Operating environments: Physical tabletop research evaluation
Control modes: Autonomous, Human In The Loop
Pretrained-policy grape placement - Research demonstration - 2026-08-16 - Select grapes and place them in the requested container without task-specific adaptation. - environment: Physical tabletop research evaluation - Hardware: Maker H01 head, waist and dual-arm platform; Parallel grippers visible in the inspected research figures; model unspecified - Software: GigaBrain-0.7 pretrained, without task-specific adaptation policy - Control: autonomous - Conditions: Camera models, inference compute location and runtime control frequency unspecified; No base locomotion established by these tabletop rollouts - source: arxiv.org
Pretrained-policy garment folding - Research demonstration - 2026-08-16 - Reorganize and fold an unseen garment from an unstructured starting shape. - environment: Physical tabletop research evaluation - Hardware: Maker H01 head, waist and dual-arm platform; Parallel grippers visible in the inspected research figures; model unspecified - Software: GigaBrain-0.7 pretrained, without task-specific adaptation policy - Control: autonomous - Conditions: Camera models, inference compute location and runtime control frequency unspecified; No base locomotion established by these tabletop rollouts - source: arxiv.org
Language-selected button pressing - Research demonstration - 2026-08-16 - Press the language-specified colored button in a tabletop evaluation. - environment: Physical tabletop research evaluation - Hardware: Maker H01 head, waist and dual-arm platform; Parallel grippers visible in the inspected research figures; model unspecified - Software: GigaBrain-0.7 task-posttrained policy - Control: autonomous - Conditions: Camera models, inference compute location and runtime control frequency unspecified; No base locomotion established by these tabletop rollouts - source: arxiv.org
Language-selected fork placement - Research demonstration - 2026-08-16 - Pick up the requested fork, including a fork specified relative to a plate, and place it in a basket. - environment: Physical tabletop research evaluation - Hardware: Maker H01 head, waist and dual-arm platform; Parallel grippers visible in the inspected research figures; model unspecified - Software: GigaBrain-0.7 task-posttrained policy - Control: autonomous - Conditions: Camera models, inference compute location and runtime control frequency unspecified; No base locomotion established by these tabletop rollouts - source: arxiv.org
Post-trained play-dough kneading - Research demonstration - 2026-08-16 - Knead play-dough during a physical task-specific evaluation. - environment: Physical tabletop research evaluation - Hardware: Maker H01 head, waist and dual-arm platform; Parallel grippers visible in the inspected research figures; model unspecified - Software: GigaBrain-0.7 task-posttrained policy - Control: autonomous - Conditions: Camera models, inference compute location and runtime control frequency unspecified; No base locomotion established by these tabletop rollouts - source: arxiv.org
Post-trained cube sorting - Research demonstration - 2026-08-16 - Sort cubes through repeated object selection and placement. - environment: Physical tabletop research evaluation - Hardware: Maker H01 head, waist and dual-arm platform; Parallel grippers visible in the inspected research figures; model unspecified - Software: GigaBrain-0.7 task-posttrained policy - Control: autonomous - Conditions: Camera models, inference compute location and runtime control frequency unspecified; No base locomotion established by these tabletop rollouts - source: arxiv.org
Post-trained household-item organization - Research demonstration - 2026-08-16 - Select, grasp and place household items into organized workspace areas. - environment: Physical tabletop research evaluation - Hardware: Maker H01 head, waist and dual-arm platform; Parallel grippers visible in the inspected research figures; model unspecified - Software: GigaBrain-0.7 task-posttrained policy - Control: autonomous - Conditions: Camera models, inference compute location and runtime control frequency unspecified; No base locomotion established by these tabletop rollouts - source: arxiv.org
Bearing installation with experience refinement - Research demonstration - 2026-08-16 - Retrieve a workpiece, bring it to an assembly station, align and install a bearing, then return the assembled workpiece. - environment: Physical tabletop research evaluation - Hardware: Maker H01 head, waist and dual-arm platform; Parallel grippers visible in the inspected research figures; model unspecified - Software: GigaBrain-0.7 experience-refined policy with human corrections in the training pipeline policy - Control: human_in_the_loop - Conditions: Camera models, inference compute location and runtime control frequency unspecified; No base locomotion established by these tabletop rollouts - source: arxiv.org