Molmo2-ER: Human-Annotated Robotics Video QA

A subset of Google DeepMind's RoboVQA dataset, re-hosted for loader compatibility. Human-annotated long-horizon robotics video question-answering data across three embodiments, used to train the allenai/Molmo2-ER-4B model. The upstream dataset is described in the paper 'RoboVQA: Multimodal Long-Horizon Reasoning for Robotics' (arXiv:2311.00899).

Use Cases

• Train multimodal models for video-based question answering based on the described human annotations.
• Benchmark long-horizon reasoning capabilities in robotics based on the described video QA tasks.
• Evaluate AI performance across different robotic embodiments based on the described three embodiments.

Strengths

• Human-annotated data, which likely provides high-quality ground truth.
• Focuses on long-horizon reasoning, a complex task in robotics.
• Includes data across three different robotic embodiments.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Google DeepMind RoboVQA

Collection Method

Re-hosted subset of the upstream dataset.

Freshness

Last updated 2026-05-05 10:11:16; freshness should be verified.