Egocentric-100K Evaluation Dataset
Cross-dataset active labor and hand-presence evaluation
An Apache-2.0 benchmark comparing Egocentric-100K, Ego4D, and EPIC-KITCHENS samples for hand visibility, active manipulation, and active labor.
Comparing egocentric corpora before ingest
Open evaluation layer; the larger source datasets may still be gated.
Inspect schema and run a bounded sample audit before committing to the full release.
Has observation and action/state proxy, but weak task-phase context.
fit 70 · confidence 55
World-model signals have not been verified.
fit 50 · confidence 15
WAM-required observation, intent, and action alignment is not fully verified.
fit 50 · confidence 15
Verified facts and provenance
Claims, metadata verification, and sample verification are shown separately.
Unknown — no machine-readable schema facts have been captured.
Unknown — metadata conclusions do not prove sample coverage, alignment, or file integrity.
Declared loop signal coverage
Signals inferred from official metadata; Data pipeline verification is still pending.
Observation / ego video
images
Action / hand pose / robot state
hand visibility · active manipulation · Auditing industrial manipulation density · Cross-dataset active labor and hand-presence evaluation
Gaze / attention
No decision-grade evidence captured yet.
Language intent / task phase
No decision-grade evidence captured yet.
Feedback / correction / failure
active manipulation · Open evaluation layer; the larger source datasets may still be gated. · Cross-dataset active labor and hand-presence evaluation · An Apache-2.0 benchmark comparing Egocentric-100K, Ego4D, and EPIC-KITCHENS samples for hand visibility, active manipulation, and active labor.
Sim-real pairing
No decision-grade evidence captured yet.
License / format / access
Open · Apache-2.0 · Parquet · JSON metadata
Model and task fit · OpenBot inference
Has observation and action/state proxy, but weak task-phase context.
fit 70 · confidence 55
World-model signals have not been verified.
fit 50 · confidence 15
WAM-required observation, intent, and action alignment is not fully verified.
fit 50 · confidence 15
Has failure/evaluation-style labels with action or manipulation context.
fit 82 · confidence 55
Good tasks
Blockers and unresolved evidence
- Gaze / attentionunknownNot enough evidence to classify this signal. Verify metadata or a bounded sample.
- Language intent / task phaseunknownNot enough evidence to classify this signal. Verify metadata or a bounded sample.
- Sim-real pairingunknownNot enough evidence to classify this signal. Verify metadata or a bounded sample.
Raw dataset signals
OpenBot fit
- Comparing egocentric corpora before ingest
- Training data-efficiency filters
- Auditing industrial manipulation density
Integration notes
- Open evaluation layer; the larger source datasets may still be gated.
- Useful for OpenBot Data as a benchmark fixture.