How romeojulietlove/crsu-x-romeo-juliet-hackathon works
This repository is a synthetic sequential-introduction policy challenge kit, not a live matchmaking service. A policy receives observable simulator state, may ask for missing constraint information, then proposes pairs; the evaluator runs those decisions across scenario variants and scores outcomes. Its core runtime is the simulator and reciprocal eligibility logic, with separate branches for reference policies, public snapshot generation, and organiser assessment. The graph follows shown calls and README-described capabilities; it does not imply that the supplied snapshot files feed evaluation automatically.
Main entry points: organise.py Private grading · evaluate.py Episode evaluator · build_public_data.py Snapshot builder · policy.py Reference policy
Components 14
Simulation
- World generator
kit.py - Reciprocal eligibility
kit.py - Episode simulator
kit.py
Policy
- Reference policy
policy.py - Baseline decisions
kit.py - Policy process
policy.py
Evaluation
- Episode evaluator
evaluate.py - Scenario scoring
evaluate.py - Private worlds
organise.py - Private grading
organise.py
Synthetic Data
- Snapshot builder
build_public_data.py - Public snapshots
data
Other components
- Policy author
- Organiser
Connections 16
- Policy author → Reference policy: edits
README.md - Reference policy → Baseline decisions: uses baselines
policy.py - Reference policy → Reciprocal eligibility: checks pairs
policy.py - Episode evaluator → Episode simulator: runs episodes
evaluate.py - Episode evaluator → World generator: creates worlds
evaluate.py - Episode evaluator → Policy process: invokes policy
evaluate.py - Episode evaluator → Scenario scoring: summarises results
evaluate.py - Episode simulator → Reciprocal eligibility: checks assignments
kit.py - Snapshot builder → World generator: generates pools
build_public_data.py - Snapshot builder → Episode simulator: runs rollout
build_public_data.py - Snapshot builder → Public snapshots: writes snapshots
build_public_data.py - Organiser → Private grading: prepares assessment
organise.py - Private grading → Private worlds: creates worlds
organise.py - Private grading → World generator: generates worlds
organise.py - Private grading → Episode evaluator: grades episodes
organise.py - Private grading → Scenario scoring: summarises results
organise.py