Opinion Alignment on American National Election Studies (ANES) 2020 Time Series (test)
45.43Mean Macro-F1SFT+GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| SFT+GRPOBase model=Magistral 24B2026.03 | 45.43 | |
| GRPOBase model=Magistral 24B2026.03 | 43.79 | |
| SFTBase model=Llama 3.1 8B2026.03 | 42.77 | |
| SFT+GRPOBase model=Llama 3.1 8B2026.03 | 40.66 | |
| SFTBase model=Magistral 24B2026.03 | 39.15 | |
| SFT+GRPOBase model=Qwen3 8B2026.03 | 38.44 | |
| SFTBase model=Qwen3 8B2026.03 | 35.14 | |
| ORPOBase model=Llama 3.1 8B2026.03 | 34.84 | |
| GRPOBase model=Llama 3.1 8B2026.03 | 34.55 | |
| randomBase model=Untrained baselines2026.03 | 33.33 | |
| GRPOBase model=Qwen3 8B2026.03 | 31.47 | |
| ORPOBase model=Qwen3 8B2026.03 | 26.95 | |
| ORPOBase model=Magistral 24B2026.03 | 24.25 | |
| iclBase model=Llama 3.1 8B2026.03 | 23.2 | |
| iclBase model=Qwen3 8B2026.03 | 23.2 | |
| majorityBase model=Untrained baselines2026.03 | 22.98 | |
| iclBase model=Magistral 24B2026.03 | 19.23 |