Open-Ended Generation on Social Bias Tasks (test)
0.47Disability Status ScoreNous-Hermes-2-Mistral-7B-DPO
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Nous-Hermes-2-Mistral-7B-DPOSteering Method=Raw, Decoding Strategy=greedy decoding2025.09 | 0.47 | 0.525 | 0.52 | 0.62 | 0.525 | 0.435 | 0.49 | 0.42 | — | — | |
| DeepSeek-R1-Distill-Llama-8BSteering Method=Raw, Decoding Strategy=greedy decoding2025.09 | 0.3859 | 0.445 | 0.405 | 0.454 | 0.4975 | 0.4575 | 0.5125 | 0.4708 | — | — | |
| iPASwoModel=DeepSeek-R1-Distill-Llama-8B, Steering Strength={15, 19}, Intervention Layer={1, 4}, Decoding Strategy=greedy decoding2025.09 | 0.045 | 0.115 | 0.0825 | 0.0667 | 0.1275 | 0.0675 | 0.0925 | 0.1458 | — | — | |
| iPASaModel=DeepSeek-R1-Distill-Llama-8B, Steering Strength={15, 19}, Intervention Layer={1, 4}, Decoding Strategy=greedy decoding2025.09 | 0.0289 | -0.0025 | 0.0225 | 0.0476 | 0.0325 | 0.015 | -0.005 | 0.0625 | — | — | |
| iPASwoModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Strength={15, 19}, Intervention Layer={1, 4}, Decoding Strategy=greedy decoding2025.09 | 0.025 | 0.08 | 0.125 | 0.07 | 0.08 | 0.215 | 0.225 | 0.05 | — | — | |
| iPASaModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Strength={15, 19}, Intervention Layer={1, 4}, Decoding Strategy=greedy decoding2025.09 | -0.015 | 0.02 | 0.07 | 0.045 | 0.02 | 0.04 | 0 | 0.005 | — | — |