Amazon FAR's StopWatch: Explicit Goal Grounding Brings Object-Goal Navigation Near-Expert Success

Amazon-FAR · hf · 2026-10-19

Amazon FAR (with NYU, UC Berkeley, CMU; advised by Pieter Abbeel) presents StopWatch, tackling the dominant failure mode in object-goal navigation.

• Explicit goal grounding, with minimal policy changes, yields near-expert success rates across vision-language navigation policies
• The key failure is the "last meter": baselines pass within 0.25 m of the goal (e.g., walking through an ottoman) and time out, while StopWatch stops correctly
• Across-task evaluations on HSSD-OV, MP3D-OV, HM3D-OV and more substantially beat OVON, LightNav, Un-NaVid and DagRL baselines; project page at stopwatch-nav.github.io

Original post →

More from Embodied

• Robot brain inference cost fell 84% in two months, from $2.36 to $0.38 per task — victor_explore · 2026-10-19
• NORINCO unveils Fuxi, an anti-riot humanoid robot that mirrors its operator's movements in real time — CyberRobooo · 2026-10-19
• HexaShell: VLM Coding Agents Drive Dual-Arm Robots, SOTA on RoboDojo — HexaFuture · 2026-10-19
• Has GPT-6-Astra Achieved Embodied Intelligence? A 12-Benchmark Evaluation — sensenova · 2026-10-19
• Robot can now recover its balance after a 0.1 N push — _Stocko_ · 2026-10-19
• Chinese blogger builds working electric hoverboard with 10 motors and real-time AI voice coaching — churchkey · 2026-10-19

Embodied channel →