arXiv:2603.00582v1 Announce Type: new Abstract: While Large Language Models (LLMs) have demonstrated proficiency in Deep Research or Wide Search, their capacity to solve highly complex questions-those requiring long-horizon planning, massive evidence gathering, and synthesis across heterogeneous sources-remains largely unexplored. We introduce Super Research, a task for complex autonomous research tasks that integrates (i) structured decomposition into a research plan, (ii) super wide retrieval for diverse perspectives, and (iii) super deep investigation to resolve uncertainties through iterative queries. To […]

Ver mais

I’ve been diving into the deep end of Reinforcement Learning and Generative Models lately, specifically trying to see if I could train a simple diffusion model from scratch using nothing but a reward signal. On paper, it sounded like a fun weekend experiment, but in reality, it was a 72-hour masterclass in frustration. By Sunday night, I was staring at a screen of pure static; every time I adjusted the hyperparameters, the model would either collapse into a […]

Ver mais

Hello everyone, I’m working on a research problem for which I need single agent ppo/sac Baselines to compare against. From my own research I could only find implementations on multi agents or safe RL envs. Also the metadrive’s own implementation is just importing already existing weights and not training which just has ppo. Is there any implementation Baselines for me to compare against, maybe from some paper which I can refer to. Any help would be appreciated! Thanks. […]

Ver mais

Hey, I’ve been going down the rabbit hole of sim-to-real RL and I’m trying to understand where the ACTUAL bottlenecks are for people doing this in practice (not just in papers). From what I’ve read, domain randomization and system identification help close the gap, but it seems like there’s still a lot of pain around rare/adversarial scenarios that you can’t really plan for in sim. For those of you actually deploying RL policies on physical robots: What part […]

Ver mais

Unlike banks, crypto networks run without a central boss and without access requirements. Anyone with a device and Internet can create their own wallet in minutes. They can also participate as “block producers” (miner, “validator,” or equivalent) without sharing their identity. Crypto is pseudonymous and open… and that, without proper protection, could cause bad things like a Sybil attack. The term and the mechanism existed well before cryptocurrencies, since it applies to reputation systems and peer-to-peer (P2P) networks. […]

Ver mais

Multi-agent reinforcement learning is a promising research area that extends established reinforcement learning approaches to problems formulated as multi-agent systems. Recently, a multitude of communication methods have been introduced to this field to address problems such as partially observable environments, non-stationarity, and exponentially growing action spaces. Communication further enables efficient cooperation among all agents interacting in an environment. This work aims at providing an overview of communication techniques in multi-agent reinforcement learning. By an in-depth analysis of 29 […]

Ver mais

arXiv:2507.09093v2 Announce Type: replace Abstract: We study convergence in high-probability of SGD-type methods in non-convex optimization and the presence of heavy-tailed noise. To combat the heavy-tailed noise, a general black-box nonlinear framework is considered, subsuming nonlinearities like sign, clipping, normalization and their smooth counterparts. Our first result shows that nonlinear SGD (N-SGD) achieves the rate $widetilde{mathcal{O}}(t^{-1/2})$, for any noise with unbounded moments and a symmetric probability density function (PDF). Crucially, N-SGD has exponentially decaying tails, matching the performance […]

Ver mais

arXiv:2603.20521v1 Announce Type: cross Abstract: Distributed reinforcement learning trains on data from stale, buggy, or mismatched actors, producing actions with high surprisal (negative log-probability) under the learner’s policy. The core difficulty is not surprising data per se, but emph{negative learning from surprising data}. High-surprisal failures can dominate the update direction despite carrying little useful signal, while high-surprisal successes reveal opportunities the current policy would otherwise miss. The textit{Delightful Policy Gradient} (DG) separates these cases by gating each update […]

Ver mais

arXiv:2602.01928v2 Announce Type: replace Abstract: Privacy preservation is a fundamental requirement in many high-stakes domains such as medicine and finance, where sensitive personal data must be analyzed without compromising individual confidentiality. At the same time, these applications often involve datasets with missing values due to non-response, data corruption, or deliberate anonymization. Missing data is traditionally viewed as a limitation because it reduces the information available to analysts and can degrade model performance. In this work, we take an […]

Ver mais