File size: 22,117 Bytes
ae35a2a 8d422bb ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a 9bf7d58 ae35a2a | 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 | {
"_about": "One example per ability of the strands decider (sources: the strands-decider README and docs/vision.md). `measured` blocks were recorded through app.answer() on the stated model and date -- never typed by hand.",
"families": [
{
"id": "pick",
"label": "Pick",
"blurb": "choice: one of N options, with probabilities"
},
{
"id": "rate",
"label": "Rate",
"blurb": "score: an ordered rubric, read as an expected value"
},
{
"id": "yesno",
"label": "Yes-No",
"blurb": "noul: P(true)"
},
{
"id": "agentic",
"label": "Agentic",
"blurb": "routing, tool checks, guardrails, rules"
},
{
"id": "vision",
"label": "Vision",
"blurb": "over pixels · Qwen3.5 models only"
},
{
"id": "code",
"label": "Code",
"blurb": "yes/no over a function"
}
],
"examples": [
{
"id": "triage-choice",
"family": "pick",
"title": "triage choice",
"golden": true,
"shows": "Route a support message to the team that should own it.",
"teaches": "A choice answer is a distribution, not a label: billing wins, 23% of the mass sits elsewhere, and the confidence (0.659) is what you act on.",
"state": "Help! My payouts have been failing for 3 days!",
"questions": [
{
"type": "choice",
"instructions": "Which team should handle this?",
"options": [
"billing",
"sales",
"retail"
]
}
],
"expected_shape": "choice -> billing (confidence ~0.659); billing ~0.773, sales ~0.121, retail ~0.106 (README, cpu)",
"measured": {
"label": "measured on cpu, strands-decider-2B-qwen3.5-v1-2610, 2026-10-10 02:20Z",
"model": "strands-decider-2B-qwen3.5-v1-2610",
"device": "cpu",
"torch": "2.14.1",
"question_ids": [
"choice_0"
],
"answers": {
"choice_0": {
"type": "choice",
"display": "billing",
"value": "billing",
"probs": {
"billing": 0.7728,
"sales": 0.1213,
"retail": 0.106
},
"confidence": 0.6591,
"vote": "billing"
}
},
"input_tokens": 68,
"latency_ms": 329.141,
"cold": true
}
},
{
"id": "urgency-noul",
"family": "yesno",
"title": "urgency yes/no",
"golden": true,
"shows": "Is this message urgent?",
"teaches": "noul returns only P(true): with two outcomes the probability already is the uncertainty.",
"state": "Help! My payouts have been failing for 3 days!",
"questions": [
{
"type": "noul",
"instructions": "Does this convey urgency?",
"options": []
}
],
"expected_shape": "noul ~0.858 (README, cpu)",
"measured": {
"label": "measured on cpu, strands-decider-2B-qwen3.5-v1-2610, 2026-10-10 02:20Z",
"model": "strands-decider-2B-qwen3.5-v1-2610",
"device": "cpu",
"torch": "2.14.1",
"question_ids": [
"noul_0"
],
"answers": {
"noul_0": {
"type": "noul",
"display": "0.8581",
"value": 0.8581,
"probs": {
"true": 0.8581,
"false": 0.14190000000000003
},
"confidence": null,
"vote": "true"
}
},
"input_tokens": 86,
"latency_ms": 262.437,
"cold": false
}
},
{
"id": "frustration-score",
"family": "rate",
"title": "frustration score",
"golden": true,
"shows": "Rate the writer on calm < frustrated < depressed.",
"teaches": "A score is the expected value over the ordered levels you type, so the levels are the scale.",
"state": "Help! My payouts have been failing for 3 days!",
"questions": [
{
"type": "score",
"instructions": "How frustrated is the writer?",
"options": [
"calm",
"frustrated",
"depressed"
]
}
],
"expected_shape": "score ~1.03 (confidence ~0.509); calm ~0.196, frustrated ~0.574, depressed ~0.230 (README, cpu)",
"measured": {
"label": "measured on cpu, strands-decider-2B-qwen3.5-v1-2610, 2026-10-10 02:20Z",
"model": "strands-decider-2B-qwen3.5-v1-2610",
"device": "cpu",
"torch": "2.14.1",
"question_ids": [
"score_0"
],
"answers": {
"score_0": {
"type": "score",
"display": "1.0345",
"value": 1.0345,
"probs": {
"calm": 0.1958,
"frustrated": 0.5738,
"depressed": 0.2303
},
"confidence": 0.5091,
"legend": {
"0": "calm",
"1": "frustrated",
"2": "depressed"
},
"vote": "frustrated"
}
},
"input_tokens": 84,
"latency_ms": 256.475,
"cold": false
}
},
{
"id": "multi-question",
"family": "pick",
"title": "three questions, one read",
"shows": "All three questions on one state, in one request.",
"teaches": "The state is encoded once, so extra questions about the same text are nearly free.",
"state": "Help! My payouts have been failing for 3 days!",
"questions": [
{
"type": "noul",
"instructions": "Does this convey urgency?",
"options": []
},
{
"type": "choice",
"instructions": "Which team should handle this?",
"options": [
"billing",
"sales",
"retail"
]
},
{
"type": "score",
"instructions": "How frustrated is the writer?",
"options": [
"calm",
"frustrated",
"depressed"
]
}
],
"expected_shape": "noul ~0.858 · choice billing (~0.659) · score ~1.03 -- the same numbers as the three single-question examples",
"measured": {
"label": "measured on cpu, strands-decider-2B-qwen3.5-v1-2610, 2026-10-10 02:20Z",
"model": "strands-decider-2B-qwen3.5-v1-2610",
"device": "cpu",
"torch": "2.14.1",
"question_ids": [
"noul_0",
"choice_0",
"score_0"
],
"answers": {
"noul_0": {
"type": "noul",
"display": "0.8581",
"value": 0.8581,
"probs": {
"true": 0.8581,
"false": 0.14190000000000003
},
"confidence": null,
"vote": "true"
},
"choice_0": {
"type": "choice",
"display": "billing",
"value": "billing",
"probs": {
"billing": 0.7728,
"sales": 0.1213,
"retail": 0.106
},
"confidence": 0.6591,
"vote": "billing"
},
"score_0": {
"type": "score",
"display": "1.0345",
"value": 1.0345,
"probs": {
"calm": 0.1958,
"frustrated": 0.5738,
"depressed": 0.2303
},
"confidence": 0.5091,
"legend": {
"0": "calm",
"1": "frustrated",
"2": "depressed"
},
"vote": "frustrated"
}
},
"input_tokens": 196,
"latency_ms": 604.615,
"cold": false
}
},
{
"id": "model-routing",
"family": "agentic",
"title": "model routing",
"shows": "Pick the LLM for a task: small and fast, or large and careful.",
"teaches": "Options come from the request, not the weights: rename them to your own tiers.",
"state": "User request: \"Draft a 3-page legal memo comparing the indemnification clauses in these two vendor contracts and flag any asymmetric liability caps.\"",
"questions": [
{
"type": "choice",
"instructions": "Which model should handle this request?",
"options": [
"small fast model",
"large careful model"
]
}
],
"expected_shape": "choice -> one of the two tiers, with P(each) and a confidence",
"measured": {
"label": "measured on cpu, strands-decider-2B-qwen3.5-v1-2610, 2026-10-10 02:20Z",
"model": "strands-decider-2B-qwen3.5-v1-2610",
"device": "cpu",
"torch": "2.14.1",
"question_ids": [
"choice_0"
],
"answers": {
"choice_0": {
"type": "choice",
"display": "large careful model",
"value": "large careful model",
"probs": {
"small fast model": 0.0925,
"large careful model": 0.9075
},
"confidence": 0.8149,
"vote": "large careful model"
}
},
"input_tokens": 85,
"latency_ms": 239.644,
"cold": false
}
},
{
"id": "argument-check",
"family": "agentic",
"title": "argument check",
"shows": "Verify a tool call's arguments before it runs.",
"teaches": "JSON goes in as state; gate the call on P(true) with your own threshold.",
"state": "Tool: send_email(to, subject, body)\nCall: {\"to\": \"\", \"subject\": \"Q3 invoice\", \"body\": \"Hi, attached is the Q3 invoice. Best, Sam\"}",
"questions": [
{
"type": "noul",
"instructions": "Are the arguments complete and ready to execute?",
"options": []
}
],
"expected_shape": "noul P(true) -- low, since `to` is empty",
"measured": {
"label": "measured on cpu, strands-decider-2B-qwen3.5-v1-2610, 2026-10-10 02:20Z",
"model": "strands-decider-2B-qwen3.5-v1-2610",
"device": "cpu",
"torch": "2.14.1",
"question_ids": [
"noul_0"
],
"answers": {
"noul_0": {
"type": "noul",
"display": "0.1268",
"value": 0.1268,
"probs": {
"true": 0.1268,
"false": 0.8732
},
"confidence": null,
"vote": "false"
}
},
"input_tokens": 120,
"latency_ms": 277.906,
"cold": false
}
},
{
"id": "grounding-guardrail",
"family": "agentic",
"title": "grounding check",
"shows": "Is a claim supported by the passage it cites?",
"teaches": "With evidence and claim in one state, noul is a hallucination detector in one pass.",
"state": "Passage: The clinic opens at 8:00 on weekdays and 9:00 on Saturdays. It is closed on Sundays and public holidays.\nClaim: The clinic is open on Sundays from 9:00.",
"questions": [
{
"type": "noul",
"instructions": "Is the claim supported by the passage?",
"options": []
}
],
"expected_shape": "noul P(true) -- low: the passage says closed on Sundays",
"measured": {
"label": "measured on cpu, strands-decider-2B-qwen3.5-v1-2610, 2026-10-10 02:20Z",
"model": "strands-decider-2B-qwen3.5-v1-2610",
"device": "cpu",
"torch": "2.14.1",
"question_ids": [
"noul_0"
],
"answers": {
"noul_0": {
"type": "noul",
"display": "0.0423",
"value": 0.0423,
"probs": {
"true": 0.0423,
"false": 0.9577
},
"confidence": null,
"vote": "false"
}
},
"input_tokens": 125,
"latency_ms": 285.565,
"cold": false
}
},
{
"id": "rule-application",
"family": "agentic",
"title": "apply a policy",
"shows": "A written policy plus a case: which outcome does the policy dictate?",
"teaches": "Rules written in plain words apply at request time, nothing to retrain.",
"state": "Policy: Refunds are approved automatically within 14 days of purchase. Between 15 and 30 days a refund needs a manager's approval. After 30 days refunds are denied unless the item arrived defective.\nCase: Purchased 22 days ago, item works, customer changed their mind.",
"questions": [
{
"type": "choice",
"instructions": "What does the policy say should happen?",
"options": [
"approve automatically",
"needs manager approval",
"deny"
]
}
],
"expected_shape": "choice -> needs manager approval (22 days falls in the 15-30 band)",
"measured": {
"label": "measured on cpu, strands-decider-2B-qwen3.5-v1-2610, 2026-10-10 02:20Z",
"model": "strands-decider-2B-qwen3.5-v1-2610",
"device": "cpu",
"torch": "2.14.1",
"question_ids": [
"choice_0"
],
"answers": {
"choice_0": {
"type": "choice",
"display": "needs manager approval",
"value": "needs manager approval",
"probs": {
"approve automatically": 0.0292,
"needs manager approval": 0.9456,
"deny": 0.0252
},
"confidence": 0.9184,
"vote": "needs manager approval"
}
},
"input_tokens": 126,
"latency_ms": 288.074,
"cold": false
}
},
{
"id": "language-id",
"family": "pick",
"title": "language id",
"shows": "Which language is 'sihamba ngokushesha'?",
"teaches": "Options can be anything nameable; a label set of languages works like a label set of teams.",
"state": "sihamba ngokushesha",
"questions": [
{
"type": "choice",
"instructions": "What language is this phrase in?",
"options": [
"English",
"Zulu",
"Dutch"
]
}
],
"expected_shape": "choice -> Zulu",
"measured": {
"label": "measured on cpu, strands-decider-2B-qwen3.5-v1-2610, 2026-10-10 02:20Z",
"model": "strands-decider-2B-qwen3.5-v1-2610",
"device": "cpu",
"torch": "2.14.1",
"question_ids": [
"choice_0"
],
"answers": {
"choice_0": {
"type": "choice",
"display": "Zulu",
"value": "Zulu",
"probs": {
"English": 0.0444,
"Zulu": 0.9112,
"Dutch": 0.0444
},
"confidence": 0.8668,
"vote": "Zulu"
}
},
"input_tokens": 66,
"latency_ms": 244.617,
"cold": false
}
},
{
"id": "judge-score",
"family": "rate",
"title": "judge an answer",
"shows": "Score a model's answer on a rubric from 0 to 4.",
"teaches": "An ordered rubric makes the decider a cheap judge: the expected value is the grade.",
"state": "Question: What is the capital of Australia?\nModel answer: The capital of Australia is Sydney, its largest city.\nRubric: 4 = correct and complete, 3 = correct with a minor flaw, 2 = partly correct, 1 = mostly wrong, 0 = wrong or no answer.",
"questions": [
{
"type": "score",
"instructions": "Grade the model answer against the rubric.",
"options": [
"0",
"1",
"2",
"3",
"4"
]
}
],
"expected_shape": "score near the low end (Canberra is the capital)",
"measured": {
"label": "measured on cpu, strands-decider-2B-qwen3.5-v1-2610, 2026-10-10 02:20Z",
"model": "strands-decider-2B-qwen3.5-v1-2610",
"device": "cpu",
"torch": "2.14.1",
"question_ids": [
"score_0"
],
"answers": {
"score_0": {
"type": "score",
"display": "2.3892",
"value": 2.3892,
"probs": {
"0": 0.1218,
"1": 0.156,
"2": 0.1965,
"3": 0.2626,
"4": 0.2631
},
"confidence": 0.3895,
"legend": {
"0": "0",
"1": "1",
"2": "2",
"3": "3",
"4": "4"
},
"vote": "4"
}
},
"input_tokens": 158,
"latency_ms": 330.912,
"cold": false
}
},
{
"id": "signature-signed",
"family": "vision",
"title": "signature: filled",
"vision_only": true,
"image": "form_signed.png",
"shows": "A release form whose signature block is signed and dated.",
"teaches": "The image enters the state as vision tokens; the readout and confidence are the text model's.",
"state": "",
"questions": [
{
"type": "noul",
"instructions": "Is the signature block filled in?",
"options": []
}
],
"expected_shape": "noul P(true) -- higher than on the empty form",
"measured": {
"label": "measured on cpu, strands-decider-2B-qwen3.5-v1-2610, 2026-10-10 02:20Z",
"model": "strands-decider-2B-qwen3.5-v1-2610",
"device": "cpu",
"torch": "2.14.1",
"question_ids": [
"noul_0"
],
"answers": {
"noul_0": {
"type": "noul",
"display": "0.6868",
"value": 0.6868,
"probs": {
"true": 0.6868,
"false": 0.31320000000000003
},
"confidence": null,
"vote": "true"
}
},
"input_tokens": 232,
"latency_ms": 1160.211,
"cold": true
}
},
{
"id": "signature-empty",
"family": "vision",
"title": "signature: empty",
"vision_only": true,
"image": "form_unsigned.png",
"shows": "The same form, signature block empty.",
"teaches": "The gap between the two P(true) values is what the model reads from the pixels.",
"state": "",
"questions": [
{
"type": "noul",
"instructions": "Is the signature block filled in?",
"options": []
}
],
"expected_shape": "noul P(true) -- lower than on the signed form",
"measured": {
"label": "measured on cpu, strands-decider-2B-qwen3.5-v1-2610, 2026-10-10 02:20Z",
"model": "strands-decider-2B-qwen3.5-v1-2610",
"device": "cpu",
"torch": "2.14.1",
"question_ids": [
"noul_0"
],
"answers": {
"noul_0": {
"type": "noul",
"display": "0.1344",
"value": 0.1344,
"probs": {
"true": 0.1344,
"false": 0.8656
},
"confidence": null,
"vote": "false"
}
},
"input_tokens": 232,
"latency_ms": 1057.52,
"cold": false
}
},
{
"id": "shape-choice",
"family": "vision",
"title": "which shape",
"vision_only": true,
"image": "shapes.png",
"shows": "One large shape is drawn; pick which.",
"teaches": "choice over images works like choice over text: the option tokens follow the image.",
"state": "",
"questions": [
{
"type": "choice",
"instructions": "Which shape is drawn in the image?",
"options": [
"circle",
"triangle",
"square"
]
}
],
"expected_shape": "choice -> triangle",
"measured": {
"label": "measured on cpu, strands-decider-2B-qwen3.5-v1-2610, 2026-10-10 02:20Z",
"model": "strands-decider-2B-qwen3.5-v1-2610",
"device": "cpu",
"torch": "2.14.1",
"question_ids": [
"choice_0"
],
"answers": {
"choice_0": {
"type": "choice",
"display": "triangle",
"value": "triangle",
"probs": {
"circle": 0.0119,
"triangle": 0.9766,
"square": 0.0115
},
"confidence": 0.9649,
"vote": "triangle"
}
},
"input_tokens": 256,
"latency_ms": 1128.372,
"cold": false
}
},
{
"id": "empty-list-code",
"family": "code",
"title": "empty list?",
"shows": "Does this Python function handle the empty list?",
"teaches": "Code is state; one forward pass answers a property of the function.",
"state": "def average(xs):\n return sum(xs) / len(xs)",
"questions": [
{
"type": "noul",
"instructions": "Does this function handle the empty list?",
"options": []
}
],
"expected_shape": "noul P(true) -- low: len([]) == 0 divides by zero",
"measured": {
"label": "measured on cpu, strands-decider-2B-qwen3.5-v1-2610, 2026-10-10 02:20Z",
"model": "strands-decider-2B-qwen3.5-v1-2610",
"device": "cpu",
"torch": "2.14.1",
"question_ids": [
"noul_0"
],
"answers": {
"noul_0": {
"type": "noul",
"display": "0.2203",
"value": 0.2203,
"probs": {
"true": 0.2203,
"false": 0.7797000000000001
},
"confidence": null,
"vote": "false"
}
},
"input_tokens": 91,
"latency_ms": 359.318,
"cold": false
}
}
]
}
|