{"data":{"kind":"file","path":"README.md","version_id":"fuw3xsx1rh4x2d6kngzrmasz","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":2163,"modified_at":"2026-08-28T13:18:55.546000","content_hash":"c450c7967b0ad9c3e2e32b3ad64953c5ceda0b1b69e2cf88c158ba62b6977272"},"entries":[],"content":"# PKB Second Brain Environment\n\nEvaluates LLM capability in Personal Knowledge Management (PKM) and Second Brain systems.\n\n## What It Tests\n\nThis environment measures how well an LLM can:\n\n1. **Design knowledge architectures** — propose PARA, Zettelkasten, or hybrid systems for specific use cases\n2. **Organize raw information** — categorize, tag, and link notes into coherent structures\n3. **Synthesize across sources** — connect ideas from disparate domains into unified summaries\n4. **Answer retrieval queries** — locate and surface relevant knowledge from a simulated knowledge base\n5. **Progressive summarization** — distill long content into layered summaries (key highlights → executive summary → full notes)\n6. **Maintain system hygiene** — detect orphan notes, broken links, duplicate content, and suggest merges\n\n## Task Types\n\n| Task | Description | Difficulty |\n|------|-------------|------------|\n| System Design | Design a PKB architecture for a given persona/need | Medium |\n| Note Organization | Categorize and link a batch of raw notes | Medium |\n| Cross-Domain Synthesis | Connect insights across 3+ unrelated sources | Hard |\n| Knowledge Retrieval | Answer questions from a structured knowledge base | Medium |\n| Progressive Summarization | Multi-layer summarization of long documents | Medium |\n| System Audit | Detect issues in an existing knowledge graph | Hard |\n\n## Reward Functions\n\n- **structure_quality**: Evaluates organizational coherence (hierarchy, tagging, linking)\n- **synthesis_depth**: Measures cross-domain connection quality\n- **retrieval_accuracy**: Tests answer correctness from knowledge base context\n- **summarization_fidelity**: Checks summary captures key ideas at each layer\n- **system_completeness**: Validates PKB design covers capture, organize, retrieve, review\n\n## Scoring\n\nAll scores normalized to [0.0, 1.0]. Composite score uses weighted average:\n- structure_quality: 0.25\n- synthesis_depth: 0.25\n- retrieval_accuracy: 0.20\n- summarization_fidelity: 0.15\n- system_completeness: 0.15\n\n## Usage\n\n```python\nimport verifiers as vf\nenv = vf.load_environment(\"pkb-second-brain\")\n```\n\n## License\n\nMIT\n","encoding":"utf-8","truncated":false,"total_bytes":2163},"status":null}