{ "continuation_coverage": { "decoded_world_translation_norm_max": 19.901873486912873, "first_heading_abs_max": 2.8601129628888535, "first_heading_abs_min": 0.004064200935457434, "generation_token_offsets_observed": [ 0, 10 ], "history_frames_observed": [ 0, 40 ], "nonzero_first_heading_count": 1920 }, "contract_validation": { "candidate_graph_sha256": { "decoder": "37085139b5572ed0536e9e231c1e7af023d7cd98e536495708258fa1642e9b1d", "denoiser": "7fd49903ec65c0831e2fa692081fd64d8818c6d1f62320d2f2b5e71467a3ca64", "text_encoder": "c9a729e5490e58e77936d96c42d2bcf9a84e17717f633d80d7c2f52bbf66d552" }, "candidate_policy_version": 3, "candidate_precision_format": "mixed-fp16", "graph_conversion_modes": { "decoder": "mixed-fp16", "denoiser": "fp32-identity", "text_encoder": "fp32-identity" }, "graph_policy_ids": { "decoder": "decoder-qk-norm-io-geometry-v3", "denoiser": "denoiser-continuation-fp32-v3", "text_encoder": "text-continuation-fp32-v3" }, "identity_graphs_byte_identical": [ "denoiser", "text_encoder" ], "non_precision_contract_equal": true, "non_precision_contract_sha256": "fa5cb308c36543d9f335c26dbd101593260efeca0d1b5553774d77b513efd84c", "reference_graph_sha256": { "decoder": "1fcd909b2997002edca559ba016ab3accc5b2d5e5ba0c7e85a6ab54d2100a80f", "denoiser": "7fd49903ec65c0831e2fa692081fd64d8818c6d1f62320d2f2b5e71467a3ca64", "text_encoder": "c9a729e5490e58e77936d96c42d2bcf9a84e17717f633d80d7c2f52bbf66d552" }, "tokenizer_payload_sha256": { "tokenizer/tokenizer.json": "2fc687b11de0bc1b3d8348f92e3b49ef1089a621506c7661fbf3248fcd54947e", "tokenizer/tokenizer_config.json": "4b91f63e0ec01ae5112b4fab0027b301a38d94f6b1fd001abf2236390168b553" }, "tokenizer_payloads_equal": true }, "cpu_timing": { "candidate_mean_seconds_per_case": 1.3008700043089145, "candidate_total_seconds": 249.76704082731158, "note": "CPU FP16 timing is diagnostic only; the deployment backend is WebGPU. Each case includes every consecutive window.", "reference_mean_seconds_per_case": 1.2313923729343514, "reference_total_seconds": 236.42733560339548, "windows_per_case": 5 }, "format": "ardy-browser-fp16-aggregate", "format_version": 2, "method": { "candidate": "mixed-FP16 ONNX Runtime CPU with graph optimizations disabled", "case_count": 192, "cfg_weight": 3.5, "frames_per_case": 200, "initial_heading": 0.35, "initial_translation": [ 1.25, 0.20000000298023224, -0.75 ], "noise_generator": "browser PortableRandom (Mulberry32 + Marsaglia polar)", "prompt_count": 64, "prompt_selection": "first-occurrence text deduplication in manifest order, then evenly spaced numpy.linspace integer indices", "prompt_sha256": "be55085c05d4b4c1a228edf5bea501ee77a33420a3d886b6e8ea0b737ad963ef", "prompt_split": "test", "reference": "FP32 ONNX Runtime CPU with graph optimizations disabled", "rollout": "5 consecutive browser windows, 10-step deterministic DDIM per window, 40-frame history, recenter, FSQ requantize, heading, and accumulated world translation", "seeds": [ 12031, 987654, 20260729 ], "window_count": 5 }, "models": { "candidate": { "id": "ardy-minilm-core40-browser-v1", "manifest_sha256": "07a0cb3984cc324e9318549d03494d3fa7193727367ae45b2087ca8a978b6d0b", "raw_size_bytes": 740128936, "revision": "82059e1b3e8c861d1f11128184c58ce17a62795e0e24289a295c0a70cd78befe", "transport_size_bytes": 684221164 }, "raw_saved_bytes": 35450964, "raw_saved_fraction": 0.04570897724399514, "reference": { "id": "ardy-minilm-core40-browser-v1", "manifest_sha256": "982095622886845ed985f353ab5e3806db5da74f3aa0b85acb0415e81e48e5a1", "raw_size_bytes": 775579900, "revision": "82059e1b3e8c861d1f11128184c58ce17a62795e0e24289a295c0a70cd78befe", "transport_size_bytes": 717533539 }, "transport_saved_bytes": 33312375, "transport_saved_fraction": 0.04642622705333921 }, "motion_fidelity": { "contact_agreement": { "max": 1.0, "mean": 0.99998046875, "p95": 1.0 }, "contact_f1": 0.9999878455411367, "contact_iou": 0.9999756913777317, "global_rotation_error_degrees": { "max": 0.10194642096757889, "mean": 0.07647083381501336, "p95": 0.09021148458123206 }, "heading_error_degrees": { "max": 0.00926036573946476, "mean": 0.006903713561769109, "p95": 0.008434763876721263 }, "joint_velocity_error_mps": { "max": 0.005097570829093456, "mean": 0.0031330614153072625, "p95": 0.003811004789895378 }, "local_rotation_error_degrees": { "max": 0.09130936861038208, "mean": 0.06878133238448451, "p95": 0.08178812451660632 }, "mpjpe_m": { "max": 0.00040933865238912404, "mean": 0.00026524110944592394, "p95": 0.0003497354977298528 }, "normalized_motion_cosine": { "max": 0.999998631978834, "mean": 0.9999957261172812, "p95": 0.9999975190029062 }, "normalized_motion_rmse": { "max": 0.0058677298948168755, "mean": 0.0026446143092471175, "p95": 0.0033932944410480555 }, "root_ade_m": { "max": 0.0, "mean": 0.0, "p95": 0.0 }, "root_aligned_mpjpe_m": { "max": 0.00040933865238912404, "mean": 0.0002652411005783506, "p95": 0.00034973549481946974 }, "root_fde_m": { "max": 0.0, "mean": 0.0, "p95": 0.0 } }, "motion_fidelity_by_window": [ { "generation_token_offset": 0, "history_frames": 0, "motion_fidelity": { "contact_agreement": { "max": 1.0, "mean": 0.9999674479166667, "p95": 1.0 }, "contact_f1": 0.9999793503624012, "contact_iou": 0.9999587015775997, "global_rotation_error_degrees": { "max": 0.1542610079050064, "mean": 0.08115143185326208, "p95": 0.09792995154857634 }, "heading_error_degrees": { "max": 0.011627002619206905, "mean": 0.00684818510368738, "p95": 0.009582063695415855 }, "joint_velocity_error_mps": { "max": 0.011050221510231495, "mean": 0.0032832376238426755, "p95": 0.004385076943435705 }, "local_rotation_error_degrees": { "max": 0.13925813138484955, "mean": 0.07251732908965398, "p95": 0.08770628459751606 }, "mpjpe_m": { "max": 0.0006877752603031695, "mean": 0.00028308230609278934, "p95": 0.000383828738995362 }, "normalized_motion_cosine": { "max": 0.9999989755044695, "mean": 0.9999955959958736, "p95": 0.9999976699565432 }, "normalized_motion_rmse": { "max": 0.011462797410786152, "mean": 0.002799639324318074, "p95": 0.0036801897920668114 }, "root_ade_m": { "max": 0.0, "mean": 0.0, "p95": 0.0 }, "root_aligned_mpjpe_m": { "max": 0.0006877753185108304, "mean": 0.0002830822916166653, "p95": 0.00038382867351174346 }, "root_fde_m": { "max": 0.0, "mean": 0.0, "p95": 0.0 } }, "window_index": 0 }, { "generation_token_offset": 10, "history_frames": 40, "motion_fidelity": { "contact_agreement": { "max": 1.0, "mean": 0.9999674479166667, "p95": 1.0 }, "contact_f1": 0.9999801921362781, "contact_iou": 0.9999603850572436, "global_rotation_error_degrees": { "max": 0.11084937304258347, "mean": 0.0757197640875044, "p95": 0.09547545947134492 }, "heading_error_degrees": { "max": 0.014484826475381851, "mean": 0.006939516270601113, "p95": 0.010117053054273127 }, "joint_velocity_error_mps": { "max": 0.004555481136776507, "mean": 0.003044415858160695, "p95": 0.004036348516819999 }, "local_rotation_error_degrees": { "max": 0.09590752422809601, "mean": 0.06787548580905423, "p95": 0.08357225395739079 }, "mpjpe_m": { "max": 0.00046821762225590646, "mean": 0.00026081187509892817, "p95": 0.0003520364392898045 }, "normalized_motion_cosine": { "max": 0.9999986234208106, "mean": 0.9999957953630636, "p95": 0.9999976744341992 }, "normalized_motion_rmse": { "max": 0.004383150488138199, "mean": 0.0025761706213719058, "p95": 0.0032808053190819917 }, "root_ade_m": { "max": 0.0, "mean": 0.0, "p95": 0.0 }, "root_aligned_mpjpe_m": { "max": 0.00046821770956739783, "mean": 0.0002608118744925984, "p95": 0.0003520363214192912 }, "root_fde_m": { "max": 0.0, "mean": 0.0, "p95": 0.0 } }, "window_index": 1 }, { "generation_token_offset": 10, "history_frames": 40, "motion_fidelity": { "contact_agreement": { "max": 1.0, "mean": 1.0, "p95": 1.0 }, "contact_f1": 1.0, "contact_iou": 1.0, "global_rotation_error_degrees": { "max": 0.10496819019317627, "mean": 0.07381617318606004, "p95": 0.09214951433241367 }, "heading_error_degrees": { "max": 0.012519109062850475, "mean": 0.006783440334402258, "p95": 0.009841464878991246 }, "joint_velocity_error_mps": { "max": 0.004801172763109207, "mean": 0.0029651702508696567, "p95": 0.003973252794821746 }, "local_rotation_error_degrees": { "max": 0.09078813344240189, "mean": 0.06634670447480555, "p95": 0.08243084140121935 }, "mpjpe_m": { "max": 0.00038963230326771736, "mean": 0.0002578581997264943, "p95": 0.0003529940586304292 }, "normalized_motion_cosine": { "max": 0.9999986625956077, "mean": 0.9999958393762132, "p95": 0.9999976617489793 }, "normalized_motion_rmse": { "max": 0.004302056040614843, "mean": 0.002542274758525309, "p95": 0.0035095547325909135 }, "root_ade_m": { "max": 0.0, "mean": 0.0, "p95": 0.0 }, "root_aligned_mpjpe_m": { "max": 0.00038963239057920873, "mean": 0.0002578581943453173, "p95": 0.00035299402952659875 }, "root_fde_m": { "max": 0.0, "mean": 0.0, "p95": 0.0 } }, "window_index": 2 }, { "generation_token_offset": 10, "history_frames": 40, "motion_fidelity": { "contact_agreement": { "max": 1.0, "mean": 0.9999674479166667, "p95": 1.0 }, "contact_f1": 0.9999794758122448, "contact_iou": 0.9999589524669568, "global_rotation_error_degrees": { "max": 0.10823715478181839, "mean": 0.07608008959020178, "p95": 0.09582429341971872 }, "heading_error_degrees": { "max": 0.01221688836812973, "mean": 0.0068974279044293025, "p95": 0.00953664625994861 }, "joint_velocity_error_mps": { "max": 0.004709368804469705, "mean": 0.0030876657365297433, "p95": 0.0040596173057565466 }, "local_rotation_error_degrees": { "max": 0.09616772830486298, "mean": 0.06883417316324388, "p95": 0.08628824278712273 }, "mpjpe_m": { "max": 0.00043023788020946085, "mean": 0.00026473069290962786, "p95": 0.0003749165509361773 }, "normalized_motion_cosine": { "max": 0.9999986695273096, "mean": 0.9999958637302813, "p95": 0.9999977874028252 }, "normalized_motion_rmse": { "max": 0.0042211017571389675, "mean": 0.002629348922710051, "p95": 0.0036556855542585233 }, "root_ade_m": { "max": 0.0, "mean": 0.0, "p95": 0.0 }, "root_aligned_mpjpe_m": { "max": 0.0004302378511056304, "mean": 0.00026473067805454775, "p95": 0.00037491636758204546 }, "root_fde_m": { "max": 0.0, "mean": 0.0, "p95": 0.0 } }, "window_index": 3 }, { "generation_token_offset": 10, "history_frames": 40, "motion_fidelity": { "contact_agreement": { "max": 1.0, "mean": 1.0, "p95": 1.0 }, "contact_f1": 1.0, "contact_iou": 1.0, "global_rotation_error_degrees": { "max": 0.10978787392377853, "mean": 0.07558671078489472, "p95": 0.09087162539362907 }, "heading_error_degrees": { "max": 0.013133668340742588, "mean": 0.007049998545577789, "p95": 0.010118653997778893 }, "joint_velocity_error_mps": { "max": 0.004520879301708192, "mean": 0.003058643160329666, "p95": 0.0039847704756539315 }, "local_rotation_error_degrees": { "max": 0.09357167035341263, "mean": 0.0683329723154505, "p95": 0.08376629874110221 }, "mpjpe_m": { "max": 0.00041242953739129007, "mean": 0.0002597224709006696, "p95": 0.0003502150691929273 }, "normalized_motion_cosine": { "max": 0.9999987097596618, "mean": 0.9999959060855707, "p95": 0.9999978618613875 }, "normalized_motion_rmse": { "max": 0.004128637723624706, "mean": 0.002588702941769346, "p95": 0.0032814822625368834 }, "root_ade_m": { "max": 0.0, "mean": 0.0, "p95": 0.0 }, "root_aligned_mpjpe_m": { "max": 0.0004124295082874596, "mean": 0.0002597224641552505, "p95": 0.0003502151652355678 }, "root_fde_m": { "max": 0.0, "mean": 0.0, "p95": 0.0 } }, "window_index": 4 } ], "prompt_manifest": { "filename": "prompts-core40-timeline.jsonl", "provenance_sidecar": { "content": { "counts": { "dropped_prompt_too_long": 2, "groups_written": 3702, "missing_propagation_references": 12, "raw_event_descriptions": 352703, "raw_overview_descriptions": 142220, "recording_components": 3702, "recording_families": 4090, "sources": { "events.description": 46538, "overview_description": 17749 }, "split_groups": { "test": 341, "train": 2977, "val": 384 }, "splits": { "test": 6095, "train": 51482, "val": 6710 }, "timeline_rows": 142220, "unique_before_sampling": 64287, "unique_descriptions": 64287, "written": 64287 }, "dataset": { "filename": "timelines.jsonl", "license": "CC BY 4.0", "owner": "NVIDIA", "repo_id": "nvidia/SEED-Timeline-Annotations", "resolved_from": "hugging_face_hub", "revision": "b2cf916d8ef7a1e49fc4f0ce9e00c1981d3b9d8f", "sha256": "379d6a5b86cea06b7201d485d19ee53512cc58449352b3cf113a95d1d27603d8", "size_bytes": 80373523, "url": "https://huggingface.co/datasets/nvidia/SEED-Timeline-Annotations" }, "format": "ardy-minilm-prompt-provenance", "format_version": 1, "manifest": { "filename": "prompts-core40-timeline.jsonl", "sha256": "c80b0656dd04c28da1d665b4b9a9422f975be5c72b0fe9011b976a3672bb1eac" }, "preparation": { "deduplication": "global case- and punctuation-insensitive alphanumeric-token match after normalization; overview_description preferred", "grouping": "NFKC/casefold filename; require __A[_M], remove actor, mirror, and terminal three-digit non-angle take suffixes; union propagated_from_filename families", "max_prompt_characters": 512, "normalization": "Unicode NFKC, trim, collapse whitespace", "sample_size": 0, "seed": 20260726, "sources": [ "overview_description", "events.description" ], "split_hash_namespace": "ardy-minilm-nvidia-timeline-split-v3", "split_ratios": { "test": 0.1, "train": 0.8, "val": 0.1 } } }, "filename": "prompts-core40-timeline.metadata.json", "sha256": "37086140d78b36b5145de36e8c64ac0558c86960c4bbcc96630f6c0556ffff64", "size_bytes": 2047 }, "sha256": "c80b0656dd04c28da1d665b4b9a9422f975be5c72b0fe9011b976a3672bb1eac", "size_bytes": 12655589 }, "runtime_environment": { "logical_cpu_count": 20, "onnxruntime": { "execution_provider": "CPUExecutionProvider", "graph_optimizations": "disabled", "intra_op_num_threads": 8 }, "packages": { "numpy": "1.26.4", "onnx": "1.18.0", "onnxruntime": "1.27.0", "transformers": "5.8.1" }, "platform": { "machine": "aarch64", "processor": "aarch64", "release": "6.17.0-1026-nvidia", "system": "Linux" }, "python": { "implementation": "CPython", "version": "3.11.15" } }, "source_report_schema_version": 3, "text_conditions": { "cosine_mean": 1.0, "cosine_min": 0.9999998807907104, "mae": 0.0, "max_abs": 0.0, "rmse": 0.0 } }