与「ReinforcementLearning」相关的搜索结果

The future of humanoids isn't just mechanical—it's neural. A humanoid robot is an embodied AI system. Motors, sensors, and actuators provide the body, but neural intelligence provides the mind. Neural models power vision, speech, language, navigation, manipulation, planning, memory, decision-making, and continuous learning. As robotics evolves, nearly every cognitive subsystem is becoming neural-first. That's why represents more than a niche—it's a foundational concept for the next generation of intelligent robotics. #HumanoidRobots# #EmbodiedAI# #PhysicalAI# #NeuralNetworks# #ArtificialIntelligence# #Robotics# #MachineLearning# #DeepLearning# #RobotLearning# #ReinforcementLearning# #ComputerVision# #GenerativeAI# #AIResearch# #Automation# #FutureOfAI#

🧵 Deli AutoResearch SKILL is now officially open source! 🎉 Alongside it, we’re dropping our 4th survey paper — this time on Self-play. Inspired by AlphaZero, we got a powerful insight: prior knowledge doesn’t always lift the ceiling. Models can discover more globally optimal solutions just by playing against themselves. The biggest change in this paper? For the first time, the AutoResearch Agent autonomously planned GPU experiments — and submitted actual RL runs on the DeepSeek 285B model. The entire RL pipeline — experiment design, code writing, running, debugging, and conclusion summarization — was 100% automated, with zero human intervention from me. This was incredibly difficult, but an incredibly important step. GRPO is the tool being called by the AutoResearch Agent here. We see this as the beginning of our Continual Learning research journey. 🚀 As always, this is my personal research project, unaffiliated with any organization. All views are my own. #AI# #ReinforcementLearning# #SelfPlay# #OpenSource# #AutoML# #ContinualLearning# #DeepSeek#

A humanoid robot is fundamentally an embodied AI system. While motors, actuators, batteries, and sensors provide the body, neural intelligence provides the mind. Without neural computation, a humanoid is simply an advanced machine. With it, the robot becomes capable of adapting to unpredictable environments, understanding language, recognizing objects, planning tasks, and continuously improving through experience. This shift positions "Humanoid Neural" as a foundational concept within the robotics ecosystem. Neural Intelligence is the Core of Future Humanoids. Modern humanoids rely on neural-network-based systems to perform nearly every cognitive function. These include: Visual perception Speech recognition Language understanding Object identification Human pose estimation Motion planning Reinforcement learning Dexterous manipulation Long-term memory Decision making Navigation Emotional recognition Social interaction As robots become more capable, nearly every subsystem transitions from traditional programming toward learned neural models. This makes "neural" less of a niche AI term and more of an umbrella for robot intelligence. A Broad and Scalable Brand One of the strongest characteristics of is that it is not confined to a single product category. #Languageunderstanding# #neuralnetwork# #socialinteraction# #neuralntelligence# #smarthumanoids# #iq#

Blockchain Sports Drift — one year later One year ago, an AI drift car was just an idea. Today, it's real — and it drives itself. Here's what we've built so far. The car drives itself Our prototype drives completely on its own. No driver. No remote control. The AI reads the car's position, speed, and angle 30 times every second — and acts on it. It doesn't just react. It acts. The AI keeps learning Instead of following the same instructions every time, the system learns through reinforcement learning. It practices thousands of times, learns from every run, and keeps getting better. Always knows where it is The car combines information from several sensors into one accurate picture of where it is on the track. Even while sliding sideways, it always knows its position. Safety comes first The car has six independent hardware safety systems. They don't rely on software or a radio connection. If something goes wrong, the car stops automatically. We've already demonstrated this during live events with no safety driver inside. Building a unique dataset We're collecting biometric data from professional drivers using our own specialized racing suit. Combined with everything the car records while driving, we're building a unique dataset for autonomous motorsport. Visitors from around the world Over the past year, dozens of guests from the United States, Japan, and across Europe visited our facility. They saw the technology, tested the car, and experienced our professional drift cars firsthand. The reaction was the same every time — disbelief, then excitement. What’s next We’re building the foundation for a new kind of motorsport. First, we’ll prove what our own AI can do. Then we’ll open our core technology to the world’s leading tech companies — so they can build their own AI-powered drift cars and compete. The destination? A world championship in the United States — where AI cars backed by the world’s biggest tech companies compete against the best human drift drivers. All powered by our core technology. This is only year one!

After Xiao Wu, meet Xiao Liu. 🤖 Tencent Robotics X’s latest robot isn’t just learning massage moves. It’s learning the touch behind them — where to press and exactly how hard. A custom data system captures vision, touch, and force from human demonstrations. Using reinforcement learning, Xiao Liu learns to reproduce both the trajectory and the pressure. Getting to the right spot is one thing. Getting the pressure right is another. My shoulders volunteer as tribute.

The smartest models don’t necessarily guarantee the best results. Vladislav Tankov, our CTO, explains how with JetBrains AI for Teams and Organizations, you can maintain quality and manage costs with smaller models through context management, code verification, and reinforcement learning. Discover additional insights:

Grok 4.5 goes public tomorrow. Here’s everything that’s known. The model runs on V9, xAI’s new 1.5 trillion parameter foundation, roughly three times the size of the v8-small architecture behind Grok 4.3. It’s the largest model the company has shipped, trained on Blackwell GPUs at the Colossus cluster in Memphis. Coding is the clear focus. Cursor’s developer data was added in supplemental training, and reinforcement learning has been improving the model daily alongside the Grok Build harness. The next foundation model already in training will have Cursor data included from the start of pre-training. Elon calls it Opus-level, and says it’s faster, more token-efficient, and cheaper. It’s been in private beta at SpaceX and Tesla since June 28, with early internal evals reportedly at or above that mark. Public benchmarks land once the model does. This is the first flagship under SpaceXAI since the merger, and the plan is a from-scratch foundation model shipping every month through the end of 2026. Tomorrow we get to use it!

想系统学强化学习，最劝退的往往是两种资料：一种只讲概念，学完照样不会做；另一种公式堆满页，两章就读不下去。 开源教材 Mathematical Foundations of Reinforcement Learning 正好卡在中间：讲得清楚，推导严谨但不“吓人”，还配套大量视频，把经典算法从定义到实现一步步讲透。 它用数学视角把 RL 的核心框架梳理得很顺，用大量例子把抽象概念落地，让那些“看起来很玄”的更新公式变得能理解、能复现、能写出来。 GitHub： 你会收获什么： - 核心概念一次打通：状态、动作、策略、价值函数等完整串联成体系 - 经典算法拆到细节：MC / TD / Q-learning 从原理到推导循序展开 - 50+ 集中英文视频配套：边看边做，理解与实现同步推进 - 大量网格世界示例：用直观实验对齐公式、更新规则和结果 - 推导规范但不绕：少跳步、不糊弄，难度控制得刚刚好 - 多语言代码实现：Python、R、C++ 等可直接对照复现 适合想把强化学习理论真正学扎实、并能落到实现的人；建议先具备基础概率论和线性代数再开始。

Aloha! 🌺 Meet Ornith-1.0, a family of open-source LLMs specialized for agentic coding. Ornith-1.0 spans the full parameter sizes including 9B Dense, 31B Dense, 35B MoE, and 397B MoE. It achieves state-of-the-art performance among open-source models of comparable size on coding benchmarks including: ✅Terminal-Bench 2.1(77.5) ✅SWE-Bench(82.4 on verified, 62.2 on pro, 78.9 on Multilingual) ✅NL2Repo(48.2) ✅SWE Atlas(41.2 on QnA, 42.6 RF, 39.1 TW) ✅ClawEval(77.1) Post-trained on top of gemma4 and qwen3.5, Ornith-1.0 employs a novel self-improving training strategy in which reinforcement learning is used to generate not only solution rollouts, but also the task-specific scaffolds that drive those rollouts. By jointly optimizing the scaffold and the resulting solution, the model generate higher-quality solutions in agentic coding.😎 All models are released under the MIT license, enabling full commercial and research use. 📖Tech Blog: 🤗Huggingface:

我认为这是三年以来AI对齐的史诗级突破。 OpenAI 团队刚刚丢下一颗重磅炸弹：最新研究论文 《Reinforcement Learning Towards Broadly and Persistently Beneficial Models》。 这一次，他们彻底颠覆了传统的 AI 对齐路径，打破了越安全越笨的魔咒。 这次，杀招是Beneficial Trait RL，我们中文翻译为益处特质强化学习。 他们直接去训练 AI 的核心行为特质，比如诚实、纠错能力、认知谦逊。 这次，OpenAI直接重塑了 AI 的底层人格。 这次，研究人员仅仅在医疗健康一个特定领域训练了 AI 的这些有益特质，结果发现： AI 在医疗以外的、完全没见过的 53 个 OOD测试中，在超过 80%的基准测试上性能全面飙升。它自动学会了拒绝Reward Hacking。科技终于不再盲目迎合，甚至学会了自动识破欺骗。这是伟大的进步。 这次，经过特质强化训练的模型，展现出了惊人的Persistence。 即使面对恶意洗脑和有害微调，它依然能够死死守住底线，拒绝退化。 我们可以确定，它拥有了真正的精神抗体。 在 AI 对齐领域，一直存在一个让人绝望的对齐税，即Alignment Tax。 你想让 AI 越安全，它的通用能力通常就会下降，或者变得极其缩手缩脚。 但 OpenAI 这次用数据证明了，给 AI 注入美德，不仅没有让它变蠢，反而让它在面对未知世界时更加强韧、更有智慧。 这次，Step-change般的胜利告诉我们，当 AI 开始拥有广义的、持久的、能够跨越领域的向善人格，我们距离真正安全、能替人类走向星辰大海的 AGI 代理，又极大地往前迈了一步。未来，当然可期。