Robo Use

GLM-5.3 · Codex · dexjoco

brave-eai/dexjoco · Sep 27, 2026 · 1 of 2 solved · 12 other trials not scored no image input

GLM-5.3 · Codex: the model's API rejected the camera images the harness attached, so its vision trials failed as infrastructure failures and are not scored (31 of its 43; the other 12 because the episode server did not start). Its scored trials saw no images.
Fold both temples of a pair of glasses and put them into a boxServer did not startInfra failureServer did not start1.7 s
Pick up a hammer with a dexterous hand and drive a nail with repeated strikesServer did not startInfra failureServer did not start1.7 s
Pick up a hammer with a dexterous hand and drive a nail with repeated strikes · second layoutServer did not startInfra failureServer did not start1.7 s
Pick up a trigger sprayer with a dexterous hand and water a plant without spraying elsewhereSolved3 min 38 s
Pick up a trigger sprayer with a dexterous hand and water a plant without spraying elsewhere · second layoutAgent gave upNot solvedAgent gave up28 min 30 s
Pick up kitchen tongs, hold them up and squeeze them shut three timesServer did not startInfra failureServer did not start1.7 s
Put a food box into a bucket, then lift the bucket by its wire handleServer did not startInfra failureServer did not start1.7 s
Put a food box into a bucket, then lift the bucket by its wire handle · second layoutServer did not startInfra failureServer did not start1.7 s
Slide a mouse onto its pad and click its left button with a dexterous handServer did not startInfra failureServer did not start2.1 s
Slide a mouse onto its pad and click its left button with a dexterous hand · second layoutServer did not startInfra failureServer did not start1.7 s
With two dexterous hands, aim a camera at a logo from a marked position and press the shutterServer did not startInfra failureServer did not start1.9 s
With two dexterous hands, insert an 8 mm peg into a socketServer did not startInfra failureServer did not start2.1 s
With two dexterous hands, open a microwave, put a hot dog in, close it and press startServer did not startInfra failureServer did not start1.9 s
With two dexterous hands, press an iPad's passcode 1-2-3 without touching a wrong digitServer did not startInfra failureServer did not start2.1 s