Reasoning on 2WikiMultiHopQA (OOD)
0Degeneration CountPrompt
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| PromptModel Scale=1.7B2026.02 | 0 | 544 | 418 | 907 | 544 | |
| MBT-SModel Scale=4B2026.02 | 0 | 460 | 456 | 491 | 460 | |
| MBT-RModel Scale=4B2026.02 | 0 | 649 | 643 | 692 | 649 | |
| BaseModel Scale=0.6B2026.02 | 1 | 569 | 427 | 719 | 566 | |
| PromptModel Scale=0.6B2026.02 | 1 | 555 | 426 | 681 | 552 | |
| GRPOModel Scale=1.7B2026.02 | 1 | 637 | 473 | 1,314 | 634 | |
| BaseModel Scale=1.7B2026.02 | 2 | 560 | 397 | 927 | 555 | |
| MBT-SModel Scale=1.7B2026.02 | 4 | 469 | 450 | 526 | 459 | |
| BaseModel Scale=4B2026.02 | 4 | 501 | 394 | 1,128 | 491 | |
| MBT-RModel Scale=1.7B2026.02 | 8 | 684 | 655 | 757 | 664 | |
| PromptModel Scale=4B2026.02 | 12 | 535 | 418 | 1,114 | 504 | |
| MBT-RModel Scale=0.6B2026.02 | 13 | 682 | 635 | 764 | 649 | |
| MBT-SModel Scale=0.6B2026.02 | 14 | 505 | 454 | 595 | 469 | |
| RSModel Scale=1.7B2026.02 | 33 | 683 | 408 | 1,432 | 599 | |
| RSModel Scale=0.6B2026.02 | 50 | 703 | 430 | 957 | 575 | |
| GRPOModel Scale=0.6B2026.02 | 59 | 1,409 | 920 | 2,359 | 1,261 | |
| GRPOModel Scale=4B2026.02 | 121 | 836 | 426 | 2,274 | 526 | |
| RSModel Scale=4B2026.02 | 121 | 815 | 397 | 1,798 | 505 | |
| gpt-oss-distillModel Scale=4B2026.02 | 193 | 1,271 | 670 | 3,450 | 780 | |
| TokenSkipModel Scale=4B2026.02 | 426 | 1,601 | 390 | 3,361 | 508 | |
| LIMOProModel Scale=0.6B2026.02 | 458 | 1,509 | 298 | 2,157 | 328 | |
| TokenSkipModel Scale=0.6B2026.02 | 602 | 2,182 | 451 | 3,196 | 644 | |
| LIMOProModel Scale=1.7B2026.02 | 629 | 1,953 | 307 | 3,090 | 331 | |
| gpt-oss-distillModel Scale=1.7B2026.02 | 664 | 2,557 | 760 | 6,499 | 873 | |
| gpt-oss-distillModel Scale=0.6B2026.02 | 768 | 2,849 | 748 | 5,824 | 903 | |
| LIMOProModel Scale=4B2026.02 | 810 | 2,434 | 317 | 4,050 | 346 | |
| TokenSkipModel Scale=1.7B2026.02 | 1,034 | 3,221 | 425 | 7,257 | 574 |