{"data":{"kind":"file","path":"README.md","version_id":"eiqdwlr5nydaisexl25t6djt","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":2005,"modified_at":"2026-08-07T12:26:17.888000","content_hash":"812e52ef7e167ebc186233c0da1014dc99120a86f121ae46943560f67bed47ee"},"entries":[],"content":"# Multi-Agent Collaboration\n\nEvaluates an LLM's ability to collaborate with multiple specialist agents on complex tasks.\n\n## What It Tests\n\n- **Task Decomposition**: Breaking complex projects into atomic sub-tasks\n- **Agent Delegation**: Assigning work to the right specialist\n- **Consensus Building**: Finding agreement between conflicting proposals\n- **Conflict Resolution**: Mediating between opposing architectural choices\n- **Coordinated Debugging**: Synthesizing partial diagnoses into root cause\n- **Scaling Strategy**: Integrating infrastructure, application, and data perspectives\n- **Code Review Collaboration**: Combining correctness, performance, and style feedback\n- **Migration Planning**: Coordinating legacy understanding, modern architecture, and risk\n\n## Environment Design\n\nMulti-turn environment (2 turns):\n1. **Turn 1**: Model produces a structured plan coordinating multiple agents\n2. **Turn 2**: Simulated agents provide feedback/concerns; model refines the plan\n\n## Scoring (5 rubric functions)\n\n| Weight | Metric | What It Measures |\n|--------|--------|------------------|\n| 0.30 | Collaboration | Agent references, delegation, dependencies, timeline |\n| 0.25 | Completeness | Trade-offs, risks, technical specifics, recommendations |\n| 0.20 | Structure | Sections, bullet points, organization |\n| 0.15 | Clarity | Formatting, consistency, no filler |\n| 0.10 | Length | Not too short (<80 words) or too long (>3000) |\n\n## Tasks (8 scenarios)\n\n1. Data Pipeline — coordinate Data/ML/DevOps engineers\n2. Conflict Resolution — mediate microservices vs monolith\n3. Consensus Building — hybrid database selection\n4. Task Delegation — healthcare portal with HIPAA\n5. Debug Collaboration — production outage investigation\n6. Scaling Strategy — 10x growth planning\n7. Code Review Collab — multi-reviewer synthesis\n8. Migration Planning — Django 2.7 to cloud-native\n\n## Usage\n\n```python\nimport verifiers as vf\nenv = vf.load_environment(\"yaya/multi-agent-collaboration\")\n```\n","encoding":"utf-8","truncated":false,"total_bytes":2005},"status":null}