This updates the overall task instructions for all 20 real-world and 15 simulation tasks in both One-Shot and Full-Shot subsets, addressing [RoboValue issue #1](https://github.com/RoboValue-Benchmark/RoboValue/issues/1).
All 35 task definitions were checked. Text changes are needed for 20 real-world and 11 simulation task types; the other four simulation instructions already match the target definitions. Across both shot subsets this changes 62 task-table instruction rows and 3,131 episode task-text rows.
For real-world `task_index=1`, the instruction explicitly specifies five individual key presses in the order **2 → 8 → 2 → 6 → 8**, with a full release and return to the waiting position after every press. For `task_index=5`, it specifies **five presses of the blue button**, with the same release and return requirements. The other instructions use the definitions from the supplied 2026-10-23 metadata release package.
Only the existing task instruction field in the four `meta/tasks.…
5 days ago
Fix training examples and align task documentation with README checksums
Use recorded again to describe demonstrations collected again. Update only the corresponding README checksum entries; all training payload and provenance are preserved.
Align all four subset READMEs with the RoboValue dataset card
Summarize the four subset sizes, use paper task names without platform IDs, keep key state/action information visible, and fold extended field mappings and task descriptions. Training data and videos are unchanged.
10 days ago
Organize bilingual dataset into real-world and simulation subsets
Rename the real-world full-shot training directory, reserve the other three subset directories, document joint and end-effector action fields, and remove operational notes from current dataset documentation.
