{ "model_revision": "e5e0406c85ec241b164907bf41ebd648a54e4ea2", "model_sha256": "81dd7609c589f3694ce75520c25af5010953231358de6aec4d03e8c6aac787e9", "wheel": { "file": "nemotron_parse-3.5.0.post2-py3-none-any.whl", "sha256": "63d6fe932dfc3cab5d029086f3c7c0ac7f75fad6f6f8c08799a60c54bfa12bdb" }, "dataset": { "repo": "llamaindex/ParseBench", "revision": "2805a1d940f95a203e0ae4b88be9934f7765b3fc", "pages": 2078, "renderer": "PyMuPDF==1.26.3", "dpi": 150, "jpeg_quality": 92, "source_page_index": 0 }, "public_input_parity": { "pages": 2078, "image_mismatches": [], "annotation_mismatches": [], "public_dataset_revision": "2805a1d940f95a203e0ae4b88be9934f7765b3fc", "renderer": "PyMuPDF==1.26.3", "dpi": 150, "jpeg_quality": 92, "source_page_index": 0 }, "scorer_revision": "9af4102e26b51f234ab5f037e19fec05c0696ec6", "executed_python_source_hashes": { "serve.py": "b5a772c9108f9ac9acffb42a872bf04694e43c0cb10416ba2bb3d5572b970f5b", "score.py": "a566accf056987c90201b6bc75e4ecf02eb56cf51a1e65fc3b619c108aa0b60e", "postprocess.py": "576db3eb97c5e43ebbb1b5cdd11aef212f7728fca0b4bfa20fc4d2f309e8eb5f", "prepare_data.py": "68d87dfac34335ab4ad34b6d7b0d04c7973d49b92afec3d9e167bf7ea24d0010", "prepare_inputs.py": "f3c0354e8559e42678d43a9071b5ba4f654c7bcd13426493913b622fa4d96795", "run.py": "116a08d75f620f7e4ccd4c9350e1cffc863b04c2c9a14e147815776b572964fa", "engine/markdown.py": "fc57566255ac632b24825973d9533a13fd717dad4e38d096098bac33ecc79957", "engine/merge.py": "b1c2e749e8f772b716c5a8a2ebeba7348474eb9368b00d37f1c8987e26d24e4e", "engine/caption_relocation.py": "3a17677b1a43c793cd817f8f3f5745be807203ff75b8e3e0af5a4d51d547dd4b", "engine/policies.py": "bff4a636cecfbb9c31be98b5a02a6af6e7cbe5bdfea6414f08b125f0d4596fff", "engine/chart_context.py": "ffe3393bc13bd54b9b65235ca51a4c15a0b40fe6404a5897379eb0dd2748d868", "engine/__init__.py": "717a1b9838b948a88d2fb5106069db57dce81dc54854cf2be5fb54150c7ac652", "engine/regions.py": "33058eec02bb378050febc577a2d85de97d34083234a08a2a8fa9c4b8dc62239", "engine/chart_caption.py": "46e8fda5d7b7ddcebc361b77eb8a86d933ec12f7721466dda55ccf7c12c9db8c", "engine/score_adapter.py": "3fb786b616baa6bfe8007e6ef8e1ca25e57e51e28a9aee9945bf7d1f629df0f1", "engine/text_cleanup.py": "f02f472d7b3cb08eca5b11d3ecf80ee6ec0636857ade17fb4cd11c13f371c2e4", "engine/geometry.py": "76671df686b574a827205acb757eecf702e970470ffa15f81b3129cbcd95e242", "engine/navit_geometry.py": "469c05b5afa19ed5669c8866185a1a149846601c6e45f6c002d55ae80cb6424b", "engine/header_inference_probe.py": "8ecb8d165525d4bf3c715bf2f5f4c840ce23905e1bd8b0551ed722d6726a2dc1", "engine/convert/semantic_styles.py": "3d92146e6c713b9de67cb71e92fe254a1344d9054c3925a85cb778ec68c0d075", "engine/convert/latex2html.py": "d583b5fa7fdbe181b190d2c8c22034318a51ef25d47203f9c7504bf35b508a72", "engine/convert/rich_table_fallback.py": "c2abbc602b293bf14a499f680de9f7d1584fdcf1438a374e3c75cf3705c0b793", "engine/convert/hybrid_cleanup.py": "f2f319de9bdabcfc5e264cdef6ff9d13235c488a372e9f58f676017f28f7c5c0", "engine/convert/__init__.py": "717a1b9838b948a88d2fb5106069db57dce81dc54854cf2be5fb54150c7ac652", "engine/convert/hybrid_inline.py": "0296f7827e74ae3ab64e9c9efbd496f096f0a3bbde7c3ad64d017539f1db7cb6", "engine/convert/rich_table_reader.py": "8a8b615681d3adcb46c5290ad631abffeccb2af0ee18f25e82f0cdb70f5befac" }, "scorer_runtime": { "aiohappyeyeballs": "2.7.1", "aiohttp": "3.14.3", "aiosignal": "1.4.0", "anls_star": "1.0.1", "annotated-doc": "0.0.4", "annotated-types": "0.7.0", "anyio": "4.12.1", "apted": "1.0.3", "attrs": "25.4.0", "autoevals": "0.1.0", "beautifulsoup4": "4.14.3", "bleach": "6.3.0", "certifi": "2026.2.25", "charset-normalizer": "3.5.1", "chevron": "0.14.0", "click": "8.3.1", "distro": "1.9.0", "filelock": "3.25.2", "fire": "0.7.1", "frozenlist": "1.8.0", "fsspec": "2026.2.0", "fuzzysearch": "0.8.1", "h11": "0.16.0", "hf-xet": "1.4.2", "html5lib": "1.1", "httpcore": "1.0.9", "httpx": "0.28.1", "huggingface_hub": "1.7.1", "idna": "3.11", "joblib": "1.5.2", "jsonschema": "4.26.0", "jsonschema-specifications": "2025.9.1", "Levenshtein": "0.27.3", "llama_cloud": "1.6.0", "lxml": "6.0.2", "Markdown": "3.10.2", "markdown-it-py": "4.0.0", "markdown2": "2.5.5", "mdurl": "0.1.2", "multidict": "6.9.1", "munkres": "1.1.4", "numpy": "2.4.3", "packaging": "26.0", "pandas": "3.0.1", "parse-bench": "1.0.4", "pdf2image": "1.16.3", "pillow": "11.1.0", "polyleven": "0.11.0", "propcache": "0.5.4", "pydantic": "2.12.5", "pydantic_core": "2.41.5", "Pygments": "2.19.2", "pymupdf": "1.26.3", "pypdf": "6.8.0", "pytesseract": "0.3.13", "python-dateutil": "2.9.0.post0", "python-dotenv": "1.2.2", "python-Levenshtein": "0.27.3", "PyYAML": "6.0.3", "RapidFuzz": "3.14.3", "referencing": "0.37.0", "requests": "2.34.2", "rich": "14.3.3", "rpds-py": "0.30.0", "scikit-learn": "1.7.2", "scipy": "1.17.1", "shellingham": "1.5.4", "six": "1.17.0", "sniffio": "1.3.1", "soupsieve": "2.8.3", "termcolor": "3.3.0", "threadpoolctl": "3.6.0", "tqdm": "4.67.3", "typer": "0.24.1", "typing-inspection": "0.4.2", "typing_extensions": "4.15.0", "Unidecode": "1.4.0", "urllib3": "2.8.0", "webencodings": "0.5.1", "yarl": "1.25.1" }, "complete": true, "coverage": { "fresh_generation_pages": 2078, "fresh_scored_pages_per_mode": 2078, "modes_scored": 4, "official_errors": 0, "portable_output_parity_records": 8312, "portable_union_region_parity_pages": 2078, "portable_http_request_function_unchanged": true, "portable_http_end_to_end_rerun": true, "final_title_guard_replay": true, "final_title_guard_replay_records": 8312, "final_code_difference_from_http_snapshot": "Line-bounded chart title parsing in both rendering paths and conservative colspan-only next-row header inference. Four pages per mode have changed Markdown; raw outputs, coordinates, and cell text/order/spans are identical.", "corrected_renderer_replay_records": 8312, "corrected_renderer_fresh_official_scoring": true }, "serving_runtime": { "versions": { "torch": "2.11.0+cu130", "torchvision": "0.26.0+cu130", "transformers": "5.14.1", "huggingface-hub": "1.24.0", "accelerate": "1.12.0", "timm": "1.0.22", "einops": "0.8.1", "Pillow": "12.2.0", "numpy": "2.2.6", "safetensors": "0.8.0", "beautifulsoup4": "4.13.3", "vllm": "0.26.0", "openai": "2.30.0", "nemotron-parse": "3.5.0.post2" }, "gpu": "NVIDIA H100 80GB HBM3", "legacy_sitecustomize_env": null, "pythonpath": null, "wheel": "nemotron_parse-3.5.0.post2-py3-none-any.whl" }, "metrics": { "single": { "charts": 0.5645987033310976, "visual_grounding": 0.7038987254479313, "tables": 0.8050745889408272, "content_faithfulness": 0.8795954910356779, "semantic_formatting": 0.6826974043965897, "overall": 0.7271729826304247 }, "double": { "charts": 0.6356501382029551, "visual_grounding": 0.7066562965971672, "tables": 0.840469160615296, "content_faithfulness": 0.8797729265983598, "semantic_formatting": 0.6816760525433225, "overall": 0.7488449149114201 }, "triple": { "charts": 0.6425329756315671, "visual_grounding": 0.7066302912429703, "tables": 0.8458237023891111, "content_faithfulness": 0.879557172435472, "semantic_formatting": 0.6826974043965897, "overall": 0.751448309219142 } }, "optional_union_metrics": { "charts": 0.6439854404202996, "visual_grounding": 0.7067302912429704, "tables": 0.8467224250086215, "content_faithfulness": 0.8795990444408648, "semantic_formatting": 0.6827725437713926, "overall": 0.7519619489768298 }, "paired_union_delta_percentage_points": 0.051363975768781156, "prediction_sha256": { "single": "f62ab4ed079b79241cce467eb1ad50577a172ca04124d642c31afba8ab7048b2", "double": "3c74420e780ffe36fa60dd3a1abe701d3405be3957242a75326e9fe868a8be6f", "triple": "80b7700cd44e9ee3694454b944bbebf7a9c45ee628b33b1af70bfd8c9fb0a207", "triple_detector": "cf93b01c97cdc8376e501282cbb2498acd174145de1c7d158e362e85df703a00" }, "portable_python_sha256": { "serve.py": "b5a772c9108f9ac9acffb42a872bf04694e43c0cb10416ba2bb3d5572b970f5b", "score.py": "a566accf056987c90201b6bc75e4ecf02eb56cf51a1e65fc3b619c108aa0b60e", "postprocess.py": "576db3eb97c5e43ebbb1b5cdd11aef212f7728fca0b4bfa20fc4d2f309e8eb5f", "run.py": "116a08d75f620f7e4ccd4c9350e1cffc863b04c2c9a14e147815776b572964fa", "prepare_data.py": "68d87dfac34335ab4ad34b6d7b0d04c7973d49b92afec3d9e167bf7ea24d0010", "prepare_inputs.py": "f3c0354e8559e42678d43a9071b5ba4f654c7bcd13426493913b622fa4d96795", "engine/chart_context.py": "ffe3393bc13bd54b9b65235ca51a4c15a0b40fe6404a5897379eb0dd2748d868", "engine/markdown.py": "fc57566255ac632b24825973d9533a13fd717dad4e38d096098bac33ecc79957", "engine/merge.py": "b1c2e749e8f772b716c5a8a2ebeba7348474eb9368b00d37f1c8987e26d24e4e", "engine/score_adapter.py": "3fb786b616baa6bfe8007e6ef8e1ca25e57e51e28a9aee9945bf7d1f629df0f1", "engine/__init__.py": "717a1b9838b948a88d2fb5106069db57dce81dc54854cf2be5fb54150c7ac652", "engine/chart_caption.py": "fe82e0eca23be96d7cc2199657432bd2ddf464313d41fed3984bf5960fddb0e6", "engine/text_cleanup.py": "f02f472d7b3cb08eca5b11d3ecf80ee6ec0636857ade17fb4cd11c13f371c2e4", "engine/geometry.py": "76671df686b574a827205acb757eecf702e970470ffa15f81b3129cbcd95e242", "engine/navit_geometry.py": "469c05b5afa19ed5669c8866185a1a149846601c6e45f6c002d55ae80cb6424b", "engine/caption_relocation.py": "3a17677b1a43c793cd817f8f3f5745be807203ff75b8e3e0af5a4d51d547dd4b", "engine/header_inference_probe.py": "02950f1aec5e2dc88a96487bd9124c97e5ce2842fbcfc5264ac2d5f697865722", "engine/policies.py": "4c1f78be4d8061e4224fb59c3dc5342dfc21d9d37826814ea3573f339777aebd", "engine/regions.py": "33058eec02bb378050febc577a2d85de97d34083234a08a2a8fa9c4b8dc62239", "engine/convert/hybrid_inline.py": "0296f7827e74ae3ab64e9c9efbd496f096f0a3bbde7c3ad64d017539f1db7cb6", "engine/convert/rich_table_fallback.py": "c2abbc602b293bf14a499f680de9f7d1584fdcf1438a374e3c75cf3705c0b793", "engine/convert/rich_table_reader.py": "8a8b615681d3adcb46c5290ad631abffeccb2af0ee18f25e82f0cdb70f5befac", "engine/convert/latex2html.py": "d583b5fa7fdbe181b190d2c8c22034318a51ef25d47203f9c7504bf35b508a72", "engine/convert/hybrid_cleanup.py": "f2f319de9bdabcfc5e264cdef6ff9d13235c488a372e9f58f676017f28f7c5c0", "engine/convert/semantic_styles.py": "3d92146e6c713b9de67cb71e92fe254a1344d9054c3925a85cb778ec68c0d075", "engine/convert/__init__.py": "717a1b9838b948a88d2fb5106069db57dce81dc54854cf2be5fb54150c7ac652" }, "source_hash_scope": "executed_python_source_hashes preserves the original H100 HTTP snapshot. portable_python_sha256 identifies the corrected final renderer, replayed on all8312 saved responses and freshly scored. The hashes differ in engine/chart_caption.py, engine/policies.py, and engine/header_inference_probe.py.", "generation_requests": { "single": { "complete": true, "pages": 2078, "passes": 1, "fullpage_requests": 2078, "detection_requests": 0, "crop_requests": 0, "generation_errors": 0, "fullpage_seconds": 142.38590936901164, "additional_mode_seconds": 2.474504841986345, "fullpage_shared_across_modes": true, "temporary_crops_removed": true }, "double": { "complete": true, "pages": 2078, "passes": 2, "fullpage_requests": 2078, "detection_requests": 0, "crop_requests": 2339, "generation_errors": 0, "fullpage_seconds": 142.38590936901164, "additional_mode_seconds": 77.63449284402304, "fullpage_shared_across_modes": true, "temporary_crops_removed": true }, "triple_union": { "complete": true, "pages": 2078, "passes": 3, "fullpage_requests": 2078, "detection_requests": 2078, "crop_requests": 2616, "generation_errors": 0, "fullpage_seconds": 142.38590936901164, "additional_mode_seconds": 174.39476006900077, "fullpage_shared_across_modes": true, "temporary_crops_removed": true } }, "detector_only_control": { "complete": true, "pages": 2078, "detector_only_crop_responses": 2418, "fresh_full_pipeline_responses": true, "new_http_requests": 0, "control": "Same portable postprocessor, same full-page/detector/original-region crop responses; exclude only union additions." }, "prediction_hash_scope": "Hashes identify the corrected saved-response replay used for fresh official scoring; original HTTP prediction hashes are retained separately.", "http_prediction_sha256": { "single": "1d9da1a592541810fafc4320a4fc3519bd4123e2bc3d45143f4fa98a8f8260a3", "double": "f9a25c89ce75dd23a1a26189c4ff67f6aed131d3a24a08492e9e6af8b8dee3f3", "triple": "e4366d479b6016866333e106d3efc0d73efbff27a62ef36c500a7c09f49825d9", "triple_detector": "297335946dfbef69387ecfae66c158941bda9e153ad6aaa809108c771b2a378f" }, "postprocessing_correction": { "new_inference_requests": 0, "fixtures": { "data_row_not_promoted": true, "explicit_rowspan_preserved": true, "bold_headers_preserved": true, "year_data_not_promoted": true, "bold_numeric_data_not_promoted": true, "empty_title_full_pipeline": true, "nonempty_title_preserved": true }, "modes": { "single": { "changed_pages": 4, "raw_and_geometry_unchanged": true, "cell_text_order_spans_unchanged": true, "prediction_sha256": "f62ab4ed079b79241cce467eb1ad50577a172ca04124d642c31afba8ab7048b2" }, "double": { "changed_pages": 4, "raw_and_geometry_unchanged": true, "cell_text_order_spans_unchanged": true, "prediction_sha256": "3c74420e780ffe36fa60dd3a1abe701d3405be3957242a75326e9fe868a8be6f" }, "triple": { "changed_pages": 4, "raw_and_geometry_unchanged": true, "cell_text_order_spans_unchanged": true, "prediction_sha256": "80b7700cd44e9ee3694454b944bbebf7a9c45ee628b33b1af70bfd8c9fb0a207" }, "triple_detector": { "changed_pages": 4, "raw_and_geometry_unchanged": true, "cell_text_order_spans_unchanged": true, "prediction_sha256": "cf93b01c97cdc8376e501282cbb2498acd174145de1c7d158e362e85df703a00" } } } }