truelabelstaging

HUGGING FACE DATASET

SandThink

XDUImageLab/SandThink — source facts cached from the per-dataset Hugging Face API.

DIRECT ANSWER

License: MIT; size: n<1K; creator: XDUImageLab; paper: No arXiv link in cached metadata.

Commercial use: Review the linked license and upstream rights.

API snapshot · Jul 14, 2026

Downloads

2,099

Hugging Face metadata

Size category

n<1K

Hugging Face metadata

Last modified

Mar 5, 2026

Hugging Face metadata

Access

Not marked as gated

SOURCE CARD

What the SandThink card says

Per-dataset API

Dataset-card excerpt

SandThink Dataset (v1.0) SandThink 是一个专为具身智能 (Embodied AI) 任务设计的大规模指令微调与偏好对齐数据集。该数据集通过结构化的 Chain-of-Thought (CoT) 推理过程,显著提升了 Vision-Language-Action (VLA) 模型在复杂环境下的任务拆解、路径规划和动作执行能力。 📊 数据集概览 (Dataset Summary) 本数据集包含三个核心组件,总计约 37,000 条高质量数据: 文件名 类型 规模 说明 Mars_CoT.jsonl SFT / CoT ~17,000 条 针对特定复杂环境(如类地/火星沙盘)的高级推理数据。 iter_DPO_data.jsonl DPO / RLHF ~16,000 条 包含 Chosen 和 Rejected 的偏好对,用于模型对齐与优化。 train_CoT.jsonl SFT / CoT ~4,000 条 基础任务的思维链训练数据,涵盖避障、导航等指令。…

Hugging Face metadata

Captured tags

  • Formats: Imagefolder
  • Modalities: Image

Hugging Face

Open the source dataset

Review the current card, files, access terms, and metadata.

ALTERNATIVES

Datasets related to SandThink

TAGS

Metadata captured for SandThink

Sources

TRUELABEL ROUTING

Need data that matches a specific deployment?

Use the source facts and gaps from this profile to define a custom physical AI data request.

Generate request spec