Summary Evaluation on Aggregated Benchmarks
42.2Average Score (3-item avg)CARL
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| CARLModel Scale=7B-Inst.2026.05 | 42.2 | 55.5 | 1,315 | 1,614 | |
| Search-R1 PPOModel Scale=7B-Inst.2026.05 | 35.5 | — | — | 2,098 | |
| CARLModel Scale=3B-Inst.2026.05 | 35.4 | 48 | 1,477 | 1,755 | |
| Search-R1 GRPOModel Scale=7B-Inst.2026.05 | 31.5 | — | — | 2,062 | |
| Search-R1 GRPOModel Scale=3B-Inst.2026.05 | 25.7 | — | — | 2,220 | |
| SFT (rej. sampled)Model Scale=7B-Inst.2026.05 | 25.1 | 43.2 | 1,681 | 2,050 | |
| Search-R1 PPOModel Scale=3B-Inst.2026.05 | 24.8 | — | — | 2,142 | |
| CoT+Tools (always)Model Scale=7B-Inst.2026.05 | 22.9 | 43.4 | 1,776 | 2,158 | |
| CoT+Tools (selective)Model Scale=7B-Inst.2026.05 | 22.7 | 43 | 1,624 | 1,990 | |
| SFT (rej. sampled)Model Scale=3B-Inst.2026.05 | 18.2 | 32.4 | 1,780 | 2,159 | |
| CoT+Tools (always)Model Scale=3B-Inst.2026.05 | 15.4 | 34.2 | 1,907 | 2,251 | |
| CoT+Tools (selective)Model Scale=3B-Inst.2026.05 | 15.2 | 32.1 | 1,739 | 2,070 |