Opinion Alignment on Wahl-O-Mat (WoM) March 2025 (test)
53.21Mean Macro-F1SFT+GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| SFT+GRPOBase model=Magistral 24B2026.03 | 53.21 | |
| SFT+GRPOBase model=Llama 3.1 8B2026.03 | 52.53 | |
| SFTBase model=Magistral 24B2026.03 | 51.86 | |
| GRPOBase model=Magistral 24B2026.03 | 51 | |
| SFT+GRPOBase model=Qwen3 8B2026.03 | 49.38 | |
| SFTBase model=Llama 3.1 8B2026.03 | 48.95 | |
| ORPOBase model=Llama 3.1 8B2026.03 | 43.29 | |
| SFTBase model=Qwen3 8B2026.03 | 42.91 | |
| GRPOBase model=Llama 3.1 8B2026.03 | 37.29 | |
| randomBase model=Untrained baselines2026.03 | 33.33 | |
| GRPOBase model=Qwen3 8B2026.03 | 31.42 | |
| iclBase model=Llama 3.1 8B2026.03 | 28.17 | |
| majorityBase model=Untrained baselines2026.03 | 27.44 | |
| iclBase model=Qwen3 8B2026.03 | 26.19 | |
| iclBase model=Magistral 24B2026.03 | 26.19 | |
| ORPOBase model=Qwen3 8B2026.03 | 25.25 | |
| ORPOBase model=Magistral 24B2026.03 | 24.73 |