File size: 28,863 Bytes
a3682cf a19847a a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf a19847a 806ff14 a3682cf 806ff14 a19847a a3682cf a19847a 806ff14 a19847a a3682cf a19847a 806ff14 a19847a 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf 806ff14 a3682cf | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 | {
"@context": {
"@language": "en",
"@vocab": "https://schema.org/",
"sc": "https://schema.org/",
"cr": "http://mlcommons.org/croissant/",
"dct": "http://purl.org/dc/terms/",
"prov": "http://www.w3.org/ns/prov#",
"rai": "http://mlcommons.org/croissant/RAI/",
"field": "cr:field",
"recordSet": "cr:recordSet",
"source": "cr:source",
"fileObject": "cr:fileObject",
"fileSet": "cr:fileSet",
"extract": "cr:extract",
"column": "cr:column",
"fileProperty": "cr:fileProperty",
"jsonPath": "cr:jsonPath",
"dataType": "cr:dataType",
"containedIn": "cr:containedIn",
"includes": "cr:includes",
"conformsTo": "dct:conformsTo",
"citeAs": "cr:citeAs",
"md5": "sc:md5",
"sha256": "sc:sha256"
},
"@type": "sc:Dataset",
"name": "Temporal Twins Benchmark",
"description": "Temporal Twins is a synthetic UPI-style transaction benchmark for temporal fraud detection. The collection contains oracle_calib, easy, medium, and hard matched-prefix benchmark slices across deterministic seeds 0, 1, 2, 3, and 4. Fraud labels are assigned through delayed temporal mechanisms rather than static per-transaction attributes, and matched fraud/benign twin examples are aligned at the same local prefix index to suppress static shortcuts while preserving order-sensitive temporal structure.",
"url": "https://huggingface.co/datasets/temporal-twins-benchmark/temporal-twins",
"license": "https://creativecommons.org/licenses/by/4.0/",
"isBasedOn": {
"@type": "sc:SoftwareSourceCode",
"name": "Temporal Twins benchmark code",
"url": "https://huggingface.co/temporal-twins-benchmark/temporal-twins-code",
"license": "https://www.apache.org/licenses/LICENSE-2.0",
"identifier": "Apache-2.0"
},
"conformsTo": "http://mlcommons.org/croissant/1.1",
"citation": "Anonymous NeurIPS 2026 submission for Temporal Twins; final citation will be added after review.",
"citeAs": "Temporal Twins Benchmark (synthetic UPI-style temporal fraud benchmark). Anonymous NeurIPS 2026 submission; final citation will be added after review. Code repository: https://huggingface.co/temporal-twins-benchmark/temporal-twins-code.",
"creator": [
{
"@type": "sc:Organization",
"name": "Temporal Twins Benchmark Contributors"
}
],
"dateCreated": "2026-05-04",
"datePublished": "2026-05-04",
"version": "1.0.0",
"keywords": [
"synthetic financial transactions",
"UPI-style benchmark",
"temporal fraud detection",
"matched temporal twins",
"matched-prefix evaluation",
"sequence modeling",
"dynamic graph learning",
"reproducible benchmark"
],
"distribution": [
{
"@id": "metadata-files",
"@type": "cr:FileSet",
"name": "Metadata files",
"description": "Metadata payload for the hosted release, including this Croissant file and companion notes.",
"includes": "metadata/*.json",
"encodingFormat": "application/ld+json"
},
{
"@id": "data-archive",
"@type": "cr:FileObject",
"name": "Temporal Twins data archive",
"description": "Zip archive containing synthetic transaction and matched-prefix parquet files for oracle_calib, easy, medium, and hard across seeds 0 through 4.",
"contentUrl": "https://huggingface.co/datasets/temporal-twins-benchmark/temporal-twins/resolve/main/temporal_twins_data.zip",
"encodingFormat": "application/zip",
"sha256": "eb5a76dfa9391be447e9aa23b0d57527ac8e0d1e9d8df0277b209734483af49c"
},
{
"@id": "transactions-files",
"@type": "cr:FileSet",
"name": "Synthetic transactions parquet files",
"description": "Expected synthetic transaction files for benchmark modes oracle_calib, easy, medium, and hard across seeds 0 through 4.",
"containedIn": {
"@id": "data-archive"
},
"includes": "data/*/seed_*/transactions.parquet",
"encodingFormat": "application/x-parquet"
},
{
"@id": "matched-prefix-files",
"@type": "cr:FileSet",
"name": "Matched-prefix example parquet files",
"description": "Expected matched-prefix benchmark examples for the hosted release. Each file contains fraud and benign twin examples evaluated at the same local prefix index.",
"containedIn": {
"@id": "data-archive"
},
"includes": "data/*/seed_*/matched_pairs.parquet",
"encodingFormat": "application/x-parquet"
},
{
"@id": "config-files",
"@type": "cr:FileSet",
"name": "Benchmark config files",
"description": "YAML configuration files for the hosted release.",
"includes": "configs/*.yaml",
"encodingFormat": "text/yaml"
},
{
"@id": "results-files",
"@type": "cr:FileSet",
"name": "Results files",
"description": "Hosted result summaries and diagnostics for the deterministic paper suite.",
"includes": "results/*.csv",
"encodingFormat": "text/csv"
},
{
"@id": "paper-suite-runs-csv",
"@type": "cr:FileObject",
"name": "Per-run paper-suite results",
"description": "Per-run deterministic results for the final 5-seed paper-scale suite.",
"contentUrl": "https://huggingface.co/datasets/temporal-twins-benchmark/temporal-twins/raw/main/results/paper_suite_runs.csv",
"encodingFormat": "text/csv",
"sha256": "1445666d207ab28d94678cdbf3625bf771700bdd1c444aa0cf01f41f6672055e"
},
{
"@id": "paper-suite-summary-csv",
"@type": "cr:FileObject",
"name": "Paper-suite summary results",
"description": "Mean and standard deviation summary of the deterministic 5-seed paper suite.",
"contentUrl": "https://huggingface.co/datasets/temporal-twins-benchmark/temporal-twins/raw/main/results/paper_suite_summary.csv",
"encodingFormat": "text/csv",
"sha256": "aabe56ba6dfcb585903b4df74c53fcbcdb82a0b48e75b1214232f2fa2daaa6e4"
},
{
"@id": "paper-suite-runtime-csv",
"@type": "cr:FileObject",
"name": "Paper-suite runtime summary",
"description": "Runtime and StaticGNN evaluation diagnostics for the final paper suite.",
"contentUrl": "https://huggingface.co/datasets/temporal-twins-benchmark/temporal-twins/raw/main/results/paper_suite_runtime.csv",
"encodingFormat": "text/csv",
"sha256": "899415b8b34962cd1029b083a6f26282fe28402f03cd3877dd4da96d7840be74"
},
{
"@id": "paper-suite-failed-checks-csv",
"@type": "cr:FileObject",
"name": "Paper-suite failed gate checks",
"description": "Gate-check and advisory-check outcomes for each run in the final paper suite.",
"contentUrl": "https://huggingface.co/datasets/temporal-twins-benchmark/temporal-twins/raw/main/results/paper_suite_failed_checks.csv",
"encodingFormat": "text/csv",
"sha256": "860940b8e6594ba0b44ecf8f68d09eb0fa9b32b5426de38be6ed094cdfe4b267"
}
],
"recordSet": [
{
"@id": "transactions",
"@type": "cr:RecordSet",
"name": "transactions",
"description": "Synthetic UPI-style transactions spanning oracle_calib, easy, medium, and hard, with deterministic seeds 0 through 4.",
"field": [
{
"@id": "transactions/sender_id",
"@type": "cr:Field",
"name": "sender_id",
"description": "Synthetic sender account identifier.",
"dataType": "https://schema.org/Text",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "sender_id"
}
}
},
{
"@id": "transactions/receiver_id",
"@type": "cr:Field",
"name": "receiver_id",
"description": "Synthetic receiver account identifier.",
"dataType": "https://schema.org/Text",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "receiver_id"
}
}
},
{
"@id": "transactions/timestamp",
"@type": "cr:Field",
"name": "timestamp",
"description": "Synthetic event timestamp used to order transactions within each sender history.",
"dataType": "https://schema.org/Float",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "timestamp"
}
}
},
{
"@id": "transactions/amount",
"@type": "cr:Field",
"name": "amount",
"description": "Synthetic transaction amount.",
"dataType": "https://schema.org/Float",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "amount"
}
}
},
{
"@id": "transactions/risk_score",
"@type": "cr:Field",
"name": "risk_score",
"description": "Synthetic noisy risk score emitted by the simulator's risk engine.",
"dataType": "https://schema.org/Float",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "risk_score"
}
}
},
{
"@id": "transactions/failed",
"@type": "cr:Field",
"name": "failed",
"description": "Indicator for whether the synthetic transaction attempt failed.",
"dataType": "https://schema.org/Boolean",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "failed"
}
}
},
{
"@id": "transactions/is_fraud",
"@type": "cr:Field",
"name": "is_fraud",
"description": "Delayed synthetic fraud label attached to specific transactions.",
"dataType": "https://schema.org/Boolean",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "is_fraud"
}
}
}
]
},
{
"@id": "matched_prefix_examples",
"@type": "cr:RecordSet",
"name": "matched_prefix_examples",
"description": "Matched fraud and benign evaluation examples. Each benign twin is evaluated at the same local prefix index as the paired fraud twin, with matched static and prefix-level summaries. The release-facing field matched_local_event_idx is the matched prefix index and may correspond to the internal eval_local_event_idx column if files are exported directly from the current pipeline.",
"field": [
{
"@id": "matched_prefix_examples/twin_pair_id",
"@type": "cr:Field",
"name": "twin_pair_id",
"description": "Matched fraud/benign twin pair identifier.",
"dataType": "https://schema.org/Integer",
"source": {
"fileSet": {
"@id": "matched-prefix-files"
},
"extract": {
"column": "twin_pair_id"
}
}
},
{
"@id": "matched_prefix_examples/sender_id",
"@type": "cr:Field",
"name": "sender_id",
"description": "Sender evaluated at the matched prefix.",
"dataType": "https://schema.org/Text",
"source": {
"fileSet": {
"@id": "matched-prefix-files"
},
"extract": {
"column": "sender_id"
}
}
},
{
"@id": "matched_prefix_examples/matched_local_event_idx",
"@type": "cr:Field",
"name": "matched_local_event_idx",
"description": "Release-facing matched-prefix event index k used for both the fraud twin and its benign control.",
"dataType": "https://schema.org/Integer",
"source": {
"fileSet": {
"@id": "matched-prefix-files"
},
"extract": {
"column": "matched_local_event_idx"
}
}
},
{
"@id": "matched_prefix_examples/label",
"@type": "cr:Field",
"name": "label",
"description": "Binary matched-prefix label where 1 denotes the fraud twin example and 0 denotes the benign matched control.",
"dataType": "https://schema.org/Boolean",
"source": {
"fileSet": {
"@id": "matched-prefix-files"
},
"extract": {
"column": "label"
}
}
},
{
"@id": "matched_prefix_examples/benchmark_mode",
"@type": "cr:Field",
"name": "benchmark_mode",
"description": "Benchmark mode identifier, e.g. temporal_twins_oracle_calib or temporal_twins.",
"dataType": "https://schema.org/Text",
"source": {
"fileSet": {
"@id": "matched-prefix-files"
},
"extract": {
"column": "benchmark_mode"
}
}
},
{
"@id": "matched_prefix_examples/difficulty",
"@type": "cr:Field",
"name": "difficulty",
"description": "Difficulty slice within the release: oracle_calib, easy, medium, or hard.",
"dataType": "https://schema.org/Text",
"source": {
"fileSet": {
"@id": "matched-prefix-files"
},
"extract": {
"column": "difficulty"
}
}
},
{
"@id": "matched_prefix_examples/seed",
"@type": "cr:Field",
"name": "seed",
"description": "Deterministic benchmark seed in the final paper-scale suite.",
"dataType": "https://schema.org/Integer",
"source": {
"fileSet": {
"@id": "matched-prefix-files"
},
"extract": {
"column": "seed"
}
}
}
]
},
{
"@id": "audit_columns",
"@type": "cr:RecordSet",
"name": "audit_columns",
"description": "Audit and probe support columns carried by the synthetic generator for analysis, oracle-style scoring, and benchmark validation. These columns are not intended for ordinary model training and should be excluded from learned baseline inputs in benchmark evaluations.",
"field": [
{
"@id": "audit_columns/twin_role",
"@type": "cr:Field",
"name": "twin_role",
"description": "Twin role label such as fraud, benign, or background; excluded from ordinary model features.",
"dataType": "https://schema.org/Text",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "twin_role"
}
}
},
{
"@id": "audit_columns/template_id",
"@type": "cr:Field",
"name": "template_id",
"description": "Identifier for the matched temporal template used to construct a twin pair; excluded from ordinary model features.",
"dataType": "https://schema.org/Integer",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "template_id"
}
}
},
{
"@id": "audit_columns/motif_hit_count",
"@type": "cr:Field",
"name": "motif_hit_count",
"description": "Count of motif hits in the generator trace; exposed only for audit or probe logic, not learned baselines.",
"dataType": "https://schema.org/Integer",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "motif_hit_count"
}
}
},
{
"@id": "audit_columns/motif_source",
"@type": "cr:Field",
"name": "motif_source",
"description": "Generator-side motif provenance label; excluded from ordinary model features.",
"dataType": "https://schema.org/Text",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "motif_source"
}
}
},
{
"@id": "audit_columns/trigger_event_idx",
"@type": "cr:Field",
"name": "trigger_event_idx",
"description": "Internal trigger event index for delayed fraud assignment; excluded from ordinary model features.",
"dataType": "https://schema.org/Integer",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "trigger_event_idx"
}
}
},
{
"@id": "audit_columns/label_event_idx",
"@type": "cr:Field",
"name": "label_event_idx",
"description": "Internal event index at which the delayed fraud label is attached; excluded from ordinary model features.",
"dataType": "https://schema.org/Integer",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "label_event_idx"
}
}
},
{
"@id": "audit_columns/label_delay",
"@type": "cr:Field",
"name": "label_delay",
"description": "Internal delay between trigger and labeled event; excluded from ordinary model features.",
"dataType": "https://schema.org/Integer",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "label_delay"
}
}
},
{
"@id": "audit_columns/fraud_source",
"@type": "cr:Field",
"name": "fraud_source",
"description": "Internal fraud-source annotation such as motif or fallback; excluded from ordinary model features.",
"dataType": "https://schema.org/Text",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "fraud_source"
}
}
},
{
"@id": "audit_columns/dynamic_fraud_state",
"@type": "cr:Field",
"name": "dynamic_fraud_state",
"description": "Latent generator-side fraud-state variable used for mechanistic analysis; excluded from ordinary model features.",
"dataType": "https://schema.org/Float",
"source": {
"fileSet": {
"@id": "transactions-files"
},
"extract": {
"column": "dynamic_fraud_state"
}
}
}
]
},
{
"@id": "paper_suite_summary_results",
"@type": "cr:RecordSet",
"name": "paper_suite_summary_results",
"description": "Deterministic 5-seed summary results for the final paper-scale Temporal Twins suite.",
"field": [
{
"@id": "paper_suite_summary_results/benchmark_group",
"@type": "cr:Field",
"name": "benchmark_group",
"description": "Benchmark slice summarized in the row, e.g. oracle_calib, easy, medium, or hard.",
"dataType": "https://schema.org/Text",
"source": {
"fileObject": {
"@id": "paper-suite-summary-csv"
},
"extract": {
"column": "benchmark_group"
}
}
},
{
"@id": "paper_suite_summary_results/matched_eval_pairs_mean",
"@type": "cr:Field",
"name": "matched_eval_pairs_mean",
"description": "Mean number of matched-prefix evaluation pairs across seeds.",
"dataType": "https://schema.org/Float",
"source": {
"fileObject": {
"@id": "paper-suite-summary-csv"
},
"extract": {
"column": "matched_eval_pairs_mean"
}
}
},
{
"@id": "paper_suite_summary_results/static_agg_auc_mean",
"@type": "cr:Field",
"name": "static_agg_auc_mean",
"description": "Mean ROC-AUC of the static aggregate shortcut audit.",
"dataType": "https://schema.org/Float",
"source": {
"fileObject": {
"@id": "paper-suite-summary-csv"
},
"extract": {
"column": "static_agg_auc_mean"
}
}
},
{
"@id": "paper_suite_summary_results/audit_roc_auc_mean",
"@type": "cr:Field",
"name": "audit_roc_auc_mean",
"description": "Mean oracle or probe ROC-AUC depending on benchmark mode.",
"dataType": "https://schema.org/Float",
"source": {
"fileObject": {
"@id": "paper-suite-summary-csv"
},
"extract": {
"column": "audit_roc_auc_mean"
}
}
},
{
"@id": "paper_suite_summary_results/raw_roc_auc_mean",
"@type": "cr:Field",
"name": "raw_roc_auc_mean",
"description": "Mean raw motif oracle or probe ROC-AUC depending on benchmark mode.",
"dataType": "https://schema.org/Float",
"source": {
"fileObject": {
"@id": "paper-suite-summary-csv"
},
"extract": {
"column": "raw_roc_auc_mean"
}
}
},
{
"@id": "paper_suite_summary_results/xgb_roc_auc_mean",
"@type": "cr:Field",
"name": "xgb_roc_auc_mean",
"description": "Mean XGBoost ROC-AUC across seeds.",
"dataType": "https://schema.org/Float",
"source": {
"fileObject": {
"@id": "paper-suite-summary-csv"
},
"extract": {
"column": "xgb_roc_auc_mean"
}
}
},
{
"@id": "paper_suite_summary_results/static_gnn_roc_auc_mean",
"@type": "cr:Field",
"name": "static_gnn_roc_auc_mean",
"description": "Mean StaticGNN ROC-AUC across seeds.",
"dataType": "https://schema.org/Float",
"source": {
"fileObject": {
"@id": "paper-suite-summary-csv"
},
"extract": {
"column": "static_gnn_roc_auc_mean"
}
}
},
{
"@id": "paper_suite_summary_results/seqgru_clean_roc_auc_mean",
"@type": "cr:Field",
"name": "seqgru_clean_roc_auc_mean",
"description": "Mean clean SeqGRU ROC-AUC across seeds.",
"dataType": "https://schema.org/Float",
"source": {
"fileObject": {
"@id": "paper-suite-summary-csv"
},
"extract": {
"column": "seqgru_clean_roc_auc_mean"
}
}
},
{
"@id": "paper_suite_summary_results/seqgru_shuffle_delta_mean",
"@type": "cr:Field",
"name": "seqgru_shuffle_delta_mean",
"description": "Mean change in SeqGRU ROC-AUC under shuffled event order.",
"dataType": "https://schema.org/Float",
"source": {
"fileObject": {
"@id": "paper-suite-summary-csv"
},
"extract": {
"column": "seqgru_shuffle_delta_mean"
}
}
},
{
"@id": "paper_suite_summary_results/tgn_clean_roc_auc_mean",
"@type": "cr:Field",
"name": "tgn_clean_roc_auc_mean",
"description": "Mean TGN ROC-AUC across seeds.",
"dataType": "https://schema.org/Float",
"source": {
"fileObject": {
"@id": "paper-suite-summary-csv"
},
"extract": {
"column": "tgn_clean_roc_auc_mean"
}
}
},
{
"@id": "paper_suite_summary_results/tgat_clean_roc_auc_mean",
"@type": "cr:Field",
"name": "tgat_clean_roc_auc_mean",
"description": "Mean TGAT ROC-AUC across seeds.",
"dataType": "https://schema.org/Float",
"source": {
"fileObject": {
"@id": "paper-suite-summary-csv"
},
"extract": {
"column": "tgat_clean_roc_auc_mean"
}
}
},
{
"@id": "paper_suite_summary_results/dyrep_clean_roc_auc_mean",
"@type": "cr:Field",
"name": "dyrep_clean_roc_auc_mean",
"description": "Mean DyRep ROC-AUC across seeds.",
"dataType": "https://schema.org/Float",
"source": {
"fileObject": {
"@id": "paper-suite-summary-csv"
},
"extract": {
"column": "dyrep_clean_roc_auc_mean"
}
}
},
{
"@id": "paper_suite_summary_results/jodie_clean_roc_auc_mean",
"@type": "cr:Field",
"name": "jodie_clean_roc_auc_mean",
"description": "Mean JODIE ROC-AUC across seeds.",
"dataType": "https://schema.org/Float",
"source": {
"fileObject": {
"@id": "paper-suite-summary-csv"
},
"extract": {
"column": "jodie_clean_roc_auc_mean"
}
}
}
]
}
],
"rai:dataLimitations": [
"Temporal Twins is fully synthetic and is not representative of real UPI fraud prevalence, transaction mix, or institutional controls.",
"The benchmark is designed to isolate temporal-order reasoning under matched static controls rather than to reproduce a production fraud environment.",
"Standard-mode probe scores are informative benchmark probes, not upper bounds on real-world fraud detectability."
],
"rai:dataBiases": [
"Behavioral patterns are simulator-defined and reflect the assumptions of the Temporal Twins generator rather than observed user behavior.",
"Difficulty slices intentionally reshape motif strength, noise, delay, and adversarial perturbations, so conclusions should be interpreted as benchmark-relative rather than population-representative."
],
"rai:personalSensitiveInformation": "None. The dataset contains no real UPI data, no real users, no real bank accounts, no real transactions, no personal financial records, and no protected demographic attributes.",
"rai:dataUseCases": [
"Intended for temporal machine learning benchmark research, including sequence models, dynamic graph models, matched-control evaluation, and shortcut auditing.",
"Suitable for studying whether a model uses causal temporal order rather than static transaction summaries."
],
"rai:dataSocialImpact": "Positive use may include more rigorous evaluation of temporal fraud-detection methods under matched static controls. Potential misuse includes treating synthetic behavior as if it were real user behavior or using the dataset to justify deployment decisions without external validation on real, appropriately governed data.",
"rai:hasSyntheticData": true,
"prov:wasGeneratedBy": {
"@type": "prov:Activity",
"name": "Temporal Twins synthetic UPI transaction generator",
"description": "Synthetic benchmark generation for oracle_calib, easy, medium, and hard using deterministic seeds [0, 1, 2, 3, 4], num_users=350, simulation_days=45, fast_mode=false, and n_checkpoints=8. The generator emits matched fraud/benign twins evaluated at matched local prefix indices and preserves paper-suite shortcut audits and summary results.",
"prov:used": [
{
"@type": "prov:Entity",
"name": "Temporal Twins benchmark code repository",
"url": "https://huggingface.co/temporal-twins-benchmark/temporal-twins-code",
"license": "https://www.apache.org/licenses/LICENSE-2.0",
"identifier": "Apache-2.0"
},
{
"@type": "prov:Entity",
"name": "Temporal Twins paper",
"description": "Not available during double-blind review; to be added after publication."
}
]
}
}
|