Legal Reasoning on PRBench-Legal Hard 250-prompt
31.5GPT-5 Grader Score
Qwen3.5-4B RL on Agentic Self-Instruct data
Updated 1mo ago
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| 31.5 | 26.6 | ||
2026.06 | 27.7 | 22.6 | |
| 25.3 | 23.3 | ||
2026.06 | 16.7 | 14.5 |
Qwen3.5-4B RL on Agentic Self-Instruct data
| Method | Links | ||
|---|---|---|---|
| 31.5 | 26.6 | ||
2026.06 | 27.7 | 22.6 | |
| 25.3 | 23.3 | ||
2026.06 | 16.7 | 14.5 |