name: constraint-solver-arena
version: 1.0.0
description: Constraint Satisfaction Benchmark - Exposing LLM failures in multi-constraint planning and scheduling

environment:
  type: openenv
  action_space: ConstraintAction
  observation_space: ConstraintObservation
  
tasks:
  - name: meeting_scheduler
    description: Schedule meetings satisfying multiple participant time windows and room constraints
    difficulty: easy
    
  - name: resource_allocator
    description: Assign tasks to workers with skill requirements and availability constraints
    difficulty: medium
    
  - name: travel_planner
    description: Plan trips with budget, time, and dependency constraints across multiple activities
    difficulty: hard

metrics:
  - name: constraint_satisfaction
    description: Percentage of hard constraints satisfied (0.0 to 1.0)
    
  - name: solution_validity
    description: Whether the solution is valid (all hard constraints met)

tags:
  - constraint-satisfaction
  - planning
  - scheduling
  - reasoning
  - benchmark