FARL: A Fully Automated Pipeline of RL Interface and Solver Design

Abstract:Existing reinforcement learning (RL) largely focuses on optimizing learning algorithms within a pre-defined task formulation, while applying RL to new tasks still relies heavily on human expertise to design the learning pipeline: what an agent observes, how it is rewarded, which algorithm it uses, and how training is configured. Automating these choices is challenging because they are tightly coupled. We introduce FARL (Fully Automated Reinforcement Learning), a framework that automates RL design by jointly searching observation and reward programs, learning algorithms, and training settings. FARL uses training statistics and observed behavior to diagnose failures, propose changes, and test whether they improve learning. Its search preserves diverse candidates and the controls needed for these tests, using a staged budget to screen designs before investing in longer training. Experimental memory records each finding together with its conditions, so an unsuccessful component can be reconsidered in a different combination. Across expert-withheld and expert-available settings, FARL outperforms the evaluated automated baselines on challenging navigation, locomotion, and manipulation tasks. On Adroit Hand Relocate, FARL achieves 97.1% success at 5M training steps per policy, compared with 0% for Eureka and LiMEN and 1.0% for native Proximal Policy Optimization (PPO) under the same policy training budget.

43 pages, including references and appendices. Code and recorded experimental results are included as ancillary files

Artificial Intelligence (cs.AI)

arXiv:2610.15691 [cs.AI]

(or arXiv:2610.15691v1 [cs.AI] for this version)

https://doi.org/10.48550/arXiv.2610.15691 Focus to learn more arXiv-issued DOI via DataCite (pending registration)

FARL/CITATION.bib

FARL/CITATION.cff

FARL/MANIFEST.sha256

FARL/RELEASE_NOTES.md

FARL/THIRD_PARTY_NOTICES.md

FARL/autorl/__init__.py

FARL/autorl/ablation_execution.py

FARL/autorl/ablation_time_control.py

FARL/autorl/action_distribution_health.py

FARL/autorl/actuator_registry.py

FARL/autorl/adapter_source_semantics.py

FARL/autorl/adaptive_exploration_protocol.py

FARL/autorl/atom_grounding.py

FARL/autorl/bounded_training_protocol.py

FARL/autorl/breadth_portfolio.py

FARL/autorl/calibration.py

FARL/autorl/champion_recovery.py

FARL/autorl/closed_loop_audit.py

FARL/autorl/context_memory.py

FARL/autorl/coordinate_governor.py

FARL/autorl/core_ablation.py

FARL/autorl/coverage_diagnostics.py

FARL/autorl/data_types.py

FARL/autorl/derived_obs_wrapper.py

FARL/autorl/diagnostic_arbitrator.py

FARL/autorl/dim_role_hypotheses.py

FARL/autorl/dormancy.py

FARL/autorl/effective_breadth.py

FARL/autorl/env_builder.py

FARL/autorl/env_builder_v2.py

FARL/autorl/env_normalization.py

FARL/autorl/evidence_faults.py

FARL/autorl/evidence_leaf_repair.py

FARL/autorl/evidence_ledger.py

FARL/autorl/evidence_repair.py

FARL/autorl/evidence_validity.py

FARL/autorl/experience_pool.py

FARL/autorl/exploration_profiles.py

FARL/autorl/expression_sandbox.py

FARL/autorl/failure_taxonomy.py

FARL/autorl/frontier_exploration.py

FARL/autorl/full_control_evidence.py

FARL/autorl/global_falsifier.py

FARL/autorl/goal_omr_verifier.py

FARL/autorl/hypothesis_bank.py

FARL/autorl/hypothesis_memory.py

FARL/autorl/interface_coupling.py

FARL/autorl/intrinsic_exploration.py

FARL/autorl/joint_search_controller.py

FARL/autorl/known_env_adapter.py

FARL/autorl/known_env_policy.py

FARL/autorl/learner_execution_contract.py

FARL/autorl/limitation.py

FARL/autorl/llm/__init__.py

FARL/autorl/llm/availability.py

FARL/autorl/llm/client.py

FARL/autorl/m1_env_contract.py

FARL/autorl/meta_kb.py

FARL/autorl/model_router.py

FARL/autorl/modules/__init__.py

FARL/autorl/modules/algorithm_selector.py

FARL/autorl/modules/base.py

FARL/autorl/modules/diagnostics.py

FARL/autorl/modules/env_synthesis.py

FARL/autorl/modules/hpo.py

FARL/autorl/modules/input_resolver.py

FARL/autorl/modules/mdp_perception.py

FARL/autorl/modules/objective_metric_resolver.py

FARL/autorl/modules/obs_selection.py

FARL/autorl/modules/outcome_evaluator.py

FARL/autorl/modules/raw_sim_discoverer.py

FARL/autorl/modules/reward_designer.py

FARL/autorl/nl_task_contract.py

FARL/autorl/obs_space_contract.py

FARL/autorl/omr_contract.py

FARL/autorl/omr_trial_health.py

FARL/autorl/p2_public_spec.py

FARL/autorl/paper_success_metric.py

FARL/autorl/parallel_seed_cohort.py

FARL/autorl/pessimistic_commit.py

FARL/autorl/pipeline.py

FARL/autorl/portfolio.py

FARL/autorl/public_reward_program.py

FARL/autorl/replay.py

FARL/autorl/representation_ablation.py

FARL/autorl/representation_objective.py

FARL/autorl/representation_proposer.py

FARL/autorl/representation_shadow_eval.py

FARL/autorl/representation_targeted.py

FARL/autorl/repro.py

FARL/autorl/restart_request_reconciliation.py

FARL/autorl/reward_bandit.py

FARL/autorl/reward_execution_evidence.py

FARL/autorl/reward_program_compiler.py

FARL/autorl/runtime_information_policy.py

FARL/autorl/runtime_interface_contract.py

FARL/autorl/scientific_execution_contract.py

FARL/autorl/scientific_hparam_matching.py

FARL/autorl/search_rung_transition.py

FARL/autorl/seed_evidence.py

FARL/autorl/semantic_execution_contract.py

FARL/autorl/stack_champion.py

FARL/autorl/stage_evidence.py

FARL/autorl/stage_omr_contract.py

FARL/autorl/task_theory.py

FARL/autorl/training_health.py

FARL/autorl/trajectory_evidence/__init__.py

FARL/autorl/trajectory_evidence/behavior_analysis.py

FARL/autorl/trajectory_evidence/cohorts.py

FARL/autorl/trajectory_evidence/diagnosis_engine.py

FARL/autorl/trajectory_evidence/event_miner.py

FARL/autorl/trajectory_evidence/firewall.py

FARL/autorl/trajectory_evidence/goal_feedback.py

FARL/autorl/trajectory_evidence/hypothesis_ledger.py

FARL/autorl/trajectory_evidence/intervention_planner.py

FARL/autorl/trajectory_evidence/llm_evidence.py

FARL/autorl/trajectory_evidence/omr_health.py

FARL/autorl/trajectory_evidence/representation_replay.py

FARL/autorl/trajectory_evidence/reward_replay.py

FARL/autorl/trajectory_evidence/schemas.py

FARL/autorl/trajectory_evidence/stats.py

FARL/autorl/trajectory_evidence/store.py

FARL/autorl/trajectory_value_alignment.py

FARL/autorl/v2_provenance.py

FARL/autorl/v2_reporting.py

FARL/autorl/v2_runtime.py

FARL/autorl/v92_stage_loop.py

FARL/autorl/v93_unbounded_omr.py

FARL/autorl/v94_stage_diagnostics.py

FARL/autorl/video.py

FARL/config/ai_coding.yaml

FARL/config/core_ablations/ant_A1.yaml

FARL/config/core_ablations/ant_A2.yaml

FARL/config/core_ablations/ant_A3.yaml

FARL/config/core_ablations/doorkey_A1.yaml

FARL/config/core_ablations/doorkey_A2.yaml

FARL/config/core_ablations/doorkey_A3.yaml

FARL/config/core_ablations/first_batch.json

FARL/config/core_ablations/verification.json

FARL/config/core_full_controls/protocol_20260910.yaml

FARL/config/known_envs.yaml

FARL/config/llm.yaml

FARL/config/llm_v7_budget.yaml

FARL/config/meta_kb.json

FARL/config/pipeline.yaml

FARL/config/pipeline_a1.yaml

FARL/config/pipeline_a2.yaml

FARL/config/pipeline_a3.yaml

FARL/config/pipeline_a3_smoke.yaml

FARL/config/pipeline_a3_v5shadow.yaml

FARL/config/pipeline_ant_p3_v9_4.yaml

FARL/config/pipeline_games_p2.yaml

FARL/config/pipeline_games_p3.yaml

FARL/config/pipeline_metaworld_e2e.yaml

FARL/config/pipeline_metaworld_e2e_reach_p2_coldstart.yaml

FARL/config/pipeline_multigrid_e2e.yaml

FARL/config/pipeline_p1_benchmarks.yaml

FARL/config/pipeline_p1_smoke.yaml

FARL/config/pipeline_smoke.yaml

FARL/config/pipeline_smoke_grid.yaml

FARL/config/pipeline_v2_autonomous.yaml

FARL/config/pipeline_v4_shortrun_validation.yaml

FARL/contracts/__init__.py

FARL/contracts/schemas.py

FARL/evaluation/__init__.py

FARL/evaluation/baselines/ant_p3_support.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/cfg/config.yaml

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/cfg/env/ant.yaml

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/cfg/env/door.yaml

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/cfg/env/doorkey.yaml

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/cfg/env/hand.yaml

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/cfg/env/pick.yaml

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/cfg/env/reacher.yaml

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/cfg/env/walker.yaml

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/continuous_final.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/envs/adroit/hand_relocate_p1.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/envs/metaworld/sawyer_door_v3.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/envs/metaworld/sawyer_pick_place_v3.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/envs/metaworld/sawyer_reach_v3.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/envs/minigrid/doorkey.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/envs/mujoco/ant_v4.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/envs/mujoco/walker2d_v4.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/eureka_mujoco.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/p1_hand_bridge.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/p2_bridge.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/create_task.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/extract_task_code.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/file_utils.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/misc.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/code_feedback.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/code_output_tip.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/execution_error_feedback.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/initial_system_ant.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/initial_system_door.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/initial_system_doorkey.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/initial_system_hand.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/initial_system_pick.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/initial_system_reach.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/initial_system_walker.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/initial_user.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/paraphrase.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/policy_feedback_doorkey.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/policy_feedback_locomotion.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/policy_feedback_manipulation.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/reward_signature_ant.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/reward_signature_door.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/reward_signature_doorkey.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/reward_signature_hand.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/reward_signature_pick.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/reward_signature_reach.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/eureka/utils/prompts/reward_signature_walker.txt

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/rl/ant_native_reward.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/rl/train_mujoco.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/rl/walker_native_reward.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/rl/wrappers/ground_truth_reward.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/rl/wrappers/metaworld_to_gym.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/rl/wrappers/reward_override_wrapper.py

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/scripts/run_ant.sh

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/scripts/run_ant_p2.sh

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/scripts/run_ant_p3_formal.sh

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/scripts/run_door.sh

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/scripts/run_door_p2.sh

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/scripts/run_doorkey_p2.sh

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/scripts/run_hand_p1.sh

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/scripts/run_pick.sh

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/scripts/run_reach.sh

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/scripts/run_reach_p2.sh

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/scripts/run_walker.sh

FARL/evaluation/baselines/eurekaP3/Eureka_mujoco/task_interfaces.py

FARL/evaluation/baselines/p2_support.py

FARL/evaluation/baselines/run_with_framework_llm.py

FARL/evaluation/experiments/README.md

FARL/evaluation/experiments/adaptive_boundary_evidence.py

FARL/evaluation/experiments/adaptive_champion_final.py

FARL/evaluation/experiments/adaptive_dispatch_failure_restart.py

FARL/evaluation/experiments/adaptive_dispatch_source_upgrade.py

FARL/evaluation/experiments/adaptive_legacy_boundary_guard.py

FARL/evaluation/experiments/adaptive_restart_state.py

FARL/evaluation/experiments/adaptive_source_upgrade.py

FARL/evaluation/experiments/adaptive_unlaunched_schedule_recovery.py

FARL/evaluation/experiments/adroit_native_longtrain.py

FARL/evaluation/experiments/adroit_native_longtrain_launch.py

FARL/evaluation/experiments/adroit_relocate_native_qualification.py

FARL/evaluation/experiments/archive_eureka_ant_p2_matrix_20260726.py

FARL/evaluation/experiments/archive_eureka_ant_p3_matrix_20260727.py

FARL/evaluation/experiments/archive_eureka_doorkey_p3_run.py

FARL/evaluation/experiments/archive_eureka_p2_run.py

FARL/evaluation/experiments/archive_eureka_reach_p2_saturation.py

FARL/evaluation/experiments/archive_limen_reach_p2_saturation.py

FARL/evaluation/experiments/bounded_champion_final.py

FARL/evaluation/experiments/confirm_eureka_doorkey_p3_champion.py

FARL/evaluation/experiments/core_ablation_boundary_handoff.py

FARL/evaluation/experiments/core_ablation_cold_restart.py

FARL/evaluation/experiments/core_ablation_launch.py

FARL/evaluation/experiments/core_ablation_recover.py

FARL/evaluation/experiments/core_ablation_supervision.py

FARL/evaluation/experiments/core_ablation_time_launch.py

FARL/evaluation/experiments/core_ablation_time_resume.py

FARL/evaluation/experiments/core_ablation_validation.py

FARL/evaluation/experiments/core_full_control_launch.py

FARL/evaluation/experiments/door_core_ablation_launch.py

FARL/evaluation/experiments/door_core_ablation_monitor.py

FARL/evaluation/experiments/door_core_ablation_repair.py

FARL/evaluation/experiments/doorkey_p2_v92_champion_deep_train.py

FARL/evaluation/experiments/exp_p1_benchmark_full_info.py

FARL/evaluation/experiments/exp_p2_games_intent_rawsim.py

FARL/evaluation/experiments/exp_p2_mujoco_rawsim_full_auto.py

FARL/evaluation/experiments/exp_p2_multigrid_e2e.py

FARL/evaluation/experiments/exp_p3_ant_full_info.py

FARL/evaluation/experiments/exp_p3_games_full_info.py

FARL/evaluation/experiments/exp_p3_metaworld_full_auto.py

FARL/evaluation/experiments/exp_p3_phase1a_mujoco.py

FARL/evaluation/experiments/hand_baselines_5m.py

FARL/evaluation/experiments/hand_closed_loop_5m.py

FARL/evaluation/experiments/hand_cold_start.py

FARL/evaluation/experiments/hand_evidence_restart.py

FARL/evaluation/experiments/hand_five_wave_resume.py

FARL/evaluation/experiments/hand_p1_champion_warmstart.py

FARL/evaluation/experiments/kinetix_native_baseline.py

FARL/evaluation/experiments/monitor_v8_2_stageoracle.py

FARL/evaluation/experiments/p2_breadth_resume.py

FARL/evaluation/experiments/p2_champion_deep_train.py

FARL/evaluation/experiments/reach_champion_deep_train.py

FARL/evaluation/experiments/reach_v9_champions_deep_train.py

FARL/evaluation/experiments/rehearse_semantic_repair.py

FARL/evaluation/experiments/resume_adaptive_exploration.py

FARL/evaluation/experiments/resume_semantic_repair_boundary.py

FARL/evaluation/experiments/resume_v95_completed_boundary.py

FARL/evaluation/experiments/resume_v95_wave_boundary.py

FARL/evaluation/experiments/run_eureka_p2_background.sh

FARL/evaluation/experiments/run_metaworld_batch.sh

FARL/evaluation/experiments/run_mujoco_unfinished.sh

FARL/evaluation/experiments/run_pillar2_5env_single_seed.sh

FARL/evaluation/experiments/run_pillar2_5env_single_seed_supervisor.py

FARL/evaluation/experiments/snapshot_semantic_release.py

FARL/evaluation/experiments/specs/spec_pillar1_co_tsp.md

FARL/evaluation/experiments/specs/spec_pillar2_obs_selection.md

FARL/evaluation/experiments/specs/spec_pillar3_breadth_verification.md

FARL/evaluation/experiments/specs/spec_pillar3_phase1a_mujoco_pilot.md

FARL/evaluation/experiments/summarize_breadth.py

FARL/evaluation/experiments/update_ant_1k_displacement_xlsx.py

FARL/evaluation/experiments/update_eureka_reach_p2_archive_xlsx.py

FARL/evaluation/experiments/update_frozen_five_eureka_doorkey_p3_xlsx.py

FARL/evaluation/metrics/README.md

FARL/evaluation/metrics/__init__.py

FARL/evaluation/metrics/bootstrap_ci.py

FARL/evaluation/metrics/field_selection_pr.py

FARL/evaluation/metrics/granger_causality.py

FARL/evaluation/metrics/iqm.py

FARL/evaluation/metrics/mutual_information.py

FARL/evaluation/metrics/significance.py

FARL/paper/data/aspire_results.json

FARL/paper/data/hand_baseline_pre5m_profiles.json

FARL/paper/data/hand_baselines_5m_results.json

FARL/paper/data/hand_champion_stage_profiles.json

FARL/paper/data/hand_early_training_curves.json

FARL/paper/data/v12_results_20260914.json

FARL/paper/data/v13_native_ant_results.json

FARL/paper/data/v14_learning_data.json

FARL/paper/data/v15_evidence.json

FARL/paper/data/v16_hand_early_search_20260923.json

FARL/paper/data/v16_hand_results.json

FARL/paper/figures/fig_additional_search.py

FARL/paper/figures/fig_v15_candidate_learning.py

FARL/paper/figures/fig_v16_ant_designs.py

FARL/paper/figures/fig_v16_hand_learning.py

FARL/paper/figures/fig_v16_main.py

FARL/paper/figures/style.py

FARL/reproduction/baseline_configs/limen_adroit_hand_relocate_p1.yaml

FARL/reproduction/baseline_configs/limen_ant_p2.yaml

FARL/reproduction/baseline_configs/limen_door_p2.yaml

FARL/reproduction/baseline_configs/limen_doorkey_p3.yaml

FARL/reproduction/baseline_configs/limen_minigrid_doorkey.yaml

FARL/reproduction/baseline_configs/limen_mujoco_ant.yaml

FARL/reproduction/costs/ant_token_estimate.json

FARL/reproduction/environment.json

FARL/reproduction/hand_5m/eureka/seed_0/candidate_final_eval.v1.json

FARL/reproduction/hand_5m/eureka/seed_0/training_receipt.json

FARL/reproduction/hand_5m/eureka/seed_0/training_settings.json

FARL/reproduction/hand_5m/eureka/seed_1/candidate_final_eval.v1.json

FARL/reproduction/hand_5m/eureka/seed_1/training_receipt.json

FARL/reproduction/hand_5m/eureka/seed_1/training_settings.json

FARL/reproduction/hand_5m/eureka/seed_2/candidate_final_eval.v1.json

FARL/reproduction/hand_5m/eureka/seed_2/training_receipt.json

FARL/reproduction/hand_5m/eureka/seed_2/training_settings.json

FARL/reproduction/hand_5m/eureka/seed_3/candidate_final_eval.v1.json

FARL/reproduction/hand_5m/eureka/seed_3/training_receipt.json

FARL/reproduction/hand_5m/eureka/seed_3/training_settings.json

FARL/reproduction/hand_5m/eureka/seed_4/candidate_final_eval.v1.json

FARL/reproduction/hand_5m/eureka/seed_4/training_receipt.json

FARL/reproduction/hand_5m/eureka/seed_4/training_settings.json

FARL/reproduction/hand_5m/limen/seed_42/metrics.json

FARL/reproduction/hand_5m/limen/seed_42/runtime_contract.json

FARL/reproduction/hand_5m/limen/seed_42/training_receipt.json

FARL/reproduction/hand_5m/limen/seed_42/training_settings.json

FARL/reproduction/hand_5m/limen/seed_43/metrics.json

FARL/reproduction/hand_5m/limen/seed_43/runtime_contract.json

FARL/reproduction/hand_5m/limen/seed_43/training_receipt.json

FARL/reproduction/hand_5m/limen/seed_43/training_settings.json

FARL/reproduction/hand_5m/limen/seed_44/metrics.json

FARL/reproduction/hand_5m/limen/seed_44/runtime_contract.json

FARL/reproduction/hand_5m/limen/seed_44/training_receipt.json

FARL/reproduction/hand_5m/limen/seed_44/training_settings.json

FARL/reproduction/hand_5m/manifest.json

FARL/reproduction/hand_5m/summary.json

FARL/reproduction/selected_designs/ant_ea/algorithm_config.json

FARL/reproduction/selected_designs/ant_ea/env_adapter.py

FARL/reproduction/selected_designs/ant_ea/exploration.json

FARL/reproduction/selected_designs/ant_ea/obs_schema.json

FARL/reproduction/selected_designs/ant_ea/reward.json

FARL/reproduction/selected_designs/ant_ea/saved_model_settings.json

FARL/reproduction/selected_designs/ant_ea/worker_summary.json

FARL/reproduction/selected_designs/ant_ew/algorithm_config.json

FARL/reproduction/selected_designs/ant_ew/env_adapter.py

FARL/reproduction/selected_designs/ant_ew/exploration.json

FARL/reproduction/selected_designs/ant_ew/obs_schema.json

FARL/reproduction/selected_designs/ant_ew/reward.json

FARL/reproduction/selected_designs/ant_ew/saved_model_settings.json

FARL/reproduction/selected_designs/ant_ew/worker_summary.json

FARL/reproduction/selected_designs/door_ew/algorithm_config.json

FARL/reproduction/selected_designs/door_ew/env_adapter.py

FARL/reproduction/selected_designs/door_ew/exploration.json

FARL/reproduction/selected_designs/door_ew/obs_schema.json

FARL/reproduction/selected_designs/door_ew/reward.json

FARL/reproduction/selected_designs/door_ew/worker_summary.json

FARL/reproduction/selected_designs/doorkey_ea/exploration.json

FARL/reproduction/selected_designs/doorkey_ea/obs_schema.json

FARL/reproduction/selected_designs/doorkey_ea/reward.json

FARL/reproduction/selected_designs/doorkey_ea/selection_receipt.json

FARL/reproduction/selected_designs/doorkey_ew/algorithm_config.json

FARL/reproduction/selected_designs/doorkey_ew/env_adapter.py

FARL/reproduction/selected_designs/doorkey_ew/exploration.json

FARL/reproduction/selected_designs/doorkey_ew/obs_schema.json

FARL/reproduction/selected_designs/doorkey_ew/reward.json

FARL/reproduction/selected_designs/doorkey_ew/worker_summary.json

FARL/reproduction/selected_designs/eureka_ant_ea/program.py

FARL/reproduction/selected_designs/eureka_ant_ew/program.py

FARL/reproduction/selected_designs/eureka_door0/program.py

FARL/reproduction/selected_designs/eureka_door1/program.py

FARL/reproduction/selected_designs/eureka_door4/program.py

FARL/reproduction/selected_designs/eureka_doorkey_ea/program.py

FARL/reproduction/selected_designs/eureka_doorkey_ew/program.py

FARL/reproduction/selected_designs/hand_baselines/eureka_reward.py

FARL/reproduction/selected_designs/hand_baselines/limen_config.yaml

FARL/reproduction/selected_designs/hand_baselines/limen_program.json

FARL/reproduction/selected_designs/hand_st/algorithm_config.json

FARL/reproduction/selected_designs/hand_st/env_adapter.py

FARL/reproduction/selected_designs/hand_st/exploration.json

FARL/reproduction/selected_designs/hand_st/obs_schema.json

FARL/reproduction/selected_designs/hand_st/reward.json

FARL/reproduction/selected_designs/hand_st/runtime_interface_contract.v1.json

FARL/reproduction/selected_designs/hand_st/saved_model_settings.json

FARL/reproduction/selected_designs/hand_st/worker_summary.json

FARL/reproduction/selected_designs/limen_ant_ea/program.py

FARL/reproduction/selected_designs/limen_ant_ew/interface.py

FARL/reproduction/selected_designs/limen_door_ew/interface.py

FARL/reproduction/selected_designs/limen_door_ew/metrics.json

FARL/reproduction/selected_designs/limen_doorkey_ea/interface.py

FARL/reproduction/selected_designs/limen_doorkey_ea/metrics.json

FARL/reproduction/selected_designs/limen_doorkey_ew/interface.py

FARL/reproduction/selected_designs/limen_hand_st/interface.py

FARL/reproduction/selected_designs/limen_hand_st/metrics.json

FARL/reproduction/summarize_results.py

FARL/reproduction/verify_package.py

FARL/requirements.txt

FARL/tests/fixtures/hand_iter17_generation_reset.json

FARL/tests/fixtures/p2_doorkey_eureka_smoke.py

FARL/tests/fixtures/p2_doorkey_limen_smoke.py

FARL/tests/test_action_distribution_health.py

FARL/tests/test_adaptive_boundary_evidence.py

FARL/tests/test_adaptive_boundary_pause.py

FARL/tests/test_adaptive_candidate_archive.py

FARL/tests/test_adaptive_champion_final.py

FARL/tests/test_adaptive_dispatch_failure_restart.py

FARL/tests/test_adaptive_dispatch_resume_entrypoint.py

FARL/tests/test_adaptive_dispatch_source_upgrade.py

FARL/tests/test_adaptive_exploration_protocol.py

FARL/tests/test_adaptive_five_wave_limit.py

FARL/tests/test_adaptive_information_reaudit.py

FARL/tests/test_adaptive_interrupted_restart.py

FARL/tests/test_adaptive_legacy_boundary_guard.py

FARL/tests/test_adaptive_restart_state_preservation.py

FARL/tests/test_adaptive_resume_boundary.py

FARL/tests/test_adaptive_source_upgrade.py

FARL/tests/test_adaptive_unlaunched_schedule_recovery.py

FARL/tests/test_adroit_native_longtrain.py

FARL/tests/test_adroit_native_longtrain_launch.py

FARL/tests/test_ant_1k_displacement_contract.py

FARL/tests/test_ant_p3_full_info.py

FARL/tests/test_bounded_training_protocol.py

FARL/tests/test_collaborator_baselines_p2.py

FARL/tests/test_completed_boundary_cleanup.py

FARL/tests/test_completed_boundary_resource_refresh.py

FARL/tests/test_core_ablation_a3_audit.py

FARL/tests/test_core_ablation_adoption.py

FARL/tests/test_core_ablation_basic_diagnosis_robustness.py

FARL/tests/test_core_ablation_boundaries.py

FARL/tests/test_core_ablation_boundary_handoff.py

FARL/tests/test_core_ablation_censored_source.py

FARL/tests/test_core_ablation_deployment.py

FARL/tests/test_core_ablation_door_prompt_boundaries.py

FARL/tests/test_core_ablation_final_budget.py

FARL/tests/test_core_ablation_finalization_recovery.py

FARL/tests/test_core_ablation_supervision.py

FARL/tests/test_core_ablation_time_budget.py

FARL/tests/test_core_ablation_time_controller.py

FARL/tests/test_core_ablation_time_launch.py

FARL/tests/test_core_ablation_time_resume.py

FARL/tests/test_core_ablation_time_status.py

FARL/tests/test_core_ablation_time_validation.py

FARL/tests/test_core_ablation_validation.py

FARL/tests/test_core_ablation_validation_recovery.py

FARL/tests/test_core_ablation_zero_winner.py

FARL/tests/test_core_full_control_budget.py

FARL/tests/test_core_full_control_evidence.py

FARL/tests/test_core_full_control_launch.py

FARL/tests/test_dim_role_hypotheses.py

FARL/tests/test_door_core_ablation_launch.py

FARL/tests/test_door_core_ablation_monitor.py

FARL/tests/test_door_core_ablation_repair.py

FARL/tests/test_door_core_ablation_runtime.py

FARL/tests/test_doorkey_p2_v92_deep_train.py

FARL/tests/test_effective_obs_contract.py

FARL/tests/test_eureka_hand_p1.py

FARL/tests/test_generation_reset_request_boundary.py

FARL/tests/test_generic_evidence_leaf_repair.py

FARL/tests/test_generic_evidence_repair.py

FARL/tests/test_generic_frontier_exploration_closure.py

FARL/tests/test_generic_pending_controller_execution.py

FARL/tests/test_generic_scientific_request_admission.py

FARL/tests/test_generic_stage_evidence_semantics.py

FARL/tests/test_generic_stage_frontier_publication.py

FARL/tests/test_hand_closed_loop_5m.py

FARL/tests/test_hand_cold_start.py

FARL/tests/test_hand_controlled_closure_repair.py

FARL/tests/test_hand_five_wave_resume.py

FARL/tests/test_health_repair_controller_integration.py

FARL/tests/test_hpo_algorithm_typed_dispatch.py

FARL/tests/test_hpo_gpu_token_allocation.py

FARL/tests/test_hpo_source_bound_narrowing.py

FARL/tests/test_information_fault_lifecycle.py

FARL/tests/test_learner_execution_health.py

FARL/tests/test_limen_ant_p2.py

FARL/tests/test_limen_door_p2_doorkey_p3.py

FARL/tests/test_limen_hand_p1.py

FARL/tests/test_limen_reach_p2.py

FARL/tests/test_llm_availability_recovery.py

FARL/tests/test_llm_diagnostic_cost_guard.py

FARL/tests/test_llm_stream_transport.py

FARL/tests/test_m2_canonical_input_boundary.py

FARL/tests/test_matched_frontier_initialization.py

FARL/tests/test_optimizer_step_counter.py

FARL/tests/test_p1_adroit_stage_oracle.py

FARL/tests/test_p2_breadth_resume.py

FARL/tests/test_p2_champion_deep_train.py

FARL/tests/test_p3_doorkey_stage_oracle.py

FARL/tests/test_p3_privileged_reward_admission.py

FARL/tests/test_paper_success_metric.py

FARL/tests/test_pending_scientific_request_lifecycle.py

FARL/tests/test_restart_request_reconciliation.py

FARL/tests/test_resume_packet_authority.py

FARL/tests/test_resume_typed_records.py

FARL/tests/test_reward_execution_evidence.py

FARL/tests/test_scientific_execution_contract.py

FARL/tests/test_scientific_typed_matching.py

FARL/tests/test_semantic_execution_closure.py

FARL/tests/test_semantic_repair_resume.py

FARL/tests/test_source_bound_reward_execution.py

FARL/tests/test_stage_frontier_handoff.py

FARL/tests/test_stage_prerequisite_monotonicity.py

FARL/tests/test_stratified_trajectory_retention.py

FARL/tests/test_trajectory_diagnostic_sampling.py

FARL/tests/test_v84_event_reward_compiler.py

FARL/tests/test_v84_oracle_label_plane.py

FARL/tests/test_v84_scientific_loop.py

FARL/tests/test_v84_trusted_stage_contract.py

FARL/tests/test_v92_stage_closed_loop.py

FARL/tests/test_v93_unbounded_dense_omr.py

FARL/tests/test_v94_stage_learning_closed_loop.py

FARL/tests/test_v95_depth_phase_promotion.py

FARL/tests/test_v95_execution_closed_loop.py

FARL/tests/test_v95_runtime_interface_closure.py

FARL/tests/test_v96_evidence_validity_closure.py

FARL/tests/test_v96_interface_generation_transaction.py

FARL/tests/test_v96_observation_runtime_interface.py

FARL/tests/test_v96_parallel_interface_generation_fence.py

FARL/tests/test_v96_reward_runtime_interface.py

FARL/tests/test_v96_runtime_information_policy.py

FARL/tests/test_v96_runtime_information_regime.py

FARL/tests/test_v9_breadth_screening.py

FARL/tests/test_v9_general_reward_program.py

FARL/tests/test_v9_joint_search_controller.py

FARL/tests/test_v9_public_reward_pipeline.py

(554 additional files not shown) You must enabled JavaScript to view entire file list.

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.