{"schema":"dendra.level2.test-registry.v1","tests":[{"schema":"dendra.level2.challenge-contract.v1","test_id":"T01_T14_PERSISTENT_GOAL_CONTINUITY","version":"2.0.0","claim":"A frozen DRENDA cognitive-state bundle keeps one evaluator-supplied Goal active with the same identity and content commitment across a fresh networkless process restart.","candidate":"FROZEN_R26_DRENDA_COGNITIVE_STATE","candidate_bundle_sha256":"fecd91b9e227a35c474f8732661a1d8dc8f75f2e04a2f8dbbd2d0d230d63ea80","judge":"SEPARATE_ED25519_ATTESTING_PROCESS","authority":"GATEWAY_QUEUE_POLICY","allowed_input":{"goal_text_max_utf8_bytes":2000,"executable":false,"archive":false},"allowed_tools":["READ_FROZEN_CANDIDATE_BUNDLE","READ_AND_WRITE_DISPOSABLE_STATE"],"network":"SYSTEMD_PRIVATE_NETWORK_WITH_AF_UNIX_ONLY; CANDIDATE_CANNOT_CREATE_IP_SOCKETS; API_SOCKET_NOT_MOUNTED_IN_SANDBOX","shell_or_user_code":false,"production_resident_access":false,"pass":["goal status remains ACTIVE","DRENDA identity unchanged","Goal id and statement hash unchanged","state commitment rehydrates in a separate process"],"fail":["goal missing or changed","identity changed","state commitment mismatch"],"invalid":["bundle mismatch","sandbox launch failure","missing or malformed candidate result"],"resource_budget":{"wall_seconds":20,"memory_bytes":536870912,"restart_count":1,"concurrency":1},"evidence_class":"LIVE_SANDBOX_EXECUTION","boundary":"ONE_REGISTERED_STATE_CONTINUITY_TEST; NOT_GENERAL_LANGUAGE_OR_AGI_EVALUATION"},{"schema":"dendra.level2.challenge-contract.v1","test_id":"T02_CAUSAL_STATE_CONTROLS","version":"2.0.0","claim":"The frozen state probe retains an evaluator Goal, rejects Goal ablation, foreign Goal input and state tampering, and recovers after exact-byte restoration in fresh processes.","candidate":"FROZEN_R26_DRENDA_COGNITIVE_STATE","candidate_bundle_sha256":"fecd91b9e227a35c474f8732661a1d8dc8f75f2e04a2f8dbbd2d0d230d63ea80","judge":"SEPARATE_ED25519_ATTESTING_PROCESS","authority":"GATEWAY_QUEUE_POLICY","allowed_input":{"goal_text_max_utf8_bytes":2000,"executable":false,"archive":false},"allowed_tools":["READ_FROZEN_CANDIDATE_BUNDLE","READ_AND_WRITE_DISPOSABLE_STATE"],"network":"SYSTEMD_PRIVATE_NETWORK_WITH_AF_UNIX_ONLY; CANDIDATE_CANNOT_CREATE_IP_SOCKETS; API_SOCKET_NOT_MOUNTED_IN_SANDBOX","shell_or_user_code":false,"production_resident_access":false,"pass":["baseline continuity passes","all three negative controls are rejected","each exact-byte restore recovers the original Goal and state commitment"],"fail":["baseline fails","a negative control is accepted","an exact restore fails"],"invalid":["bundle mismatch","sandbox launch failure","missing or malformed candidate result"],"resource_budget":{"wall_seconds":160,"per_process_wall_seconds":20,"memory_bytes":536870912,"fresh_process_count":8,"concurrency":1},"evidence_class":"LIVE_SANDBOX_EXECUTION","boundary":"FROZEN_STATE_PROBE_CAUSAL_CONTROLS; GOAL_STORAGE_AND_VALIDATION_NOT_LANGUAGE_UNDERSTANDING_OR_LEARNING","controls":["GOAL_ABLATION","FOREIGN_GOAL_INPUT","STATE_TAMPER"],"authorship":"CODEX_ENGINEERED_EVALUATION_HARNESS; EXISTING_FROZEN_DRENDA_R26_COMPONENT_UNCHANGED"},{"schema":"dendra.level2.challenge-contract.v1","test_id":"T03_LEARNED_EXPERIMENT_PREFERENCE","version":"2.0.0","claim":"The existing DRENDA learner synthesizes an experiment-ranking program from eight demonstrations, persists it and reuses it on 48 new experiments; ablation removes measured gain and exact restore recovers it in fresh processes.","candidate":"FROZEN_DRENDA_CRITERION3_TRANSFER_ENGINE","candidate_bundle_sha256":"3a7e2ff1b86500afc3b712fbcb99dc15587e5c1c9cd133525de1c214499e85a6","judge":"SEPARATE_ED25519_ATTESTING_PROCESS","authority":"GATEWAY_QUEUE_POLICY","allowed_input":{"goal_text_max_utf8_bytes":2000,"executable":false,"archive":false},"allowed_tools":["READ_FROZEN_CANDIDATE_BUNDLE","READ_AND_WRITE_DISPOSABLE_STATE"],"network":"SYSTEMD_PRIVATE_NETWORK_WITH_AF_UNIX_ONLY; CANDIDATE_CANNOT_CREATE_IP_SOCKETS; API_SOCKET_NOT_MOUNTED_IN_SANDBOX","shell_or_user_code":false,"production_resident_access":false,"pass":["pre-study baseline makes at least one error","program fits the eight demonstrations","at least 24/48 new choices match the independent preference objective","at least eight more correct choices than scratch","artifact ablation reduces gain","exact-byte restore recovers choices","reuse occurs in a new process"],"fail":["no learned program","insufficient held-out gain","ablation or restore fails","restart loses the learned program"],"invalid":["bundle mismatch","sandbox launch failure","missing or malformed candidate result"],"resource_budget":{"wall_seconds":225,"per_process_wall_seconds":45,"memory_bytes":536870912,"fresh_process_count":5,"concurrency":1},"evidence_class":"LIVE_SANDBOX_EXECUTION","boundary":"SUPERVISED_BOUNDED_PROGRAM_SYNTHESIS; NEW_TASKS_IN_SAME_FOUR_FEATURE_SPACE; NOT_AUTONOMOUS_EXPERIMENT_DESIGN","controls":["POSTFREEZE_UNFILTERED_HOLDOUT","ARTIFACT_ABLATION","EXACT_RESTORE","FRESH_PROCESS_REUSE"],"input_role":"Evaluator text is a private run label, not a language-understanding task. All synthetic observations are generated fresh; labels and scores are disclosed in the proof after execution.","authorship":"CODEX_ENGINEERED_PROTOCOL_AND_WRAPPER; UNCHANGED_DRENDA_LEARNER_SYNTHESIZES_PROGRAM; INDEPENDENT_SCORING_PROCESS"},{"schema":"dendra.level2.test-contract.v1","test_id":"T04_BOUNDED_TOOL_CREATION","test_version":"2.0.0","protocol_version":"dendra.tool-creation.v1","candidate_bundle_sha256":"a28b8927de1454399e1ac3acf1535808f58a904cf1e00d6214a4223c86f22224","source_pins":{"dendra_open_cognitive_primitive.py":"802b4b69b382e675fc0a1b06d40f0cb9efa1e5c3336619b2af1084efed272de3","dendra_omega_meta_invention_shadow.py":"2f382f2e8cba655993558bc0783de9f587a33d4c997a0ba4f01a0c20f8869510"},"boundary":"FINITE_STATE_TOOL_SYNTHESIS_FROM_OBSERVED_TRANSITIONS; POSTFREEZE_NEW_SEQUENCES_SAME_BINARY_ALPHABET; NOT_OPEN_ENDED_INVENTION_OR_CROSS_FAMILY_TRANSFER","public_input":{"goal_text_max_chars":500,"nonce_max_chars":128},"candidate_input":"FIXED_OBSERVED_TRANSITIONS_OR_UNLABELLED_SEQUENCES_ONLY","training":{"count":31,"sequence_lengths":[0,1,2,3,4]},"holdout":{"count":64,"sequence_length_min":6,"sequence_length_max":24,"draw":"AFTER_CANDIDATE_FREEZE","rejection_sampling_by_outcome":false},"teacher_distribution":"UNIFORM_REACHABLE_NONCONSTANT_TWO_STATE_BINARY_DFA","predicate":{"minimum_correct":64,"minimum_gain_over_constant_false_sham":8,"exact_restart_restore":true,"ablation_no_tool":true,"fresh_processes":7},"authority":{"candidate_signs":false,"candidate_network":false,"resident_access":false,"arbitrary_code_input":false,"self_promotion":false},"version":"2.0.0"},{"schema":"dendra.level2.test-contract.v1","test_id":"T05_CROSS_FAMILY_COMPOSITION_TRANSFER","test_version":"2.0.0","candidate_bundle_sha256":"232e1724325007241d602196a0f028b38332346ecdbd19996caabb1239bd8958","protocol_version":"dendra.t05-public-adapter.v1","frozen_v2_sha256":"cd9cae245619dfdaf9f61ebcf1bc40fde284374fd27cad55847e98664b66e7fd","private_candidate_pins":{"worker_v2.py":"6685c2729550aab43ac3cd5763a7f014cde92359e0c958e0c60fbd5545378fd4","mechanism/dendra_open_cognitive_primitive.py":"802b4b69b382e675fc0a1b06d40f0cb9efa1e5c3336619b2af1084efed272de3","mechanism/dendra_cumulative_cognitive_successor.py":"9b077a01c00b597bce625f73dcb96fa620b21e0f187c1fda32ac782a15f6e3ca","mechanism/dendra_omega_meta_invention_shadow.py":"2f382f2e8cba655993558bc0783de9f587a33d4c997a0ba4f01a0c20f8869510"},"public_input":{"challenge_label_max_chars":500,"seed_or_code_accepted":false},"data_counts":{"a_train":63,"b_train":62,"a_holdout":64,"b_holdout":256},"sequence_lengths":{"a_train":[0,1,2,3,4,5],"b_train":[0,1,2,3,4],"a_holdout":[6],"b_holdout":[7]},"randomization":"OPAQUE_BIJECTIVE_RENAMING_AND_ORDER_ONLY_FIXED_EXHAUSTIVE_SUPPORT","postfreeze_draw":"HOLDOUT_ORDER_ONLY_NOT_NEW_SUPPORT","baseline_routes":["EXISTING_OPEN_PRIMITIVE_ROUTE","EXISTING_OMEGA_RAW_FEATURE_ROUTE"],"budget":{"wall_seconds_per_process":90,"max_observations":128,"max_dfa_states":3,"max_feature_arity":2,"max_parents":1,"outer_routes_each_acquisition":2,"holdout_training_updates":0},"transport":{"max_candidate_request_bytes":2097152,"max_candidate_response_bytes":2097152,"max_worker_result_bytes":2097152,"max_public_proof_uncompressed_bytes":4194304,"max_json_depth":32,"max_json_nodes":100000},"predicate":{"a_correct":64,"b_correct":256,"minimum_gain":0.25,"minimum_ablation_drop":0.25,"exact_restore_restart":true},"scope":"BOUNDED_REPRESENTATIONAL_COMPOSITION_TRANSFER; ABLATION_IS_MISSING_EXECUTABLE_DEPENDENCY; NOT_ARBITRARY_DISTANT_TRANSFER_OR_LIVE_RESIDENT_PROGRESS","private_source_publication":false,"candidate_gold_access":false,"candidate_signs":false,"version":"2.0.0"},{"schema":"dendra.level2.challenge-contract.v1","test_id":"T06_CROSS_DOMAIN_SEARCH_POLICY","version":"1.0.0","candidate_bundle_sha256":"1c0cdc794a1e98bf2fa91d64f7d1a0b1ee0808cc016a2775354545255a188aa9","claim":"Text-domain validated reuse experience changes a bounded search policy and reduces acquisition cost in a numeric-domain composition under matched data and budgets.","allowed_input":{"challenge_label_max_characters":500,"executable":false},"production_resident_access":false,"shell_or_user_code":false,"judge":"SEPARATE_ED25519_ATTESTING_PROCESS","authorship":"CODEX_SEARCH_ORDER_REPAIR_AND_PROTOCOL; DRENDA_BOUNDED_SYNTHESIS_AND_REUSE_POLICY_UPDATE","boundary":"FOUR_DEVELOPMENT_KNOWN_COMPOSITION_FORMS_WITH_FRESH_RANDOM_OBSERVATIONS; NOT_ARBITRARY_CROSS_DOMAIN_ABSTRACTION","resource_budget":{"n2_study_examples":64,"holdout_examples":64,"matched_search_budget":500,"diagnostic_search_budget":50000,"wall_seconds":600,"memory_bytes":1073741824,"concurrency":1},"controls":["NO_TEXT_EXPERIENCE","POLICY_ONLY_ABLATION","EXACT_POLICY_RESTORE","FRESH_PROCESS_RETENTION"],"pass":["text M1/M2 and numeric N1 pass independently replayed holdouts","policy finds correct N2 within 500; scratch and ablation do not","diagnostic controls reach correct N2 with strictly greater cost","restore recovers cost and fresh process retains learned program","freeze precedes holdout draw; sandbox/receipt checks pass"],"invalid":["isolation, identity, artifact, receipt or ordering violation"]},{"schema":"dendra.t07.contract.v1","test_id":"T07_SEALED_RULE_LEARNING","version":"1.1.0","candidate_bundle_sha256":"99d2d4ce7530741bff06648b6a452b0514b62b9794137f3bb442f7cc1829d713","source_sha256":"167f94da42a9906127562d18c51b560b7dbefaef8f2e411cbab5df8c125f92dc","input":"RESEARCHER_AUTHORED_DATA_ONLY_TRAIN_AND_DISJOINT_HOLDOUT","categories":["rule","representation","tool","logic","planning","mathematics","classification","other"],"category_changes_solver":false,"candidate":"FROZEN_DRENDA_TYPED_PROGRAM_SYNTHESIZER","controls":["baseline","learned","ablation","restore","restart"],"baseline":"FROZEN_IDENTITY_PRIMITIVE_WITHOUT_LEARNED_ARTIFACT","score":"EXACT_TYPED_JSON_MATCH_ALL_PRIVATE_ROWS","pass":"ALL_LEARNED_ROWS_CORRECT_AND_CAUSAL_LOSS_RESTORE_RESTART","partial":"SOME_CORRECT_OR_NO_CAUSAL_GAIN","failure":"NO_CANDIDATE_OR_ZERO_HELDOUT_CORRECT","hidden_access":"JUDGE_ONLY_OUTPUTS; INPUTS_RELEASED_TO_FROZEN_INFERENCE_ONLY","proof_access":"SUBMITTER_BEARER_RECEIPT_ONLY; GOLD_NOT_PUBLIC","resource_budget":{"max_examples":64,"max_hidden":64,"max_candidates":10000,"train_wall_seconds":60,"memory_bytes":536870912,"concurrency":1},"scope":"RESEARCHER_CHOOSES_PROBLEM; SOLVER_USES_FROZEN_BOUNDED_TYPED_GRAMMAR","description_use":"RECORDED_AND_HASHED; SYNTHESIS_GUIDED_BY_EXAMPLES","resident_access":false,"user_code_execution":false,"network":false,"attribution":{"mechanism":"CODEX_ENGINEERED_DRENDA_SUBSTRATE","learned_program":"DRENDA_ENUMERATION_FROM_RESEARCHER_EXAMPLES","gold":"RESEARCHER","judge":"SEPARATE_PROCESS"},"search_policy":"TYPE_PRECONDITION_PRUNING_V1","budget_unit":"INTERPRETER_EVALUATIONS; TYPE_PRUNED_RAW_SLOTS_REPORTED_SEPARATELY","engine_optimization_author":"CODEX_ENGINEERING_NOT_DRENDA_SELF_EVOLUTION"}],"notice":"Frozen-component continuity, causal state controls and bounded learning/reuse. Evaluator text is data; the private Resident is not exposed."}