Question-Type Diversity Alignment on Legalbench Taxonomy
0.036Jensen Shannon Divergencegpt-oss-120b
Evaluation Results
| Method | Links | |
|---|---|---|
| gpt-oss-120bSimulator Variant=AGENT2026.03 | 0.036 | |
| gpt-oss-120bSimulator Variant=SCOTUS_DEFAULT2026.03 | 0.061 | |
| gpt4oSimulator Variant=SCOTUS_DEFAULT2026.03 | 0.069 | |
| gemini-2.5-proSimulator Variant=AGENT2026.03 | 0.072 | |
| gemini-2.5-proSimulator Variant=SCOTUS_DEFAULT2026.03 | 0.087 | |
| Llama-3.3-70B-InstructSimulator Variant=SCOTUS_DEFAULT2026.03 | 0.122 | |
| gpt4oSimulator Variant=AGENT2026.03 | 0.124 | |
| Qwen3-32BSimulator Variant=SCOTUS_DEFAULT2026.03 | 0.131 |