https://techgazette.com.br/research/#journals This week's papers in Journals and labs, each with what its authors did and found, in sentences from its own abstract. en 180 Fri, 16 Oct 2026 17:29:00 GMT https://www.anthropic.com/research/designing-efficient-quantum-circuits https://www.anthropic.com/research/designing-efficient-quantum-circuits Fri, 16 Oct 2026 17:29:00 GMT Anthropic Research Journals and labs Anthropic Research — Did: Working largely on its own, Claude designed quantum circuits that improve on the best published versions, in some cases by an order of magnitude or more. https://dl.acm.org/doi/10.1145/3850128 https://dl.acm.org/doi/10.1145/3850128 Fri, 16 Oct 2026 12:00:00 GMT Communications of the ACM Journals and labs Communications of the ACM https://dl.acm.org/doi/10.1145/3849084 https://dl.acm.org/doi/10.1145/3849084 Fri, 16 Oct 2026 12:00:00 GMT Communications of the ACM Journals and labs Communications of the ACM — Did: Frontier labs say their AI models now help build their own successors. https://research.nvidia.com/publication/2026-12_imitation-learning-avoidance-modes-efficient-multimodal-model-predictive https://research.nvidia.com/publication/2026-12_imitation-learning-avoidance-modes-efficient-multimodal-model-predictive Fri, 16 Oct 2026 06:50:00 GMT NVIDIA Research Journals and labs NVIDIA Research https://research.nvidia.com/publication/2026-12_edgewise-bringing-long-horizon-agents-edge-selective-cloud-delegation https://research.nvidia.com/publication/2026-12_edgewise-bringing-long-horizon-agents-edge-selective-cloud-delegation Fri, 16 Oct 2026 06:45:00 GMT NVIDIA Research Journals and labs NVIDIA Research — Did: EdgeWise: Bringing Long-Horizon Agents to the Edge via Selective Cloud Delegation Complex and long-horizon agent tasks often rely on frontier models deployed in the cloud to achieve strong performance. https://www.nature.com/articles/d41586-026-03338-y https://www.nature.com/articles/d41586-026-03338-y Fri, 16 Oct 2026 00:00:00 GMT Nature: Computer Science Journals and labs Nature: Computer Science https://arxiv.org/abs/2610.22372 https://arxiv.org/abs/2610.22372 Thu, 15 Oct 2026 17:59:00 GMT arXiv cs.RO Journals and labs arXiv cs.RO — Did: We introduce LIBERO-PeRM, a benchmark for preference learning in embodied agents. Found: Mixed training achieves better no-preference retention than sequential adaptation and improves performance with contextual prompts compared with preference-only training, but can weaken motion-preference learning; preference generalization, decomposition, and conflict handling remain difficult. By Zhixu Li, Keqian Tang, Litian Gong, Jingyu Yao, Wenqian Zhang, Tianze Xu. https://arxiv.org/abs/2610.22367 https://arxiv.org/abs/2610.22367 Thu, 15 Oct 2026 17:59:00 GMT arXiv quant-ph Journals and labs arXiv quant-ph — Did: Here we introduce an evolutionary framework in which large language models (LLMs) propose circuit designs and an independent evaluator verifies and scores them. Found: By handling reversibility differently, the architecture makes divsteps -- a classical algorithm previously impractical for reversible integer arithmetic -- viable and substantially improves the Toffoli-qubit trade-off for modular inversion. By Andi Gu, Alek H. Kemeny, Susanne F. Yelin, Mikhail D. Lukin. https://arxiv.org/abs/2610.22364 https://arxiv.org/abs/2610.22364 Thu, 15 Oct 2026 17:59:00 GMT arXiv cs.RO Journals and labs arXiv cs.RO — Did: We introduce T2-Z for test-time optimization of environment latents, which augments a robotic world model with a learnable environment latent to condition world predictions. Found: Ablations further show that T2-Z can efficiently infer environment latents from as little as a single test-time interaction trajectory. By Chenyuan Zhou, Yining Hong, Huang Huang, Yunfan Jiang, Wenlong Huang, David Held. https://arxiv.org/abs/2610.22362 https://arxiv.org/abs/2610.22362 Thu, 15 Oct 2026 17:59:00 GMT arXiv cs.RO Journals and labs arXiv cs.RO — Did: We introduce CoDA, a framework in which a simulation data engine and an action policy co-evolve through a shared heterogeneous memory of environment construction, policy failures, and learning progress. Found: On LIBERO-Pro, 15 rounds of generation and training raise success from 25% to 84% under position perturbations and from 1% to 76% under instruction perturbations. By Tianjiao Yu, Zhilin Ou, Yifan Shen, Xinzhuo Li, Yuanzhe Liu, Onkar Kishor Susladkar. https://arxiv.org/abs/2610.22340 https://arxiv.org/abs/2610.22340 Thu, 15 Oct 2026 17:55:00 GMT arXiv cs.RO Journals and labs arXiv cs.RO — Did: We present DSP: Dexterity from supervised play, a framework in which a robot learns a play policy from minutes of human play through behavior cloning, then deploys it autonomously to collect hours of interaction data to train a world model. Found: These results demonstrate that supervised play can autonomously collect interaction data, guide model-based planning, and enable policy learning without any task-specific demonstrations or sim-to-real transfer. By Botao He, Yiming Dou, Yiannis Aloimonos, Jitendra Malik, Akash Sharma, Ruoshi Liu. https://arxiv.org/abs/2610.22296 https://arxiv.org/abs/2610.22296 Thu, 15 Oct 2026 17:45:00 GMT arXiv cs.SD Journals and labs arXiv cs.SD — Did: In this paper the concept of backward compatibility in zero-bit audio watermarking is formalized. By Admir Krnić, Kosta Pavlović, Petar Nedić, Željko Bolević, Slavko Kovačević, Igor Djurović. https://arxiv.org/abs/2610.22285 https://arxiv.org/abs/2610.22285 Thu, 15 Oct 2026 17:41:00 GMT arXiv cs.RO Journals and labs arXiv cs.RO — Did: We propose shifting this burden to the robot. Found: In simulation, the informed prior beats the standard active preference learning baseline by up to $30.8\%$ in the low-query regime, where reducing human effort matters most. By Helena Merker, Nick Walker, Andreea Bobu. https://arxiv.org/abs/2610.22278 https://arxiv.org/abs/2610.22278 Thu, 15 Oct 2026 17:39:00 GMT arXiv (systems & security) cs.DS Journals and labs arXiv (systems & security) cs.DS — Did: We study the problem of approximately counting subgraphs under local node differential privacy (LNDP), a recently introduced model that enforces strong privacy guarantees in analyses of distributed social networks. Found: Thus, our algorithm achieves optimal error for counting constant-size subgraphs, up to $\mathrm{poly}(1/\varepsilon)$ factors. By Sofya Raskhodnikova, Adam Smith, Connor Wagaman, Anatoly Zavyalov. https://arxiv.org/abs/2610.22258 https://arxiv.org/abs/2610.22258 Thu, 15 Oct 2026 17:27:00 GMT arXiv (vision, retrieval & HCI) eess.IV Journals and labs arXiv (vision, retrieval & HCI) eess.IV — Did: In this paper, we analyze the behavior of state-of-the-art (SOTA) denoisers on texture datasets and show that performance improvements are possible only by fine-tuning the denoisers on the texture datasets. By Saptarshi Mandal, Oindrila Haldar, Chirayata Bhattacharya, Chandra Sekhar Seelamantula, Zhiwei Xu, Thrasyvoulos N. Pappas. https://arxiv.org/abs/2610.22201 https://arxiv.org/abs/2610.22201 Thu, 15 Oct 2026 16:52:00 GMT arXiv cs.GT Journals and labs arXiv cs.GT — Did: We present a scalable framework for solving multi-step tasks through a decentralized market of self-interested agents. Found: Across a diverse set of domains, we demonstrate that imperfect competition successfully sustains necessary agent profit margins while maintaining high task success rates at low costs. By Ian Gemp, Shubhendu Trivedi, Adam Fisch, Fantine Huot, Michael Kaisers, Mirella Lapata. https://arxiv.org/abs/2610.22198 https://arxiv.org/abs/2610.22198 Thu, 15 Oct 2026 16:50:00 GMT arXiv (vision, retrieval & HCI) eess.SY Journals and labs arXiv (vision, retrieval & HCI) eess.SY — Did: In this work, we propose a unified control/communication strategy that provides safe control to guarantee inter-agent safety under intermittent communication and demonstrate its effectiveness in a pursuit-evasion scenario. Found: We implement and simulate a variety of pursuit-evasion scenarios and show that preemptive communication reduces messages by 77.5% compared to periodic and event-triggered baselines with zero collisions. By Madison Book, Justin Bradley. https://arxiv.org/abs/2610.22191 https://arxiv.org/abs/2610.22191 Thu, 15 Oct 2026 16:43:00 GMT arXiv cond-mat.dis-nn Journals and labs arXiv cond-mat.dis-nn — Did: In Hebbian learning from examples a Hopfield network of $N$ neurons is not supplied with the $K$ archetypes it has to retrieve but with $M$ blurred examples of each, of quality $r$, and whether it can infer them depends on the load $α=K/N$ (interference among them) and on the dataset entropy $ρ=(1-r^2)/(Mr^2)$ (how little the examples… By Francesco Alemanno. https://arxiv.org/abs/2610.22184 https://arxiv.org/abs/2610.22184 Thu, 15 Oct 2026 16:40:00 GMT arXiv cs.RO Journals and labs arXiv cs.RO — Did: We introduce Whole-Body Dexterity (WBD), a general framework that uses the robot's body and appendages as contact surfaces to grasp and manipulate objects. Found: Across real-world whole-body manipulation tasks, the distilled controller completes 90% of trials, compared with 18% for direct retargeting and 4% for a generalist whole-body controller. By Conor Mc Gartoll, An Dang, Pranay Thangeda, Arjun Gupta, Carolina Higuera, Nima Fazeli. https://arxiv.org/abs/2610.22178 https://arxiv.org/abs/2610.22178 Thu, 15 Oct 2026 16:34:00 GMT arXiv cs.RO Journals and labs arXiv cs.RO — Did: We introduce REKI, a 2.4B vision-language-action model (VLA) pretrained exclusively on public data that sets a new Pareto frontier between performance and model size: on LIBERO-Plus, RoboTwin Clean2Rand and RoboCasa365, it is outperformed only by models about twice its size. Found: This memory improves success in simulation and on real robots, and makes oscillations between poses about four times rarer. By Juliette Marrie, Paul Pacaud, Adrien Ramanana Rahary, Shizhe Chen, Moritz Böhle. https://dl.acm.org/doi/10.1145/3856804 https://dl.acm.org/doi/10.1145/3856804 Thu, 15 Oct 2026 12:00:00 GMT ACM Computing Surveys Journals and labs ACM Computing Surveys — Did: Endowing humanoid robots with expressive, controllable, and robust facial expressions and behaviors is essential for advancing human-like humanoid-human interaction across social, healthcare, and service domains. https://dl.acm.org/doi/10.1145/3793868 https://dl.acm.org/doi/10.1145/3793868 Thu, 15 Oct 2026 12:00:00 GMT Communications of the ACM Journals and labs Communications of the ACM — Did: Women remain underrepresented in computing programs globally, despite the growing relevance and reach of the field. https://arxiv.org/abs/2610.21540 https://arxiv.org/abs/2610.21540 Thu, 15 Oct 2026 09:06:00 GMT arXiv (systems & security) q-fin.TR Journals and labs arXiv (systems & security) q-fin.TR — Did: We introduce VDISC, a reproducible framework that records reported loss, exposed value, public evidence, returns, and pool flows separately. By Doohwi Cha, Hanhee Kim, Hyeonjun Yeo, Minsuk Sung, Minjae Lee, Seunghan Son. https://arxiv.org/abs/2610.21491 https://arxiv.org/abs/2610.21491 Thu, 15 Oct 2026 08:28:00 GMT arXiv (systems & security) q-fin.TR Journals and labs arXiv (systems & security) q-fin.TR — Did: We develop NEST, a stale-robust protocol that separates this shared-history fit from conditional measurement value. Found: A trade-level reconstruction on Gate gives 25.2%; NVDAX and AAPLX clear their frontiers, whereas TSLAX does not. By Doohwi Cha, Minsuk Sung, Minjae Lee, Hyeonjun Yeo, Hanhee Kim, Seunghan Son. https://arxiv.org/abs/2610.21488 https://arxiv.org/abs/2610.21488 Thu, 15 Oct 2026 08:25:00 GMT arXiv (systems & security) math.NT Journals and labs arXiv (systems & security) math.NT — Did: Let $E$ be chosen uniformly from the geometric isomorphism classes of supersingular elliptic curves over $\overline{\mathbb{F}}_p$ and $δ_E$ be the minimum degree of isogenies from $E$ to its Frobenius conjugate $E^{(p)}$. By Yi Ouyang, Wei Wei, Zheng Xu.