研究背景
?不对,不对,要准确,不能随便加,哦不对,标题要贴合内容,而且不能改原文信息,哦不对,小节的标题是基于原文的内容,不能瞎编,哦对,要语义化,比如首先研究背景与目标
,然后里面放第一段介绍的内容:A legged robot can learn to walk inside a learned simulator: a neural network that predicts the robot's next state and is run forward on its own predictions. They are usually trained and checked one step ahead, although a policy uses them for a hundred steps or more. We audit one published system end to end, without a GPU: the dynamics model of the Robotic World Model (RWM) for the ANYmal D quadruped, and the uncertainty estimate of its follow-up, RWM-U.
然后下一个模型重建与实验设置
,然后里面的段落:We rebuilt the model to match the released code's outputs, losses and gradients exactly, measured 101 trained simulators, 200 compressed copies and 38 simulators of three other architectures, and trained policies inside simulators that we tested in a hand-built MuJoCo model of the robot.
然后下一个训练效果对比
,里面的内容:Training on the model's own rollouts beat one-step training at 368 steps on all ten held-out episodes (registered), although one-step training was better one step ahead. One-step checks therefore misjudge design choices: pruning 30% of the weights kept 90% of the model's skill one step ahead and 39% at 2 s.
然后下一个不确定性与评估方法验证
,里面的内容:Ensemble disagreement ranked rollout error well (Spearman +0.90 at 100 steps), yet covered 4.6% of outcomes at one sigma, against 68% for a calibrated estimate, as each network's own variance collapses under the implemented loss. Two checks that need no recorded future ranked some pools of simulators well, but neither was shown better than 8-step error in a primary test.
然后机器人层面测试结果
,内容:At the robot level, accuracy has not predicted walking: all 18 fine-tuned copies we tested predicted better than the released model one and eight steps ahead, yet 13 gave failing policies, and registered tests of whether any check ranks simulators by their policies were not settled. In a secondary test, running the authors' policy inside each simulator ranked them well.
然后研究声明与范围
,里面的内容:We also report nine of our own claims that we withdrew.
Scope: one robot, one gait, simulated data, no hardware.
不对?等下,有没有需要列表的地方?哦原文里的那些测量的数量?不对,原文没有列表内容,也没有引用代码,哦对,原文里没有引用、代码、列表,所以不用那些标签。然后检查有没有增删改内容,所有信息都要保留。 等下,不对,小节标题是不是要准确,有没有更合适的?比如不要随便加标题?不对,规则说长文用/
划分小节,所以需要合理划分,而且标题不能改原文内容,只是语义化的归纳。 等下再核对原文的每一部分: 原文内容顺序: 1. 有足机器人可在学习型模拟器中学习行走,模拟器是预测机器人下一状态并基于自身预测前向运行的神经网络。通常训练和检查只前向一步,但策略会使用它们上百步甚至更多。我们端到端审计了一项已发表的系统(无需GPU):面向ANYmal D四足机器人的机器人世界模型(RWM)的动力学模型,以及其后续版本RWM-U的不确定性估计。 2. 我们完全复现了该模型,使其与发布代码的输出、损失和梯度完全匹配,测量了101个训练完成的模拟器、200个压缩副本以及3种其他架构的38个模拟器,还在我们手工搭建的该机器人MuJoCo模型中测试了在模拟器内训练的策略。 3. 在全部10个预留测试集上,基于模型自身rollout的训练在368步时表现优于单步训练(已注册登记),尽管单步训练在单步预测上表现更好。因此单步检查会误判设计选择:裁剪30%的权重后,模型在单步预测时保留了90%的能力,在2秒时仅保留39%的能力。 4. 集成分歧对rollout误差的排序效果良好(100步时斯皮尔曼相关系数为+0.90),但在一倍标准差下仅覆盖了4.6%的结果,而校准后的估计覆盖了68%,这是因为在实现的损失函数下每个网络自身的方差会出现坍塌。两项无需已记录未来数据的检查对部分模拟器池的排序表现较好,但在主要测试中均未显示出优于8步误差的效果。 5. 在机器人层面,精度并未预测行走能力:我们测试的所有18个微调副本在1步和8步预测时的表现都优于已发布的模型,但有13个生成了失败的策略,且关于是否有检查方法能根据策略对模拟器进行排序的注册测试尚未得出结论。在一项次要测试中,在每个模拟器内运行原作者的策略对模拟器的排序效果良好。 6. 我们还报告了我们自行撤回的9项主张。 7. 研究范围:单一机器人、单一步态、模拟数据、无硬件测试。 现在对应划分HTML: 首先
研究背景与审计目标
A legged robot can learn to walk inside a learned simulator: a neural network that predicts the robot's next state and is run forward on its own predictions. They are usually trained and checked one step ahead, although a policy uses them for a hundred steps or more. We audit one published system end to end, without a GPU: the dynamics model of the Robotic World Model (RWM) for the ANYmal D quadruped, and the uncertainty estimate of its follow-up, RWM-U.
然后模型重建与实验设置
We rebuilt the model to match the released code's outputs, losses and gradients exactly, measured 101 trained simulators, 200 compressed copies and 38 simulators of three other architectures, and trained policies inside simulators that we tested in a hand-built MuJoCo model of the robot.
然后训练效果对比结果
Training on the model's own rollouts beat one-step training at 368 steps on all ten held-out episodes (registered), although one-step training was better one step ahead. One-step checks therefore misjudge design choices: pruning 30% of the weights kept 90% of the model's skill one step ahead and 39% at 2 s.
然后评估方法验证
Ensemble disagreement ranked rollout error well (Spearman +0.90 at 100 steps), yet covered 4.6% of outcomes at one sigma, against 68% for a calibrated estimate, as each network's own variance collapses under the implemented loss. Two checks that need no recorded future ranked some pools of simulators well, but neither was shown better than 8-step error in a primary test.
然后机器人层面测试结果
At the robot level, accuracy has not predicted walking: all 18 fine-tuned copies we tested predicted better than the released model one and eight steps ahead, yet 13 gave failing policies, and registered tests of whether any check ranks simulators by their policies were not settled. In a secondary test, running the authors' policy inside each simulator ranked them well.
然后研究声明与范围
We also report nine of our own claims that we