Logical Reasoning on Logic Benchmarks
52.1SynLogic ScoreEvolve
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| EvolveData@Step=Evolve@240, Setting=(C) Training on Qwen3-8B(Thinking)2026.01 | 52.1 | 7.9 | 89.3 | 31.1 | 39.5 | 1.7 | |
| Qwen3-8B(Thinking)Setting=(C) Training on Qwen3-8B(Thinking)2026.01 | 51.1 | 7.6 | 89.3 | 30 | 39.4 | — | |
| SeedData@Step=Seed@240, Setting=(C) Training on Qwen3-8B(Thinking)2026.01 | 49.4 | 3.8 | 89.2 | 30.1 | 38.9 | — | |
| +SSLogic-EvolveData@Step=+SSLogic-Evolve@240, Setting=(B) Synergistic Mathematical Training with Data-Mixing Ladder2026.01 | 22.1 | 1.9 | 79.5 | 16.6 | 18.5 | 3 | |
| +ARC-AGI+EvolveData@Step=+ARC-AGI+Evolve@240, Setting=(B) Synergistic Mathematical Training with Data-Mixing Ladder2026.01 | 20 | 11.2 | 76.1 | 15.6 | 15.4 | 3 | |
| +SSLogic-EvolveData@Step=+SSLogic-Evolve@200, Setting=(B) Synergistic Mathematical Training with Data-Mixing Ladder2026.01 | 19.5 | 2.9 | 78.9 | 16.1 | 16.3 | 1.4 | |
| +ARC-AGI+EvolveData@Step=+ARC-AGI+Evolve@200, Setting=(B) Synergistic Mathematical Training with Data-Mixing Ladder2026.01 | 19.5 | 8.6 | 75.9 | 15.1 | 14.6 | 1.4 | |
| EvolveData@Step=Evolve@240, Setting=(A) Seed vs. Evolve under Fixed Training Steps2026.01 | 18.7 | 3.1 | 75.5 | 15 | 13.3 | 2 | |
| +SSLogic-EvolveData@Step=+SSLogic-Evolve@160, Setting=(B) Synergistic Mathematical Training with Data-Mixing Ladder2026.01 | 18.5 | 1.9 | 77.8 | 16 | 15.6 | 1.1 | |
| AIMEData@Step=AIME@160, Setting=(B) Synergistic Mathematical Training with Data-Mixing Ladder2026.01 | 16.7 | 3.8 | 75.2 | 14.8 | 13.7 | — | |
| +ARC-AGI+EvolveData@Step=+ARC-AGI+Evolve@160, Setting=(B) Synergistic Mathematical Training with Data-Mixing Ladder2026.01 | 16.7 | 9.3 | 74.6 | 14.4 | 14.1 | 1 | |
| AIMEData@Step=AIME@200, Setting=(B) Synergistic Mathematical Training with Data-Mixing Ladder2026.01 | 16.5 | 5.3 | 76.1 | 14.6 | 14.1 | — | |
| EvolveData@Step=Evolve@200, Setting=(A) Seed vs. Evolve under Fixed Training Steps2026.01 | 16.1 | 5 | 74.8 | 14.8 | 12.9 | 1.7 | |
| EvolveData@Step=Evolve@160, Setting=(A) Seed vs. Evolve under Fixed Training Steps2026.01 | 14.8 | 3.8 | 74.2 | 14.7 | 13.4 | 0.6 | |
| SeedData@Step=Seed@160, Setting=(A) Seed vs. Evolve under Fixed Training Steps2026.01 | 14.6 | 3.6 | 72.9 | 13.5 | 13.4 | — | |
| AIMEData@Step=AIME@240, Setting=(B) Synergistic Mathematical Training with Data-Mixing Ladder2026.01 | 13.3 | 3.8 | 76.5 | 15.4 | 14.4 | — | |
| SeedData@Step=Seed@240, Setting=(A) Seed vs. Evolve under Fixed Training Steps2026.01 | 13.1 | 3.1 | 72.1 | 13.6 | 13.9 | — | |
| SeedData@Step=Seed@200, Setting=(A) Seed vs. Evolve under Fixed Training Steps2026.01 | 12.9 | 1.7 | 73.5 | 13.3 | 13.5 | — | |
| Qwen3-8B-BaseSetting=(C) Training on Qwen3-8B(Thinking)2026.01 | 8.4 | 1.4 | 60.9 | 10.1 | 7.5 | — |