Reasoning on HotpotQA (Length and Degeneration Metrics)
0Degeneration CountBase
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| BaseModel Scale=0.6B2026.02 | 0 | 477 | 402 | 619 | 477 | |
| PromptModel Scale=0.6B2026.02 | 0 | 472 | 404 | 594 | 472 | |
| MBT-SModel Scale=1.7B2026.02 | 0 | 451 | 446 | 491 | 451 | |
| MBT-SModel Scale=4B2026.02 | 0 | 448 | 445 | 489 | 448 | |
| MBT-RModel Scale=4B2026.02 | 0 | 627 | 623 | 672 | 627 | |
| MBT-RModel Scale=0.6B2026.02 | 1 | 616 | 606 | 651 | 612 | |
| BaseModel Scale=1.7B2026.02 | 1 | 605 | 478 | 994 | 601 | |
| GRPOModel Scale=1.7B2026.02 | 1 | 683 | 565 | 1,327 | 679 | |
| MBT-RModel Scale=1.7B2026.02 | 1 | 643 | 633 | 694 | 639 | |
| MBT-SModel Scale=0.6B2026.02 | 3 | 472 | 453 | 531 | 459 | |
| PromptModel Scale=1.7B2026.02 | 4 | 623 | 508 | 1,030 | 606 | |
| PromptModel Scale=4B2026.02 | 10 | 556 | 453 | 1,120 | 512 | |
| BaseModel Scale=4B2026.02 | 12 | 551 | 432 | 1,217 | 499 | |
| RSModel Scale=0.6B2026.02 | 18 | 561 | 399 | 792 | 483 | |
| GRPOModel Scale=0.6B2026.02 | 28 | 1,187 | 878 | 1,991 | 1,067 | |
| RSModel Scale=1.7B2026.02 | 32 | 747 | 485 | 1,464 | 608 | |
| GRPOModel Scale=4B2026.02 | 52 | 722 | 433 | 1,747 | 495 | |
| gpt-oss-distillModel Scale=4B2026.02 | 67 | 979 | 622 | 2,387 | 689 | |
| RSModel Scale=4B2026.02 | 109 | 956 | 425 | 1,983 | 481 | |
| TokenSkipModel Scale=0.6B2026.02 | 131 | 1,058 | 407 | 1,619 | 487 | |
| LIMOProModel Scale=4B2026.02 | 139 | 913 | 295 | 1,624 | 304 | |
| LIMOProModel Scale=0.6B2026.02 | 176 | 1,071 | 288 | 1,778 | 299 | |
| TokenSkipModel Scale=4B2026.02 | 196 | 1,335 | 419 | 2,747 | 480 | |
| LIMOProModel Scale=1.7B2026.02 | 222 | 1,270 | 288 | 2,236 | 297 | |
| gpt-oss-distillModel Scale=0.6B2026.02 | 227 | 1,759 | 715 | 3,841 | 778 | |
| gpt-oss-distillModel Scale=1.7B2026.02 | 248 | 1,805 | 663 | 4,403 | 732 | |
| TokenSkipModel Scale=1.7B2026.02 | 324 | 1,959 | 476 | 4,102 | 549 |