Cartender: Verifiable Language-Conditioned Driving by Learning the Drives Not
Overview
Cartender is a research paper proposing a method for language-conditioned driving that learns from unexecuted actions. By constructing counterfactual branching from real recordings, the approach improves instruction success rates and RFS scores, achieving better performance than traditional imitation learning methods on the WOD-E2E benchmark.
From arXiv Robotics — research abstracts
Event evidence and corrections
0 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Reported quantity · mean RFS score: 8.203 other · Basis not reported
“f 8.642 across 3,603 feasible instruction conditions. Across 8,974 admitted proposals, mean RFS is 8.203 versus 7.906 for matched logged trajectories, and 67% score higher than their logged counterpart. Compared with matched-only imitation using the same backbone, Cartender improves same-scene paired instruction success from 23.67% to 57.00% and RFS from 6.32 to 7.10; its gate reduces infeasible-instruction execution”
Exact source · revision 2Source owner not reported
Reported quantity · RFS score for matched logged trajectories: 7.906 other · Basis not reported
“f 8.642 across 3,603 feasible instruction conditions. Across 8,974 admitted proposals, mean RFS is 8.203 versus 7.906 for matched logged trajectories, and 67% score higher than their logged counterpart. Compared with matched-only imitation using the same backbone, Cartender improves same-scene paired instruction success from 23.67% to 57.00% and RFS from 6.32 to 7.10; its gate reduces infeasible-instruction execution”
Exact source · revision 2Source owner not reported
Reported quantity · RFS score for instruction success rate: 6.32 other · Basis not reported
“f 8.642 across 3,603 feasible instruction conditions. Across 8,974 admitted proposals, mean RFS is 8.203 versus 7.906 for matched logged trajectories, and 67% score higher than their logged counterpart. Compared with matched-only imitation using the same backbone, Cartender improves same-scene paired instruction success from 23.67% to 57.00% and RFS from 6.32 to 7.10; its gate reduces infeasible-instruction execution”
Exact source · revision 2Source owner not reported
Evidence dependency changes (4)
- 2026-10-30T22:18:10.997Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T22:18:01.786Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-22T15:40:11.387Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-21T11:17:10.301Z · evidence updated · source revision 1. Evidence extraction was updated; current source attributions are shown above.
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsSignalCartender: Verifiable Language-Conditioned Driving by Learning the Drives Not Taken
Cartender is a research paper proposing a method for language-conditioned driving that learns from unexecuted actions. By constructing counterfactual branching from real recordings, the approach improves instruction success rates and RFS scores, achieving better performance than traditional imitation learning methods on the WOD-E2E benchmark.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).