{ "agent_trace": { "codex_logs_available_local_only": true, "raw_logs_published": false, "summary_path": "agent-summary.json" }, "codex_version": "codex-cli 0.131.0", "collected_at": "2026-05-19T20:21:33.584758+00:00", "docker_cpus": 16, "docker_memory": "30g", "eval": { "error_code": null, "failed_tests": [ "eval.tests.test_15_final_massive_push.test_flag_combinations_matrix[True-True]", "eval.tests.test_remove_nodes.test_remove_nodes_all_filtered_out", "eval.tests.test_remove_nodes.test_remove_nodes_filters_selected_elements", "tests.test_cli_edge.test_multiple_attribute_flags_error", "tests.test_gaps.test_empty_href_attribute", "tests.test_cli_edge.test_pretty_print_formatting", "tests.test_gaps.test_pretty_print_with_comments", "tests.test_cli_edge.test_invalid_selector_panics", "tests.test_gaps.test_quadruple_slash_href", "tests.test_cli_edge.test_nonexistent_file_panics", "tests.test_harvest.test_remove_links", "tests.test_gaps.test_pretty_print_nested_inline_elements", "tests.test_gaps.test_non_link_element_with_href", "tests.test_cli_edge.test_binary_data_in_stdin", "tests.test_io.test_file_input_dash_means_stdin", "tests.test_integration.test_pretty_print_with_comments", "tests.test_integration.test_malformed_html_parsing", "tests.test_io.test_pretty_print_mode", "tests.test_output.test_pretty_print_indentation", "tests.test_integration.test_selector_matches_root_element", "tests.test_integration.test_quad_slash_url_handling", "tests.test_gaps.test_triple_slash_href", "tests.test_integration.test_navigation_links_with_base_detection", "tests.test_cli_edge.test_pretty_print_inline_elements", "tests.test_output.test_table_pretty_print", "tests.test_integration.test_selector_pretty_print", "tests.test_output.test_inline_elements_in_pretty_print", "tests.test_io.test_output_dash_means_stdout", "tests.test_output.test_section_element_pretty_print", "tests.test_output.test_list_with_block_children_pretty", "tests.test_output.test_inline_spans_pretty_print", "tests.test_output.test_head_elements_pretty_print", "tests.test_removal.test_elements_after_removed_appear", "tests.test_removal.test_multiple_divs_some_filtered", "tests.test_output.test_pretty_print_with_complex_nested_structure", "tests.test_removal.test_complex_css_selector_for_removal", "tests.test_cli_edge.test_four_slash_url_special_handling", "tests.test_integration.test_whitespace_only_input", "tests.test_removal.test_removal_with_pretty_print", "tests.test_output.test_script_style_elements_in_pretty_print", "tests.test_removal.test_attribute_selector_for_removal", "tests.test_removal.test_deeply_nested_removal", "tests.test_removal.test_invalid_removal_selector_ignored", "tests.test_output.test_void_elements_in_pretty_print", "tests.test_removal.test_large_document_with_removal_at_end", "tests.test_output.test_code_elements_pretty_print_inline", "tests.test_removal.test_element_containing_removable_node_filtered", "tests.test_selectors.test_universal_selector", "tests.test_removal.test_multiple_remove_selectors_with_clean_element_first", "tests.test_urls.test_basic_url_resolution_with_base", "tests.test_removal.test_removal_with_class_selector", "tests.test_removal.test_multiple_remove_flags_independently_specified", "tests.test_output.test_html_comments_in_pretty_print", "tests.test_removal.test_not_pseudo_class_in_removal", "tests.test_removal.test_unicode_content_with_removal", "tests.test_urls.test_empty_href_attribute_resolution", "tests.test_urls.test_four_slash_href_special_handling", "tests.test_removal.test_removal_with_from_file_input", "tests.test_removal.test_removal_preserves_element_order", "tests.test_urls.test_non_link_elements_not_rewritten", "tests.test_urls.test_non_link_elements_html_output_selective_rewriting", "tests.test_removal.test_nested_divs_inner_appears_when_outer_filtered", "tests.test_removal.test_adjacent_sibling_combinator_in_removal", "tests.test_output.test_deeply_nested_inline_elements_pretty", "tests.test_urls.test_unicode_paths_url_encoded", "eval.tests.test_argparse_validation.test_missing_value_for_value_options[args5]", "eval.tests.test_argparse_validation.test_missing_value_for_value_options[args6]", "eval.tests.test_argparse_validation.test_missing_value_for_value_options[args4]", "eval.tests.test_argparse_validation.test_double_dash_treats_following_as_positional", "eval.tests.test_argparse_validation.test_missing_value_for_value_options[args8]", "eval.tests.test_argparse_validation.test_missing_value_for_value_options[args7]", "eval.tests.test_argparse_validation.test_option_value_forms_attribute[args1]", "eval.tests.test_argparse_validation.test_missing_value_for_value_options[args3]", "eval.tests.test_argparse_validation.test_missing_value_for_value_options[args1]", "eval.tests.test_argparse_validation.test_missing_value_for_value_options[args2]", "eval.tests.test_argparse_validation.test_missing_value_for_value_options[args0]", "eval.tests.test_argparse_validation.test_combined_short_flags_are_accepted", "eval.tests.test_argparse_validation.test_missing_value_for_value_options[args9]", "eval.tests.test_argparse_validation.test_remove_nodes_requires_at_least_one_value_errors", "eval.tests.test_argparse_validation.test_output_file_with_equals_form_accepts_value" ], "failed_tests_omitted": 3, "public_eval_path": "eval.json", "summary_path": "eval-summary.json", "test_branch_errors": {}, "test_records": 2058, "warnings": [] }, "host_machine": "x86_64", "host_system": "Linux", "inference_mode": "paper-prompt-nointernet", "instance_id": "mgdm__htmlq.6e31bc8", "metrics": { "almost_resolved": "False", "cached_input_tokens": "2619776", "calls": "61", "codex_version": "codex-cli 0.131.0", "created_at": "2026-05-19T03:29:57.059116+00:00", "docker_cpus": "16", "docker_memory": "30g", "error_code": "", "estimated_cost_usd": "3.3700", "evaluator_problem": "False", "host_machine": "x86_64", "host_system": "Linux", "inference_mode": "paper-prompt-nointernet", "input_tokens": "2761748", "instance_id": "mgdm__htmlq.6e31bc8", "long_context_calls": "0", "model": "gpt-5.5", "n_resolved_tests": "1374", "n_system_errors": "0", "n_tests": "1455", "n_warnings": "0", "output_tokens": "45007", "paper_compliant": "False", "pricing_source": "https://developers.openai.com/api/docs/models/gpt-5.5", "reasoning_effort": "xhigh", "reasoning_output_tokens": "12212", "resolved": "False", "run_name": "cpx62-paperprompt-xhigh-shard-0-of-5-20260518T232006Z-mgdm-htmlq", "run_version": "20260518T232006Z", "score": "0.9443298969072165", "session_ended_at": "2026-05-19T03:42:21.197Z", "session_started_at": "2026-05-19T03:30:04.806Z", "test_branch_errors": "{}", "total_tokens": "2806755", "wall_clock_seconds": "736" }, "model": "gpt-5.5", "package": { "contents": [ ".", "./.git", "./.git/COMMIT_EDITMSG", "./.git/HEAD", "./.git/branches", "./.git/config", "./.git/description", "./.git/hooks", "./.git/hooks/applypatch-msg.sample", "./.git/hooks/commit-msg.sample", "./.git/hooks/fsmonitor-watchman.sample", "./.git/hooks/post-update.sample", "./.git/hooks/pre-applypatch.sample", "./.git/hooks/pre-commit.sample", "./.git/hooks/pre-merge-commit.sample", "./.git/hooks/pre-push.sample", "./.git/hooks/pre-rebase.sample", "./.git/hooks/pre-receive.sample", "./.git/hooks/prepare-commit-msg.sample", "./.git/hooks/push-to-checkout.sample", "./.git/hooks/update.sample", "./.git/index", "./.git/info", "./.git/info/exclude", "./.git/logs", "./.git/logs/HEAD", "./.git/logs/refs", "./.git/logs/refs/heads", "./.git/logs/refs/heads/master", "./.git/objects", "./.git/objects/04", "./.git/objects/04/89421fb753a9d4dd334aa026b709c8dc297573", "./.git/objects/43", "./.git/objects/43/804d736cbec3b43aac35e03f1851fd44c5e610", "./.git/objects/4c", "./.git/objects/4c/3de4dbaade52a92a13458d2fac866cbc7293e7", "./.git/objects/61", "./.git/objects/61/86c20d286556cd3b2728f55aca50cd110055e1", "./.git/objects/info", "./.git/objects/pack", "./.git/refs", "./.git/refs/heads", "./.git/refs/heads/master", "./.git/refs/tags", "./LICENSE.md", "./README.md", "./compile.sh", "./executable", "./htmlq.py" ], "contents_omitted": 0, "submission_available_local_only": true }, "paper_compliant": false, "paper_mode": true, "reasoning_effort": "xhigh", "run_name": "cpx62-paperprompt-xhigh-shard-0-of-5-20260518T232006Z-mgdm-htmlq", "run_version": "20260518T232006Z", "usage_audit": { "available_local_artifact": true, "public_path": "usage-audit.json" } }