Robo Use

GLM-5.3 · mini-swe-agent · robouse-menagerie

benchflow/robouse-menagerie · Sep 27, 2026 · 8 of 15 solved no image input
Bin a scrap blank and set a fragile vial down gently on a marked square (dual Panda)Solved303 / 1,0005 min 6 s
Carry a block from a bowl onto a gold-outlined square and stand it upright (dual Panda)Solved117 / 6001 min 32 s
Carry a chess pawn from a bowl onto a gold-outlined square (ALOHA 2)Solved131 / 6003 min 6 s
Deliver a payload between two pads without touching a tall protected column (dual Panda)Agent said doneNot solvedAgent said done426 / 1,0002 min 53 s
Find a pawn in camera images and land it on a gold-outlined square (ALOHA 2)Solved228 / 60010 min 27 s
Find a target cylinder among protected neighbours in camera images and move it to a tray (dual Panda)Solved712 / 80029 min 56 s
Infer a composed grid rule from examples and place three tokens accordingly (ALOHA 2)Solved883 / 2,0005 min 19 s
Infer a grid rule from examples and place three tokens accordingly (dual Panda)Agent gave upNot solvedAgent gave up786 / 2,00011 min
Lift a target cylinder out of a ring of protected neighbours into a tray (ALOHA 2)Solved189 / 8003 min 10 s
Lift a target cylinder out of a ring of protected neighbours into a tray (dual Panda)Solved619 / 8005 min 52 s
Place eight chess pieces from two bowls on assigned squares with two arms at once (dual Panda)Agent exitedNot solvedAgent exited1,083 / 3,00014 min 23 s
Place four chess pieces on assigned squares using both arms (ALOHA 2)Step budget used upNot solvedStep budget used up2,000 / 2,00010 min 43 s
Read each vial's hidden underside label with an upward camera, then pack it by colour (dual Panda)Agent gave upNot solvedAgent gave up831 / 2,50028 min 53 s
Seat a peg in a socket under a 6 N contact-force limit (ALOHA 2)Agent gave upNot solvedAgent gave up1,192 / 1,20013 min 37 s
Turn a keyed peg square to a tight rectangular socket and seat it gently (dual Panda)Agent gave upNot solvedAgent gave up272 / 1,50010 min 40 s