{"data":{"kind":"file","path":"README.md","version_id":"pcoffry6e271u38eqsu90j6a","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":1801,"modified_at":"2026-09-12T12:57:06.854000","content_hash":"3aecc18814a389b0d10794629fdb9b76cc53175b83c8a18d3b2daa3fe0fe8b44"},"entries":[],"content":"# Robot Gesture Training Environment\n\nA Verifiers environment for training and evaluating LLMs on robotic gesture planning and manipulation tasks.\n\n## Overview\n\nThis environment challenges models to:\n\n1. **Generate** robot gesture sequences for manipulation tasks (pick-and-place, assembly, handover)\n2. **Analyze** existing gesture plans for safety, efficiency, and feasibility\n3. **Debug** failed gesture executions by identifying error sources\n4. **Optimize** gesture trajectories for energy efficiency and collision avoidance\n5. **Adapt** gestures to different robot morphologies (6-DOF arm, humanoid hand, gripper)\n\n## Task Types\n\n| Task | Description | Difficulty |\n|------|-------------|------------|\n| Pick-and-Place | Generate grasp + lift + move + place sequences | Easy |\n| Assembly | Insert peg, screw thread, snap-fit with precision | Medium |\n| Handover | Human-robot object handover with safety zones | Medium |\n| Bimanual | Coordinated two-arm manipulation | Hard |\n| Error Recovery | Diagnose and fix failed gesture sequences | Hard |\n\n## Scoring\n\nThe rubric evaluates on 5 dimensions:\n\n- **Correctness** (0.30): Are joint angles, waypoints, and frames physically valid?\n- **Safety** (0.25): Collision avoidance, joint limits, force thresholds respected?\n- **Efficiency** (0.20): Minimal trajectory length, smooth velocity profiles?\n- **Completeness** (0.15): All required steps present, proper approach/retreat?\n- **Adaptability** (0.10): Handles constraints, different morphologies?\n\n## Usage\n\n```python\nimport verifiers as vf\n\nenv = vf.load_environment(\"robot-gesture-training\")\n# or\nfrom robot_gesture_training import load_environment\nenv = load_environment(level=1)\n```\n\n## Evaluation\n\n```bash\nprime eval run leminho/robot-gesture-training -m gpt-4o-mini\n```\n\n## Version\n\nv1.0.0\n","encoding":"utf-8","truncated":false,"total_bytes":1801},"status":null}