{"data":{"kind":"file","path":"README.md","version_id":"pbenzqrjsi33xlewbm6v5d4k","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":3340,"modified_at":"2026-09-18T06:43:23.530000","content_hash":"82c6925aeed1a7bd7a2dd76e16333d5748096fd254d03614f415a5dde90238b6"},"entries":[],"content":"# Construction Management Evaluation Environment (v0.1.0)\n\nA high-fidelity benchmark and Reinforcement Learning (RL) verification environment built for the **Prime Intellect Environments Hub** using the `verifiers` framework.\n\n## Overview\n\nThe `construction-management-eval` environment evaluates LLMs and AI reasoning agents on real-world civil engineering, project controls, contract administration, and construction jobsite management tasks.\n\n### Core Evaluated Competencies\n\n1. **Cost Control & Earned Value Management (EVM)**:\n   - Schedule & Cost Variances: $CV = EV - AC$, $SV = EV - PV$\n   - Performance Indices: $CPI = EV / AC$, $SPI = EV / PV$\n   - Forecasting: Estimate at Completion ($EAC = BAC / CPI$), Estimate to Complete ($ETC$), Variance at Completion ($VAC$)\n2. **Schedule Optimization & Critical Path Method (CPM)**:\n   - Forward / Backward pass network analysis (ES, EF, LS, LF)\n   - Total Float ($TF$) and Free Float ($FF$) determination\n   - Schedule compression tradeoffs: Schedule Crashing vs. Fast-Tracking\n3. **Site Safety & OSHA 29 CFR 1926 Compliance**:\n   - Subpart P deep excavation and trenching standards\n   - Spoil pile setback boundaries (minimum 2 feet)\n   - Egress ladder lateral distance (maximum 25 feet)\n   - Soil classifications (Type A, B, C) and allowable sloping / benching ratios\n4. **Materials Quality Control & Structural Concrete (ACI 318 / ASTM)**:\n   - 28-day cylinder compressive strength test evaluation\n   - Mass concrete thermal differential controls (maximum 35°F / 19.4°C core vs surface delta)\n   - Delayed Ettringite Formation (DEF) mitigation and Supplementary Cementitious Materials (SCMs)\n5. **Contract Administration & Differing Site Conditions (AIA A201)**:\n   - Type I vs. Type II Differing Site Condition (DSC) claim substantiation\n   - Blue Book standby equipment rate quantification\n   - Time Impact Analysis (TIA) for compensable vs. excusable schedule extensions\n6. **VDC / BIM Spatial Clash Resolution**:\n   - Multi-trade coordination hierarchy (Structural > Gravity Drainage > HVAC Ductwork > Pressurized Piping)\n   - Constructability reviews and beam web penetration guidelines\n\n---\n\n## Rubrics & Multi-Signal Verification\n\nThe environment employs a 4-signal composite verification rubric:\n\n| Metric | Weight | Description |\n| :--- | :--- | :--- |\n| `metric_technical_calculation_accuracy` | **0.35** | Verifies exact mathematical solutions, formulas, and numeric thresholds. |\n| `metric_regulatory_and_standards_compliance` | **0.25** | Checks citations of industry codes (OSHA, ACI, ASTM, AIA, PMBOK). |\n| `metric_actionable_management_recommendations` | **0.20** | Rewards concrete, field-ready executive decisions and structured actions. |\n| `metric_comprehensive_reasoning_depth` | **0.20** | Assesses root-cause depth, derivation clarity, and professional rigor. |\n\n---\n\n## Installation & Usage\n\n### Local Installation\n\n```bash\ncd construction-management-eval\npip install -e .\n```\n\n### Loading in Python\n\n```python\nimport verifiers as vf\nimport construction_management_eval\n\nenv = construction_management_eval.load_environment()\nprint(f\"Loaded dataset with {len(env.dataset)} construction engineering challenges.\")\n```\n\n---\n\n## Publishing\n\nPublished to the Prime Intellect Environments Hub:\n```bash\nprime env push jeryhand/construction-management-eval -p .\n```\n","encoding":"utf-8","truncated":false,"total_bytes":3340},"status":null}