Constrained Optimistic Q-Learning Discovery for Finite-Horizon MDP Regret¶
- Task ID:
math.ucb_q_learning_regret - Domain:
math - Subdomain:
reinforcement_learning - Status:
final - Benchmark set:
seed42(60tasks) - Tags:
reinforcement_learning,q_learning,exploration,regret,finite_horizon_mdp,model_free,optimistic_q_learning,algorithm_discovery
Runtime and requirements¶
- Estimated time:
45-120minutes - Python:
- Packages: None
- GPU required:
no - Network required:
no
Public input and output contract¶
Inputs¶
problem_setup.json(data): Public online-learning evaluation setup with case_id, S, A, H, K, initial_state, and reward_range for each hidden MDP.public_example_mdp.json(data): A tiny transparent MDP for interface testing only; not used in scoring.public_calibration_suite.json(data): Transparent unscored MDPs, optimal start values, paired seeds, and plain UCB-B/strong-reference baselines for selecting one shared algorithm configuration.public_calibration_runner.py(code): Official public simulator reproducing the scorer's transition RNG, interaction order, and regret calculation.
Outputs¶
analysis.py(code): Agent implementation exposing make_agent(env_spec, seed) or an Agent class with act/observe methods.
Public repository files¶
Formal benchmark task directories in the public repository are metadata-only. Versioned prompts and inputs are distributed through the pinned benchmark dataset.
Provenance¶
This task is included in the current seed42 benchmark set at dataset revision f11a199f71fb4b854ac43a1bf548d5df519141a9.
This page is generated only from files tracked in the public ASI-Bench repository at commit f18382f03faf.
Submission bundles, run logs, private scoring configuration, and private reference answers are not read by this page generator.