robotics
The Navigator Verified Most Steps Instead of Explaining Them
VerNav batches action checks and calls an autoregressive generator only when uncertainty rises.
Summary
VerNav batches action checks and calls an autoregressive generator only when uncertainty rises.
The verifier-first path replaces a full generated rationale at every navigation step with compact action verification, while an entropy trigger requests state evidence for uncertain decisions. On Room-to-Room, the verifier-only path remained competitive with representative language-model navigators and reduced average decision-stage model latency per step by more than tenfold. Results are benchmark-specific.
Why it matters
VerNav batches action checks and calls an autoregressive generator only when uncertainty rises.
Limits and context
- On Room-to-Room, the verifier-only path remained competitive with representative language-model navigators and reduced average decision-stage model latency per step by more than tenfold.
Key claims
VerNav batches action checks and calls an autoregressive generator only when uncertainty rises.
Qualification: On Room-to-Room, the verifier-only path remained competitive with representative language-model navigators and reduced average decision-stage model latency per step by more than tenfold.
Evidence: source-2026-09-02-012
Sources
- arXiv preprint 2609.00920arXiv · primary research
Corrections
No corrections have been recorded for this story.