Skip to content

Watch & compare Benchmarks Models

Rather

Loading benchmark results…

rather / World model benchmarks

Sources & attribution Compare benchmark scoresCuriosity first. Evidence always.

WBench / Navigation / Evidence

gd_world

Source-listed configuration · gd_cv

Current report

Reported scores

Average: 81.20points / 100
Quality: 82.20points / 100
Setting: 79.50points / 100
Interaction: 85.30points / 100
Consistency: 89.50points / 100
Physical: 69.60points / 100

Read the original leaderboard ↗ (opens in a new tab)

Provenance

Result origin: Published by WBench; not run by Rather
Evaluation source: Organizer evaluated
Source publication date: Oct 12, 2026
Imported: Oct 13, 8:38 AM UTC
Protocol: Navigation · 158 cases
Checkpoint: Not pinned by source

Generated by: Not established by this leaderboard row
Scored by: organizer-evaluation

Turn boundaries follow submitted turns.json; original inference sidecars were not supplied for independent reconstruction verification.

Organizer evaluation

Source revision & integrity

Upstream revision: 004aa621b4e526f733b3e5cc8305ee241f04048f
Snapshot SHA-256: 14f13b9229d883ed8c5a7b5d9d20759333e5838e226204b0f7a19d422604ac20
Original label: gd_cv | gd_world |