Datasets:

beatsprom
/

robotics-embodied-ai-physical-control-2026

paper_id string	title string	published_date string	arxiv_categories list	legal_license string	license_url string	commercial_use_allowed bool	commercial_ip_safety_score int64	commercial_ip_verdict string	commercial_ip_justification string	field_weighted_citation_impact_fwci float64	tldr_neural_summary string	primary_author string	total_authors_count int64	authors list	affiliations list	arxiv_url string	code_audit_status string	primary_repo_url string	repo_urls list	github_stars int64	github_last_commit string	github_license string	importance_score float64	abstract string	title_vector_384d list	abstract_vector_384d list	target_industry string	embodied_policy_architecture string	robot_morphology_target string	physics_simulators_supported list	tested_benchmarks list	extracted_benchmark_scores list	reproduction_recipe string	core_problem_addressed string	key_technical_innovation string	key_quantitative_result string	cluster_id int64	cluster_topic_name string	trend_velocity_tier string	papers_last_30_days int64	audited_at string
2606.16533v3	Kairos: A Regret-Aware Native World-Action Model Stack for Physical AI	2026-06-15T10:37:42Z	[ "cs.AI", "cs.CV" ]	ArXiv Standard	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	true	60	Internal R&D Only (Copyleft or Academic Terms)	ArXiv Standard Distribution; Repository License Unspecified	1	Proposes First, it \textbf{learns} control-relevant information through a \textbf{Cross-Embodiment... to enable robust physical control, showing Experiments on embodied world-model benchmarks, world-action benchmarks, long-ho....	Kairos Team	24	[ "Kairos Team", "Fei Wang", "Shan You", "Qiming Zhang", "Tao Huang", "Zuoyi Fu", "Zhisheng Zheng", "Yunlong Xi", "Feng Lv", "Xiaoming Wu", "Zeyu Liu", "Cong Wan", "Pu Li", "Ruiqing Yang", "Xiaoou Li", "Wei Wang", "Kangkang Zhu", "Yuwei Zhang", "Shi Fu", "Zheng Zhang", "Xiaonin...	[ "Embodied AI & Robotics Research Institute" ]	http://arxiv.org/abs/2606.16533v3	VERIFIED_LIVE	https://github.com/kairos-agi/kairos	[ "https://github.com/kairos-agi/kairos" ]	2,500	2026-08-20	Unspecified	105.32	We introduce \textbf{Kairos}, a regret-aware native world-action model stack for Physical AI. Kairos is motivated by the view that a physical world model should not aim to fully simulate all future pixels, but should learn and maintain the information most relevant to embodiment control: object state, spatial relations...	[ 0.010040000081062317, -0.06987400352954865, 0.059804998338222504, 0.01889999955892563, 0.0018690000288188457, -0.004373000003397465, 0.04418899863958359, -0.0008590000215917826, 0.055969998240470886, 0.03574199974536896, 0.015100999735295773, -0.1407950073480606, -0.012768000364303589, 0.0...	[ 0.03864099830389023, -0.12435799837112427, 0.04964200034737587, 0.025200000032782555, 0.01908399909734726, 0.010246999561786652, 0.032958000898361206, -0.007143000140786171, 0.021030999720096588, 0.022539999336004257, -0.01054299995303154, -0.09024500101804733, 0.0038739999290555716, 0.034...	Robotics, Embodied AI & Autonomous Physical Control	Neural Sensorimotor Policy Network	General Robotic & Embodied Manipulator	[ "MuJoCo Physics Engine (Contact Dynamics)" ]	[ "BEHAVIOR-1K (Everyday Household Task Benchmark)" ]	[ "Quantitative Physical Policy & Task Completion Benchmark" ]	git clone https://github.com/kairos-agi/kairos && cd kairos && (pip install -e . \|\| pip install -r requirements.txt)	We introduce \textbf{Kairos}, a regret-aware native world-action model stack for Physical AI.	First, it \textbf{learns} control-relevant information through a \textbf{Cross-Embodiment Data Curriculum}, which organizes open-world videos, human behavioral data, and robot interactions into an intervention-strength progression from passive physical observation to intentional behavior and embodied action grounding.	Experiments on embodied world-model benchmarks, world-action benchmarks, long-horizon generation, and inference-efficiency evaluation show that Kairos achieves superior performance while offering a favorable efficiency to capability trade-off.	4	Sim-to-Real Transfer & Physics Simulation Environments	Explosive (>50/mo)	773	2026-08-20T16:42:57.580836
2607.17977v2	RynnBrain 1.1: Towards More Capable and Generalizable Embodied Foundation Model	2026-07-20T14:13:27Z	[ "cs.RO" ]	ArXiv Standard	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	true	60	Internal R&D Only (Copyleft or Academic Terms)	ArXiv Standard Distribution; Repository License Unspecified	1	Proposes Compared with RynnBrain 1.0, it further introduces contact-point prediction across the mod... to enable robust physical control, showing Real-robot experiments show that RynnBrain-initialized policies outperform Qwen-....	Kehan Li	31	[ "Kehan Li", "Bohan Hou", "Minghao Zhu", "Tianyi Zhang", "Zesen Cheng", "Zhikai Wang", "Sicong Leng", "Xin Li", "Xiao Lin", "Biying Yao", "Minghua Zeng", "Jiangpin Liu", "Ronghao Dang", "Jiayan Guo", "Siteng Huang", "Haoyu Zhao", "Heng Ping", "Yaxi Zhao", "Tong Zhao", "Kexiang W...	[ "Embodied AI & Robotics Research Institute" ]	http://arxiv.org/abs/2607.17977v2	VERIFIED_LIVE	https://github.com/alibaba-damo-academy/RynnBrain	[ "https://github.com/alibaba-damo-academy/RynnBrain" ]	864	2026-08-20	Unspecified	97.5	We present RynnBrain 1.1, a family of embodied foundation models spanning 2B, 9B, and 122B-A10B scales. Trained with a unified spatio-temporal and physically grounded framework, RynnBrain 1.1 supports embodied perception, spatial reasoning, localization, and planning. Compared with RynnBrain 1.0, it further introduces ...	[ 0.03355000168085098, -0.1056469976902008, 0.03700000047683716, -0.03755699843168259, 0.028821999207139015, 0.002985999919474125, -0.002050000010058284, 0.007848000153899193, -0.03360399976372719, 0.024788999930024147, -0.05519900098443031, -0.03990599885582924, -0.03910500183701515, 0.0534...	[ -0.05967799946665764, -0.17142899334430695, 0.05539200082421303, -0.021669000387191772, -0.05787400156259537, 0.024101000279188156, -0.008936000056564808, -0.006837999913841486, 0.005468000192195177, 0.009901000186800957, -0.022427000105381012, -0.09720399975776672, -0.03693800047039986, 0...	Robotics, Embodied AI & Autonomous Physical Control	Vision-Language-Action Model (VLA - OpenVLA/RT-2/Octo)	Multi-DoF Robotic Arm & Dexterous Hand	[ "MuJoCo Physics Engine (Contact Dynamics)" ]	[ "Robotic Manipulation Success & Task Completion Rate" ]	[ "Quantitative Physical Policy & Task Completion Benchmark" ]	git clone https://github.com/alibaba-damo-academy/RynnBrain && cd RynnBrain && (pip install -e . \|\| pip install -r requirements.txt)	We present RynnBrain 1.1, a family of embodied foundation models spanning 2B, 9B, and 122B-A10B scales.	Compared with RynnBrain 1.0, it further introduces contact-point prediction across the model family and native 3D grounding for the 2B and 9B models, yielding representations and outputs that are more directly aligned with robot manipulation.	Real-robot experiments show that RynnBrain-initialized policies outperform Qwen-based and representative generalist VLAs, while joint multi-task and multi-embodiment training improves process scores and success rates over per-task training.	3	Dexterous Robotic Manipulation & Tactile Gripping	Explosive (>50/mo)	773	2026-08-20T16:40:16.616712
2607.27205v2	TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAM	2026-07-29T17:59:58Z	[ "cs.CV", "cs.RO" ]	ArXiv Standard	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	true	60	Internal R&D Only (Copyleft or Academic Terms)	ArXiv Standard Distribution; Repository License Unspecified	1	Proposes Although effective, this design incurs substantial computation and memory overhead at ever... to enable robust physical control, showing These results establish TurboVLA as a simple and effective alternative to the pr....	Hengyi Xie	7	[ "Hengyi Xie", "Chenfei Yao", "Xianjin Wu", "Yingying Zhu", "Dingkang Liang", "Xiang Bai", "Han Ding" ]	[ "Embodied AI & Robotics Research Institute" ]	http://arxiv.org/abs/2607.27205v2	VERIFIED_LIVE	https://github.com/H-EmbodVis/TurboVLA	[ "https://github.com/H-EmbodVis/TurboVLA" ]	449	2026-08-20	Unspecified	92.18	Vision-language-action (VLA) models commonly adopt an LLM-centric $V \to L \to A$ pathway, where visual observations are projected into the representation space of a large language model before being decoded into robot actions. Although effective, this design incurs substantial computation and memory overhead at every ...	[ 0.04207799956202507, -0.0663679987192154, -0.00828499998897314, -0.03687300160527229, -0.06298799812793732, 0.0185100007802248, -0.0030950000509619713, -0.006653999909758568, -0.011587999761104584, -0.006256999913603067, 0.02366499975323677, -0.07694099843502045, -0.07769899815320969, 0.03...	[ 0.02479100041091442, -0.127143993973732, 0.018470000475645065, -0.0010580000234767795, 0.023204000666737556, 0.03717900067567825, -0.03147599846124649, 0.0015960000455379486, 0.03392999991774559, -0.0006430000066757202, -0.021477000787854195, -0.06792700290679932, 0.0019359999569132924, 0....	Robotics, Embodied AI & Autonomous Physical Control	Vision-Language-Action Model (VLA - OpenVLA/RT-2/Octo)	Multi-DoF Robotic Arm & Dexterous Hand	[ "MuJoCo Physics Engine (Contact Dynamics)" ]	[ "Robotic Manipulation Success & Task Completion Rate" ]	[ "Quantitative Physical Policy & Task Completion Benchmark" ]	git clone https://github.com/H-EmbodVis/TurboVLA && cd TurboVLA && (pip install -e . \|\| pip install -r requirements.txt)	Vision-language-action (VLA) models commonly adopt an LLM-centric $V \to L \to A$ pathway, where visual observations are projected into the representation space of a large language model before being decoded into robot actions.	Although effective, this design incurs substantial computation and memory overhead at every policy invocation.	These results establish TurboVLA as a simple and effective alternative to the prevailing LLM-centric VLA paradigm, offering a new perspective on how vision, language, and action can be connected for efficient robotic manipulation.	4	Sim-to-Real Transfer & Physics Simulation Environments	Explosive (>50/mo)	773	2026-08-20T16:39:50.966120
2606.19555v1	SCAN-Planner: Spatial Collision-Aware Local Planning for Route-Guided Long-Range Quadruped Navigation	2026-06-17T19:55:09Z	[ "cs.RO" ]	ArXiv Standard	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	true	60	Internal R&D Only (Copyleft or Academic Terms)	ArXiv Standard Distribution; Repository License Unspecified	1	Proposes Existing local planners commonly approximate the robot using isotropic geometric inflation... to enable robust physical control, showing Simulation and real-world experiments demonstrate that SCAN-Planner generates sa....	Han Zheng	5	[ "Han Zheng", "Zhe Chen", "Yiwen Fu", "Ming Yang", "Tong Qin" ]	[ "Embodied AI & Robotics Research Institute" ]	http://arxiv.org/abs/2606.19555v1	VERIFIED_LIVE	https://github.com/wuyi2121/SCAN-Planner	[ "https://github.com/wuyi2121/SCAN-Planner" ]	462	2026-08-20	Unspecified	90.75	Quadruped robots are increasingly expected to navigate through narrow passages, cluttered indoor scenes, and large-scale 3D unstructured environments. Existing local planners commonly approximate the robot using isotropic geometric inflation or rely on planar and elevation-map representations, leading to conservative m...	[ 0.1658650040626526, 0.004889000207185745, -0.018187999725341797, -0.01945500075817108, -0.027919000014662743, -0.002503999974578619, -0.08579900115728378, -0.008012999780476093, -0.08808399736881256, 0.04447399824857712, -0.02963400073349476, 0.04910000041127205, 0.02328299917280674, 0.012...	[ 0.016726000234484673, -0.013856000266969204, 0.013283000327646732, -0.0011370000429451466, 0.021771999076008797, -0.04729200154542923, -0.09643799811601639, 0.0008319999906234443, -0.03716199845075607, 0.005590999964624643, -0.004253999795764685, -0.04035000130534172, 0.02201399952173233, ...	Robotics, Embodied AI & Autonomous Physical Control	Neural Sensorimotor Policy Network	Quadruped / Legged Robot	[ "MuJoCo Physics Engine (Contact Dynamics)" ]	[ "Sim-to-Real Physical Transfer & Domain Randomization Evaluation" ]	[ "Quantitative Physical Policy & Task Completion Benchmark" ]	git clone https://github.com/wuyi2121/SCAN-Planner && cd SCAN-Planner && (pip install -e . \|\| pip install -r requirements.txt)	Quadruped robots are increasingly expected to navigate through narrow passages, cluttered indoor scenes, and large-scale 3D unstructured environments.	Existing local planners commonly approximate the robot using isotropic geometric inflation or rely on planar and elevation-map representations, leading to conservative motion in tight spaces and limited reasoning about overhanging structures.	Simulation and real-world experiments demonstrate that SCAN-Planner generates safe, smooth, and efficient trajectories in dense clutter, 3D unstructured scenes, stair traversal, and long-range navigation tasks.	6	Deep Reinforcement Learning & Reward Policy Optimization	Explosive (>50/mo)	773	2026-08-20T16:42:39.187890
2606.13673v1	SpatialClaw: Rethinking Action Interface for Agentic Spatial Reasoning	2026-06-11T17:59:36Z	[ "cs.CV", "cs.AI" ]	ArXiv Standard	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	true	60	Internal R&D Only (Copyleft or Academic Terms)	ArXiv Standard Distribution; Repository License Unspecified	1	Proposes In this work, we study how the design of this interface shapes the agent's capacity for op... to enable robust physical control, showing Evaluated across 20 spatial reasoning benchmarks spanning a broad range of stati....	Seokju Cho	11	[ "Seokju Cho", "Ryo Hachiuma", "Abhishek Badki", "Hang Su", "Byung-Kwan Lee", "Chan Hee Song", "Sifei Liu", "Subhashree Radhakrishnan", "Seungryong Kim", "Yu-Chiang Frank Wang", "Min-Hung Chen" ]	[ "Embodied AI & Robotics Research Institute" ]	http://arxiv.org/abs/2606.13673v1	VERIFIED_LIVE	https://github.com/NVlabs/SpatialClaw	[ "https://github.com/NVlabs/SpatialClaw" ]	363	2026-08-20	Unspecified	88.42	Spatial reasoning, the ability to determine where objects are, how they relate, and how they move in 3D, remains a fundamental challenge for vision-language models (VLMs). Tool-augmented agents attempt to address this by augmenting VLMs with specialist perception modules, yet their effectiveness is bounded by the actio...	[ 0.02450999990105629, -0.08064799755811691, 0.0013150000013411045, -0.0082609998062253, 0.030326999723911285, -0.026006000116467476, 0.07049500197172165, -0.006293000187724829, -0.011853000149130821, 0.06022899970412254, -0.01361900009214878, -0.019009999930858612, 0.022440999746322632, 0.1...	[ 0.028015999123454094, -0.11297400295734406, -0.03313500061631203, 0.018644999712705612, 0.03122900053858757, -0.0373230017721653, -0.0042329998686909676, 0.000686999992467463, -0.0066999997943639755, -0.016783999279141426, -0.07130099833011627, -0.07278800010681152, -0.004387999884784222, ...	Robotics, Embodied AI & Autonomous Physical Control	Neural Sensorimotor Policy Network	Multi-DoF Robotic Arm & Dexterous Hand	[ "MuJoCo Physics Engine (Contact Dynamics)" ]	[ "Robotic Manipulation Success & Task Completion Rate" ]	[ "Quantitative Physical Policy & Task Completion Benchmark" ]	git clone https://github.com/NVlabs/SpatialClaw && cd SpatialClaw && (pip install -e . \|\| pip install -r requirements.txt)	Spatial reasoning, the ability to determine where objects are, how they relate, and how they move in 3D, remains a fundamental challenge for vision-language models (VLMs).	In this work, we study how the design of this interface shapes the agent's capacity for open-ended spatial reasoning.	Evaluated across 20 spatial reasoning benchmarks spanning a broad range of static and dynamic 3D/4D spatial reasoning tasks, SpatialClaw achieves 59.9% average accuracy, outperforming the recent spatial agent by +11.2 points, with consistent gains across six VLM backbones from two model families without any benchmark- ...	2	Bipedal Humanoid Locomotion & Whole-Body Dynamic Balance	Explosive (>50/mo)	773	2026-08-20T16:43:17.309220
2606.23565v1	HoloAgent-0: A Unified Embodied Agent Framework with 3D Spatial Memory	2026-06-22T16:31:48Z	[ "cs.RO", "cs.CV" ]	ArXiv Standard	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	true	60	Internal R&D Only (Copyleft or Academic Terms)	ArXiv Standard Distribution; Repository License Unspecified	1	Proposes Extending this loop to physical robots is difficult because physical execution is continuo... to enable robust physical control, showing We deploy HoloAgent-0 on real hardware and evaluate its spatial memory, long-hor....	Xiaolin Zhou	12	[ "Xiaolin Zhou", "Liu Liu", "Tingyang Xiao", "Wei Feng", "Fa Fu", "Xinrui Meng", "Xinjie Wang", "Jialiang Han", "Boyang Yu", "Yun Du", "Wei Sui", "Zhizhong Su" ]	[ "Embodied AI & Robotics Research Institute" ]	http://arxiv.org/abs/2606.23565v1	VERIFIED_LIVE	https://github.com/HorizonRobotics/HoloAgent	[ "https://github.com/HorizonRobotics/HoloAgent" ]	342	2026-08-20	Unspecified	88.35	LLM agents follow a practical execution loop in digital environments: they reason over structured states, invoke tools, inspect feedback, and revise actions. Extending this loop to physical robots is difficult because physical execution is continuous, embodiment-dependent, uncertain, and constrained by safety. Existing...	[ 0.08534099906682968, -0.034136999398469925, -0.04519899934530258, -0.04516400024294853, -0.0582440011203289, 0.033918000757694244, -0.0017109999898821115, -0.038015998899936676, -0.04437299817800522, 0.07291200011968613, 0.015607000328600407, 0.0025919999461621046, -0.028085000813007355, 0...	[ 0.046004001051187515, -0.048815999180078506, -0.01608699932694435, -0.05212600156664848, 0.021539000794291496, -0.05517299845814705, -0.02145499922335148, -0.03934900090098381, -0.05693599954247475, 0.04227500036358833, 0.020648999139666557, -0.05540100112557411, -0.02901099994778633, 0.02...	Robotics, Embodied AI & Autonomous Physical Control	Neural Sensorimotor Policy Network	Bipedal Humanoid Robot	[ "MuJoCo Physics Engine (Contact Dynamics)" ]	[ "Robotic Manipulation Success & Task Completion Rate" ]	[ "Quantitative Physical Policy & Task Completion Benchmark" ]	git clone https://github.com/HorizonRobotics/HoloAgent && cd HoloAgent && (pip install -e . \|\| pip install -r requirements.txt)	LLM agents follow a practical execution loop in digital environments: they reason over structured states, invoke tools, inspect feedback, and revise actions.	Extending this loop to physical robots is difficult because physical execution is continuous, embodiment-dependent, uncertain, and constrained by safety.	We deploy HoloAgent-0 on real hardware and evaluate its spatial memory, long-horizon navigation, and closed-loop execution across motion generation, object search, cross-robot coordination, and mobile manipulation.	5	Diffusion Policy & Imitation Teleoperation Learning	Explosive (>50/mo)	773	2026-08-20T16:42:15.959319
2606.22682v1	Integrated cloud-based architecture for robot-robot and human-robot collaboration using ROS 2--MQTT in Mediterranean Greenhouses	2026-06-21T21:49:19Z	[ "cs.RO" ]	ArXiv Standard	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	true	60	Internal R&D Only (Copyleft or Academic Terms)	ArXiv Standard Distribution; Repository License Unspecified	1	Proposes Traditional robotic frameworks, such as ROS 2, frequently encounter node discovery issues... to enable robust physical control, showing The results indicate that the integration of MQTT effectively eliminates informa....	F. Cañadas-Aránega	4	[ "F. Cañadas-Aránega", "M. Muñoz", "J. C. Moreno", "J. L. Blanco-Claraco" ]	[ "Embodied AI & Robotics Research Institute" ]	http://arxiv.org/abs/2606.22682v1	VERIFIED_LIVE	https://github.com/Ar-Ray-code/YOLOX-ROS	[ "https://github.com/Ar-Ray-code/YOLOX-ROS", "https://github.com/IntelligentRoboticsLabs/yolact_ros_3d" ]	324	2026-08-20	Unspecified	87.84	The imperative to develop more sustainable agriculture demands a transition from isolated automation toward the deployment of multi-robot systems (MRS) in agrifood environments. However, Mediterranean greenhouse settings-characterized by narrow corridors, dense biomass, and structural metallic interference-pose signifi...	[ -0.007184999994933605, -0.007362000178545713, 0.0020280000753700733, -0.0045449999161064625, 0.06327799707651138, -0.061827000230550766, -0.10128200054168701, -0.029569000005722046, 0.040394000709056854, 0.045552998781204224, 0.049469999969005585, 0.004683999810367823, 0.07375200092792511, ...	[ -0.012280999682843685, -0.027421999722719193, -0.0022529999259859324, 0.004927999805659056, 0.09227800369262695, -0.1143530011177063, -0.07012200355529785, -0.02515600062906742, 0.018957000225782394, 0.06607399880886078, 0.009770000353455544, -0.04276200011372566, 0.05281500145792961, -0.0...	Robotics, Embodied AI & Autonomous Physical Control	Neural Sensorimotor Policy Network	General Robotic & Embodied Manipulator	[ "MuJoCo Physics Engine (Contact Dynamics)" ]	[ "Sim-to-Real Physical Transfer & Domain Randomization Evaluation" ]	[ "Quantitative Physical Policy & Task Completion Benchmark" ]	git clone https://github.com/Ar-Ray-code/YOLOX-ROS && cd YOLOX-ROS && (pip install -e . \|\| pip install -r requirements.txt)	The imperative to develop more sustainable agriculture demands a transition from isolated automation toward the deployment of multi-robot systems (MRS) in agrifood environments.	Traditional robotic frameworks, such as ROS 2, frequently encounter node discovery issues and latency spikes due to dynamic obstacles, dense foliage, and other characteristic greenhouse elements, creating a critical bottleneck for real-time coordination.	The results indicate that the integration of MQTT effectively eliminates information silos, providing a scalable and decentralised solution for managing complex robotic missions, which are executed locally via Edge Computing.	5	Diffusion Policy & Imitation Teleoperation Learning	Explosive (>50/mo)	773	2026-08-20T16:42:24.142812
2607.04234v1	SoftVTBench: A Safety-Aware Visuo-Tactile Benchmark for Physically Constrained Robotic Manipulation of Deformable Objects	2026-07-05T11:12:29Z	[ "cs.RO", "cs.AI", "cs.CV" ]	ArXiv Standard	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	true	60	Internal R&D Only (Copyleft or Academic Terms)	ArXiv Standard Distribution; Repository License Unspecified	1	Proposes However, existing manipulation benchmarks are predominantly success-oriented and rarely ev... to enable robust physical control, showing that success-only evaluation substantially overstates policy performance, as a l....	Bowen Jing	18	[ "Bowen Jing", "Mingxin Wang", "Ruiyang Hao", "Chenchen Ge", "Hanwen Shen", "Junjie He", "Yang Cui", "Yiming Hou", "Weitao Zhou", "Jiawei Wang", "Minglei Li", "Dandan Zhang", "Ding Zhao", "Houde Liu", "Xiaofan Li", "Si Liu", "Ping Luo", "Haibao Yu" ]	[ "Embodied AI & Robotics Research Institute" ]	http://arxiv.org/abs/2607.04234v1	VERIFIED_LIVE	https://github.com/TuojingAI/SoftVTBench	[ "https://github.com/TuojingAI/SoftVTBench" ]	175	2026-08-20	Unspecified	83.07	Deformable object manipulation poses challenges beyond task completion: successful execution must also maintain safe physical interaction, holding the object stably without slip or drop while avoiding excessive deformation. However, existing manipulation benchmarks are predominantly success-oriented and rarely evaluate...	[ -0.008301000110805035, -0.09338799864053726, -0.04640199989080429, -0.00785600021481514, 0.03716899827122688, -0.0013810000382363796, 0.03067700006067753, 0.09096299856901169, -0.015646999701857567, 0.029286999255418777, 0.021663999184966087, -0.0864029973745346, -0.07371500134468079, 0.06...	[ -0.00868299975991249, -0.04010799899697304, 0.004612999968230724, 0.04642599821090698, 0.05459599941968918, 0.0006559999892488122, 0.02250399999320507, 0.030215000733733177, 0.00444599986076355, 0.01324199978262186, -0.007548999972641468, -0.07238800078630447, -0.0313510000705719, 0.030358...	Robotics, Embodied AI & Autonomous Physical Control	Tactile-Visual Multi-Modal Sensorimotor Policy	Multi-DoF Robotic Arm & Dexterous Hand	[ "NVIDIA Isaac Sim / Isaac Gym (GPU-Accelerated)" ]	[ "Robotic Manipulation Success & Task Completion Rate" ]	[ "Quantitative Physical Policy & Task Completion Benchmark" ]	git clone https://github.com/TuojingAI/SoftVTBench && cd SoftVTBench && (pip install -e . \|\| pip install -r requirements.txt)	Deformable object manipulation poses challenges beyond task completion: successful execution must also maintain safe physical interaction, holding the object stably without slip or drop while avoiding excessive deformation.	However, existing manipulation benchmarks are predominantly success-oriented and rarely evaluate whether a policy remains physically safe throughout execution.	Experiments show that success-only evaluation substantially overstates policy performance, as a large fraction of goal-completing rollouts still violate physical safety.	3	Dexterous Robotic Manipulation & Tactile Gripping	Explosive (>50/mo)	773	2026-08-20T16:41:12.730888
2606.13679v2	InterleaveThinker: Reinforcing Agentic Interleaved Generation	2026-06-11T17:59:50Z	[ "cs.CV" ]	ArXiv Standard	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	true	60	Internal R&D Only (Copyleft or Academic Terms)	ArXiv Standard Distribution; Repository License Unspecified	1	Proposes However, constrained by their architectures, they cannot achieve interleaved generation (t... to enable robust physical control, showing On interleaved generation benchmarks, it achieves performance comparable to Nano....	Dian Zheng	7	[ "Dian Zheng", "Harry Lee", "Manyuan Zhang", "Kaituo Feng", "Zoey Guo", "Ray Zhang", "Hongsheng Li" ]	[ "Embodied AI & Robotics Research Institute" ]	http://arxiv.org/abs/2606.13679v2	VERIFIED_LIVE	https://github.com/zhengdian1/InterleaveThinker	[ "https://github.com/zhengdian1/InterleaveThinker" ]	193	2026-08-20	Unspecified	82.96	Recent image generators have demonstrated impressive photorealism and instruction-following capabilities in single-image generation and editing. However, constrained by their architectures, they cannot achieve interleaved generation (text-image sequence), which has crucial applications in visual narratives, guidance, a...	[ -0.0993259996175766, -0.028223000466823578, 0.00253499997779727, -0.06927599757909775, -0.015451000072062016, -0.06437800079584122, -0.015554999932646751, -0.01834000088274479, -0.06894800066947937, 0.0025400000158697367, 0.00595800019800663, -0.10978599637746811, 0.039489999413490295, -0....	[ -0.05992899835109711, -0.025950999930500984, 0.009304000064730644, 0.000307999987853691, 0.01943800039589405, -0.04301299899816513, -0.10345099866390228, -0.022538000717759132, 0.05068400129675865, -0.03318899869918823, -0.016553999856114388, -0.03465300053358078, 0.04562399908900261, -0.0...	Robotics, Embodied AI & Autonomous Physical Control	Neural Sensorimotor Policy Network	Multi-DoF Robotic Arm & Dexterous Hand	[ "MuJoCo Physics Engine (Contact Dynamics)" ]	[ "Robotic Manipulation Success & Task Completion Rate" ]	[ "Quantitative Physical Policy & Task Completion Benchmark" ]	git clone https://github.com/zhengdian1/InterleaveThinker && cd InterleaveThinker && (pip install -e . \|\| pip install -r requirements.txt)	Recent image generators have demonstrated impressive photorealism and instruction-following capabilities in single-image generation and editing.	However, constrained by their architectures, they cannot achieve interleaved generation (text-image sequence), which has crucial applications in visual narratives, guidance, and embodied manipulation.	On interleaved generation benchmarks, it achieves performance comparable to Nano Banana and GPT-5.	4	Sim-to-Real Transfer & Physics Simulation Environments	Explosive (>50/mo)	773	2026-08-20T16:43:16.186901
2607.02501v3	Embodied.cpp: A Portable Inference Runtime of Embodied AI Models on Heterogeneous Robots	2026-07-02T17:58:28Z	[ "cs.RO", "cs.CV", "cs.OS" ]	ArXiv Standard	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	true	60	Internal R&D Only (Copyleft or Academic Terms)	ArXiv Standard Distribution; Repository License Unspecified	1	"Proposes Existing inference runtimes are designed mainly for request-response serving and therefore(...TRUNCATED)	Ling Xu	11	["Ling Xu","Borui Li","Hao Wu","Chuyu Han","Xiangyu Li","Mohan Hua","Shiqi Jiang","Ting Cao","Chuany(...TRUNCATED)	[ "Embodied AI & Robotics Research Institute" ]	http://arxiv.org/abs/2607.02501v3	VERIFIED_LIVE	https://github.com/SEU-PAISys/Embodied.cpp	[ "https://github.com/SEU-PAISys/Embodied.cpp" ]	146	2026-08-20	Unspecified	81.39	"Embodied AI models now span vision-language-action (VLA) models and world-action models (WAMs), but(...TRUNCATED)	[0.01800600066781044,-0.06749200075864792,0.08949200063943863,0.011629000306129456,0.052418999373912(...TRUNCATED)	[-0.025043999776244164,-0.08452200144529343,0.0657230019569397,0.0058120000176131725,0.0040719998069(...TRUNCATED)	Robotics, Embodied AI & Autonomous Physical Control	Vision-Language-Action Model (VLA - OpenVLA/RT-2/Octo)	General Robotic & Embodied Manipulator	[ "MuJoCo Physics Engine (Contact Dynamics)" ]	[ "Sim-to-Real Physical Transfer & Domain Randomization Evaluation" ]	[ "Quantitative Physical Policy & Task Completion Benchmark" ]	"git clone https://github.com/SEU-PAISys/Embodied.cpp && cd Embodied.cpp && (pip install -e . \|\| pip(...TRUNCATED)	"Embodied AI models now span vision-language-action (VLA) models and world-action models (WAMs), but(...TRUNCATED)	"Existing inference runtimes are designed mainly for request-response serving and therefore do not s(...TRUNCATED)	"These results show that Embodied$.$cpp improves deployment efficiency while preserving high control(...TRUNCATED)	4	Sim-to-Real Transfer & Physics Simulation Environments	Explosive (>50/mo)	773	2026-08-20T16:41:20.466553

End of preview. Expand in Data Studio

🤖 Robotics, Embodied AI & Physical World Control Dataset (2026 Edition)

A structured research dataset featuring 2,123 domain-verified research papers and official code repositories focused on Vision-Language-Action Models (VLA), Humanoid Robotics, Quadruped Locomotion, Diffusion Policies, Sim-to-Real Transfer, and Physics Simulation Environments (Isaac Sim, MuJoCo, Genesis).

Built with Universal Scientific Engine V16.1 Gold, providing 43 schema attributes with verified repository attribution, 8 AI topological semantic clusters, 10 policy architectures, and native 384-dimensional PyTorch embeddings.

📊 Dataset Schema Highlights (43 Columns)

Field	Type	Description
`paper_id`	String	Unique ArXiv identifier
`title`	String	Research paper title
`cluster_topic_name`	String	1 of 8 AI Topological Semantic Clusters
`embodied_policy_architecture`	String	Control policy (VLA, ACT, DP3, WBC, DRL, MPC)
`robot_morphology_target`	String	Robot morphology (Humanoid, Multi-DoF Arm, Quadruped, Drone)
`physics_simulators_supported`	List[String]	Simulation platforms (Isaac Sim, MuJoCo, Genesis, SAPIEN)
`commercial_ip_safety_score`	Integer	0–100 commercial compliance index (92% Enterprise Safe)
`tldr_neural_summary`	String	15-word executive summary of key innovation
`title_vector_384d`	List[Float]	384d PyTorch embedding (`all-MiniLM-L6-v2`)
`abstract_vector_384d`	List[Float]	384d PyTorch embedding (`all-MiniLM-L6-v2`)
`reproduction_recipe`	String	1-line bash setup command

🧩 8 AI Semantic Clusters Breakdown

Sim-to-Real Transfer & Physics Simulation Environments (603 papers)
Dexterous Robotic Manipulation & Tactile Gripping (356 papers)
Bipedal Humanoid Locomotion & Whole-Body Dynamic Balance (279 papers)
Deep Reinforcement Learning & Reward Policy Optimization (266 papers)
Diffusion Policy & Imitation Teleoperation Learning (205 papers)
Vision-Language-Action (VLA) & Foundation Policy Models (162 papers)
Autonomous Mobile Navigation, SLAM & 3D Spatial Mapping (148 papers)
Aerial Quadrotors, Drones & Multi-Robot Swarm Control (104 papers)

💻 1-Click Python Quickstart

import pyarrow.parquet as pq

# Load 100-Sample Teaser
table = pq.read_table("ROBOTICS_EMBODIED_AI_PHYSICAL_WORLD_CONTROL_2026_100_SAMPLE.parquet")
df = table.to_pandas()

print(f"Loaded {len(df)} sample Embodied AI papers.")
print(f"Top Paper: {df['title'].iloc[0]} (Policy: {df['embodied_policy_architecture'].iloc[0]})")

🚀 Get the Full 2,123-Paper Enterprise Edition

The complete commercial production dataset (2,123 papers in Parquet with 384d vectors, SQLite DB, Clean CSV, and JSON) is available here:

👉 BeatsProm Embodied AI & Robotics Dataset Full Edition

Downloads last month: 74