[
 {
  "id": "step",
  "arxiv_id": "2611.08097",
  "url": "https://arxiv.org/abs/2611.08097",
  "title": "STEP: Spatiotemporal Training with Explicit Prediction for Perceptive Humanoid Locomotion",
  "authors": "Youhao Lyu et al.",
  "affiliation": "University of Science and Technology of China, Noetix Robotics and others",
  "date": "2026-11-05",
  "capability": "walking",
  "robot": "Noetix E1 (23 joints)",
  "setting": "Physical robot and simulation",
  "evidence": "Strong",
  "what_they_did": "A walking controller that uses a depth camera to predict the terrain ahead, and keeps track of terrain it can no longer see, such as the stairs behind it when it walks backward.",
  "result": "98 of 100 hardware trials succeeded across stairs, a 40 cm platform, a 70 cm gap, stepping stones and walking backward down stairs. In simulation it beat other ways of combining vision with body sensing.",
  "caveat": "Short trials on a fixed course, run by the authors. Detailed simulation figures are in the full paper.",
  "checked_against": "Abstract and paper text",
  "peer_reviewed": false,
  "added": "2026-11-13"
 },
 {
  "id": "dkwbc",
  "arxiv_id": "2611.09605",
  "url": "https://arxiv.org/abs/2611.09605",
  "title": "DKWBC: Decoupling Motion Tracking and Balancing for Humanoid Whole-Body Control",
  "authors": "Dingkun Wei et al.",
  "affiliation": "",
  "date": "2026-11-06",
  "capability": "walking",
  "robot": "Not named in the abstract",
  "setting": "Physical robot and simulation",
  "evidence": "Moderate",
  "what_they_did": "Splits whole-body control in two: a physics model makes the robot follow a motion, and a learned policy adds small torque corrections so it stays balanced when standing, carrying a load or being touched.",
  "result": "The authors report that one learned balance policy worked across different motion datasets, in simulation and on real hardware.",
  "caveat": "The abstract reports no numbers and does not name the robot or the comparison methods.",
  "checked_against": "Abstract",
  "peer_reviewed": false,
  "added": "2026-11-13"
 },
 {
  "id": "ergocub-terrain",
  "arxiv_id": "2611.05585",
  "url": "https://arxiv.org/abs/2611.05585",
  "title": "Contact-Implicit Whole-Body Trajectory Optimization for Dynamic Humanoid Locomotion on Parameterizable Uneven Terrain",
  "authors": "Roberto Mauceri et al.",
  "affiliation": "",
  "date": "2026-11-03",
  "capability": "walking",
  "robot": "ergoCub",
  "setting": "Physical robot and simulation",
  "evidence": "Preliminary",
  "what_they_did": "Plans whole-body walking motions over uneven ground without scripting in advance where each foot will land.",
  "result": "The authors describe preliminary feasibility tests in simulation and on the real ergoCub robot over non-flat terrain.",
  "caveat": "No success rates or baselines are reported; the authors call the tests preliminary.",
  "checked_against": "Abstract",
  "peer_reviewed": false,
  "added": "2026-11-13"
 },
 {
  "id": "hug",
  "arxiv_id": "2611.01659",
  "url": "https://arxiv.org/abs/2611.01659",
  "title": "HuG: Learning Humanoid Whole-Body Affordance-Aware Grasping with Active Perception",
  "authors": "Xuanbin Peng et al.",
  "affiliation": "UC San Diego",
  "date": "2026-11-01",
  "capability": "hands",
  "robot": "Unitree G1 with Inspire hands",
  "setting": "Physical robot and simulation",
  "evidence": "Strong",
  "what_they_did": "Teaches a humanoid to find an object on a shelf, turn its head to see it better, step and bend into position, and grasp it with dexterous hands. Trained only in simulation, with no real demonstrations.",
  "result": "87.5% success over 40 real trials with 10 household objects on a multi-level shelf, and 85% on objects it had not seen in training. The best baseline reached 52.5%; baselines trained on 70 human demonstrations per object and shelf level reached 32.5% to 50%.",
  "caveat": "Lab test with standard benchmark objects. The model ran on an external desktop GPU connected by cable, not on the robot.",
  "checked_against": "Abstract and paper text",
  "peer_reviewed": false,
  "added": "2026-11-13"
 },
 {
  "id": "forpc",
  "arxiv_id": "2611.08664",
  "url": "https://arxiv.org/abs/2611.08664",
  "title": "ForPC: Force-Aware Whole-Body Control for Precise Humanoid Cart Pushing",
  "authors": "ZhenYu Zou et al.",
  "affiliation": "",
  "date": "2026-11-05",
  "capability": "hands",
  "robot": "Unitree G1",
  "setting": "Physical robot and simulation",
  "evidence": "Moderate",
  "what_they_did": "Pushes a loaded cart along a route with both hands, estimating the forces from the robot's own joint readings instead of using force sensors.",
  "result": "On a real G1, it pushed a cart carrying up to 30 kg at 0.5 m/s without extra tuning. In simulation it cut speed-tracking error by up to 68.2% and path error by up to 74.2% against a walking-only baseline.",
  "caveat": "The baseline comparison is simulation-only; the real-robot test reports no success rate.",
  "checked_against": "Abstract",
  "peer_reviewed": false,
  "added": "2026-11-13"
 },
 {
  "id": "anchor-1",
  "arxiv_id": "2611.00193",
  "url": "https://arxiv.org/abs/2611.00193",
  "title": "Anchor 1: Precise and Stable Humanoid Control for Whole-Body Manipulation",
  "authors": "Archon Robotics, Caojun Wang et al.",
  "affiliation": "Archon Robotics",
  "date": "2026-10-30",
  "capability": "hands",
  "robot": "Not named in the abstract",
  "setting": "Physical robot and simulation",
  "evidence": "Moderate",
  "what_they_did": "A whole-body controller that copies a human operator's motion precisely and stably, to make teleoperated data collection faster.",
  "result": "30% to 80% lower tracking error than existing controllers in the authors' strictest real-world scoring, using fewer look-ahead inputs.",
  "caveat": "The comparison controllers and the robot are not named in the abstract. This is teleoperation, not autonomy.",
  "checked_against": "Abstract",
  "peer_reviewed": false,
  "added": "2026-11-13"
 },
 {
  "id": "lhnav",
  "arxiv_id": "2611.12136",
  "url": "https://arxiv.org/abs/2611.12136",
  "title": "LHNav: Long-Horizon Humanoid Navigation via Multirate Coordination and Physics-Grounded Action Data",
  "authors": "Xueyang Wang et al.",
  "affiliation": "",
  "date": "2026-11-09",
  "capability": "navigation",
  "robot": "Unitree G1",
  "setting": "Physical robot and simulation",
  "evidence": "Moderate",
  "what_they_did": "A vision-language-action model that walks a humanoid to goals described in words, and chains several goals into one long trip.",
  "result": "75.56% success on long, multi-goal trips in simulation after training only on single-goal examples, and better than two earlier navigation models on single goals. On a real G1 it completed single-goal and multi-goal trips, including with rearranged objects.",
  "caveat": "The headline number is from simulation; no success rates are given for the real-robot runs.",
  "checked_against": "Abstract",
  "peer_reviewed": false,
  "added": "2026-11-13"
 },
 {
  "id": "reflexarc",
  "arxiv_id": "2611.04424",
  "url": "https://arxiv.org/abs/2611.04424",
  "title": "ReflexArc: Bottom-Up Integration of Instantaneous Tactile Reflexes into Whole-Body Humanoid Control",
  "authors": "Mingyu Wang et al.",
  "affiliation": "",
  "date": "2026-11-03",
  "capability": "working-with-people",
  "robot": "Unitree G1",
  "setting": "Physical robot",
  "evidence": "Strong",
  "what_they_did": "Gives a humanoid a fast touch reflex: skin sensors trigger a pull-away motion sent straight to the motors, skipping the slower main controller.",
  "result": "The reflex reaches the motors within 10 ms. Against the same reflex delayed by 20 ms, the first force spike fell 34.1% and contact impulse fell from 0.263 to 0.123 N·s, about 2.15 times lower, across three kinds of hazardous contact.",
  "caveat": "Controlled lab contact tests. This is not a safety certification.",
  "checked_against": "Abstract",
  "peer_reviewed": false,
  "added": "2026-11-13"
 },
 {
  "id": "affectivevla",
  "arxiv_id": "2611.07712",
  "url": "https://arxiv.org/abs/2611.07712",
  "title": "AffectiveVLA: Safe and Intent-Aware VLA Control for Human–Humanoid Social Interaction",
  "authors": "Nikita Kuzmin et al.",
  "affiliation": "Skolkovo Institute of Science and Technology",
  "date": "2026-11-05",
  "capability": "working-with-people",
  "robot": "Unitree G1 with BrainCo hands",
  "setting": "Physical robot",
  "evidence": "Strong",
  "what_they_did": "Reads what a person intends from video and responds with a handshake, fist bump or hug, with a spoken safe-stop command.",
  "result": "With intent recognition added, it scored 0.90 on a combined accuracy measure (macro-F1, 0 to 1) over 150 real-robot trials: hug 1.00, handshake 0.92, fist bump 0.78. The best model without it scored 0.75. The voice stop worked in 10 of 10 trials.",
  "caveat": "Training data came from 19 people, and the user study on how safe and comfortable it feels is still running.",
  "checked_against": "Paper text",
  "peer_reviewed": false,
  "added": "2026-11-13"
 },
 {
  "id": "compliantbot",
  "arxiv_id": "2611.01748",
  "url": "https://arxiv.org/abs/2611.01748",
  "title": "CompliantBot: Learning Whole-Body Compliance across Task Interfaces for Humanoid Control",
  "authors": "Haoyang Li et al.",
  "affiliation": "",
  "date": "2026-11-01",
  "capability": "working-with-people",
  "robot": "Not named in the abstract",
  "setting": "Physical robot and simulation",
  "evidence": "Moderate",
  "what_they_did": "Teaches a humanoid to give way when pushed or pulled, by moving its arms, stepping or turning its body, instead of resisting.",
  "result": "The authors report it yielded farther under steady pulls, with lower interaction forces, lower joint power and the highest task completion under strong pulls among the methods compared. Real-robot demonstrations are shown.",
  "caveat": "The abstract gives no numbers or named baselines, and does not say which comparisons were run on hardware.",
  "checked_against": "Abstract",
  "peer_reviewed": false,
  "added": "2026-11-13"
 }
]