NeuTouch、Prophesee 事件相机与 VT-SNN 构成可复查的事件驱动视觉触觉感知节点；作者公开代码和数据入口，并报告容器/重量分类与旋转滑移检测，但无 WAM/VLA/WBC 动作输出或人形/足式闭环证据。
- 代码 · 已核验官方资源
审核码：verified_official; 官方关系：official_author_repository_fixed_commit; 最近核验：2026-08-20代码：source available with preprocessing, fold generation and VT-SNN/MLP-GRU/CNN3D training configurations
权重：未披露
数据：未披露
许可证：not_disclosed；No LICENSE, COPYING or NOTICE file was found; repository visibility is not a software reuse grant.
依赖：Python 3.6.10、PyTorch 1.4.0、torchvision 0.2.2、Guild AI、SLAYER PyTorch fork pinned at 856e010
硬件/传感器：CUDA-capable GPU for training、Intel Loihi for the inference-only comparison、Panda+Robotiq+NeuTouch+Prophesee for equivalent collection
实机证据：Author data were collected on a Franka Emika Panda with Robotiq 2F-140, two NeuTouch sensors and Prophesee event vision.；The released code is an offline perception stack; no closed-loop regrasp or physical controller is supplied.
独立复现：No independent fixed-version reproduction or cross-embodiment deployment was located.
限制/反面证据：No pretrained checkpoints or Loihi image.；No software licence.；No complete NeuTouch electronics/manufacturing release or robot controller.
当前建议：证据观察；Useful for event-stream perception baselines, but licensing, weights and closed-loop deployment remain unresolved.
重点标签：事件视觉触觉感知奠基路线、开放不完整
核验来源：来源 1 · 来源 2 · 来源 3
- 数据集 · 已核验官方资源
审核码：verified_official; 官方关系：official_author_download_page; 最近核验：2026-08-20代码：未披露
权重：未披露
数据：download page lists container-weight V2, YCB, rotational-slip, preprocessing packages, taxel coordinates, camera bias files and mount CAD
许可证：not_disclosed；No standalone dataset licence was verified; the missing code licence cannot be imputed to data.
依赖：7 GB container Lite or 86 GB with RGB、9 GB YCB Lite or 100 GB with RGB、300 MB slip Lite or 4.4 GB full、VT_SNN preprocessing code
硬件/传感器：No robot required for offline parsing、Equivalent collection requires Panda, Robotiq 2F-140, NeuTouch, Prophesee and OptiTrack
实机证据：The files are author-collected real sensor recordings.；The later V2 container release has 800 recordings, while the paper benchmark used 300; the counts must remain distinct.
独立复现：No independent replay report, checksum manifest or cross-platform reproduction was located.
限制/反面证据：No standalone data licence.；Large archives were not mirrored into the public ECS payload.；Sample-stratified splits do not prove unseen-object or cross-embodiment generalization.
当前建议：可复查数据·权利待核；Suitable for offline fusion benchmarking after rights review; not evidence of a deployable tactile policy.
重点标签：事件视觉触觉数据、许可证待核
核验来源：来源 1 · 来源 2 · 来源 3
- 权重 · verified_hold
审核码：verified_hold; 官方关系：official_repository_no_pretrained_checkpoint; 最近核验：2026-08-20代码：未披露
权重：not released: no official pretrained checkpoint, Loihi image or model revision was located
数据：未披露
许可证：not_disclosed；No model-weight licence because no weight release was verified.
依赖：未披露
硬件/传感器：Loihi result reported, reusable deployment image absent
实机证据：Author real-sensor data exist, but no trained model was found.
独立复现：No independent trained-weight release was located.
限制/反面证据：Training code availability must not be rendered as pretrained-weight availability.
当前建议：证据观察；Retain as an explicit absence until an official checkpoint appears.
重点标签：权重未发布
核验来源：来源 1