a secure and reproducible way. Starting from Gemini Enterprise Agent-to-Agent (A2A) invocation, we implement an A2A Hub orchestrator on Cloud Run that routes queries to four paths: a public A2A agent deployed in a different project, an IAM-protected Cloud Run A2A agent in a different account, a retrieval-augmented generation path combining Discovery Engine and Vertex […]

Ver mais

arXiv:2601.09896v1 Announce Type: new Abstract: Visual generative AI models are trained using a one-size-fits-all measure of aesthetic appeal. However, what is deemed “aesthetic” is inextricably linked to personal taste and cultural values, raising the question of whose taste is represented in visual generative AI models. In this work, we study an aesthetic evaluation model–LAION Aesthetic Predictor (LAP)–that is widely used to curate datasets to train visual generative image models, like Stable Diffusion, and evaluate the quality of AI-generated […]

Ver mais

arXiv:2602.09236v1 Announce Type: new Abstract: Version control is critical in mechanical computer-aided design (CAD) to enable traceability, manage product variation, and support collaboration. Yet, its implementation in modern CAD software as an essential information infrastructure for product development remains plagued by issues due to the complexity and interdependence of design data. This paper presents a systematic review of user-reported challenges with version control in modern CAD tools. Analyzing 170 online forum threads, we identify recurring socio-technical issues that […]

Ver mais

Pre-training large language models is expensive enough that even modest efficiency improvements can translate into meaningful cost and time savings. Nous Research is releasing Token Superposition Training (TST), a method that substantially reduces pre-training wall-clock time at fixed compute without touching the model architecture, optimizer, tokenizer, parallelism strategy, or training data. At the 10B-A1B mixture-of-experts scale, TST reaches a lower final training loss than a matched-FLOPs baseline while consuming 4,768 B200-GPU-hours versus the baseline’s 12,311 — roughly a […]

Ver mais

arXiv:2508.07473v3 Announce Type: replace-cross Abstract: In Online Convex Optimization (OCO), when the stochastic gradient has a finite variance, many algorithms provably work and guarantee a sublinear regret. However, limited results are known if the gradient estimate has a heavy tail, i.e., the stochastic gradient only admits a finite $mathsf{p}$-th central moment for some $mathsf{p}inleft(1,2right]$. Motivated by it, this work examines different old algorithms for OCO (e.g., Online Gradient Descent) in the more challenging heavy-tailed setting. Under the standard […]

Ver mais

arXiv:2602.13224v1 Announce Type: new Abstract: The term “hallucination” in large language models conflates distinct phenomena with different geometric signatures in embedding space. We propose a taxonomy identifying three types: unfaithfulness (failure to engage with provided context), confabulation (invention of semantically foreign content), and factual error (incorrect claims within correct conceptual frames). We observe a striking asymmetry. On standard benchmarks where hallucinations are LLM-generated, detection is domain-local: AUROC 0.76-0.99 within domains, but 0.50 (chance level) across domains. Discriminative directions […]

Ver mais