Harmlessness evaluation on HarmfulQA
69.4Helpfulness ScoreAligner
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AlignerAligner size=7B, Upstream LLM=Claude 2, Zero-shot=true2024.02 | 69.4 | 42.1 | |
| AlignerAligner size=2B, Upstream LLM=Claude 2, Zero-shot=true2024.02 | 48 | 14.4 | |
| AlignerAligner size=13B, Upstream LLM=Claude 2, Zero-shot=true2024.02 | 45.9 | 28.6 | |
| AlignerAligner size=7B, Upstream LLM=Alpaca-7B, Zero-shot=true2024.02 | 38.2 | 70.7 | |
| AlignerAligner size=13B, Upstream LLM=Vicuna-13B, Zero-shot=true2024.02 | 30.9 | 18.9 | |
| AlignerAligner size=13B, Upstream LLM=Llama2-13B-Chat, Zero-shot=true2024.02 | 30.7 | 35 | |
| AlignerAligner size=2B, Upstream LLM=Vicuna-33B, Zero-shot=true2024.02 | 26.9 | 32.6 | |
| AlignerAligner size=13B, Upstream LLM=GPT-4, Zero-shot=true2024.02 | 25.1 | 20.1 | |
| AlignerAligner size=2B, Upstream LLM=Alpaca-7B, Zero-shot=true2024.02 | 22.6 | 65.3 | |
| AlignerAligner size=7B, Upstream LLM=Vicuna-13B, Zero-shot=true2024.02 | 21.9 | 18.9 | |
| AlignerAligner size=2B, Upstream LLM=Llama2-70B-Chat, Zero-shot=true2024.02 | 21.3 | 7.2 | |
| AlignerAligner size=13B, Upstream LLM=Vicuna-7B, Zero-shot=true2024.02 | 19.5 | 31 | |
| AlignerAligner size=2B, Upstream LLM=Llama2-13B-Chat, Zero-shot=true2024.02 | 18.6 | 27.6 | |
| AlignerAligner size=2B, Upstream LLM=Vicuna-7B, Zero-shot=true2024.02 | 17.7 | 27.1 | |
| AlignerAligner size=2B, Upstream LLM=Vicuna-13B, Zero-shot=true2024.02 | 17.1 | 16.1 | |
| AlignerAligner size=7B, Upstream LLM=GPT-4, Zero-shot=true2024.02 | 16.3 | 28 | |
| AlignerAligner size=7B, Upstream LLM=Llama2-13B-Chat, Zero-shot=true2024.02 | 15.5 | 28.6 | |
| AlignerAligner size=13B, Upstream LLM=Llama2-7B-Chat, Zero-shot=true2024.02 | 12.6 | 32.3 | |
| AlignerAligner size=2B, Upstream LLM=GPT-4, Zero-shot=true2024.02 | 12.5 | 29.2 | |
| AlignerAligner size=7B, Upstream LLM=Vicuna-7B, Zero-shot=true2024.02 | 12 | 29.3 | |
| AlignerAligner size=2B, Upstream LLM=Llama2-7B-Chat, Zero-shot=true2024.02 | 11.3 | 25.9 | |
| AlignerAligner size=7B, Upstream LLM=Beaver-7B, Zero-shot=true2024.02 | 8.9 | 6 | |
| AlignerAligner size=7B, Upstream LLM=GPT-3.5, Zero-shot=true2024.02 | 8.4 | 7 | |
| AlignerAligner size=13B, Upstream LLM=Beaver-7B, Zero-shot=true2024.02 | 8 | 11.6 | |
| AlignerAligner size=2B, Upstream LLM=Beaver-7B, Zero-shot=true2024.02 | 7.8 | 7.6 | |
| AlignerAligner size=13B, Upstream LLM=GPT-3.5, Zero-shot=true2024.02 | 7.6 | 7.7 | |
| AlignerAligner size=13B, Upstream LLM=Vicuna-33B, Zero-shot=true2024.02 | 7.3 | 33.3 | |
| AlignerAligner size=13B, Upstream LLM=Llama2-70B-Chat, Zero-shot=true2024.02 | 6.3 | 7.6 | |
| AlignerAligner size=2B, Upstream LLM=GPT-3.5, Zero-shot=true2024.02 | 3.6 | 4.4 | |
| AlignerAligner size=13B, Upstream LLM=Alpaca-7B, Zero-shot=true2024.02 | 3.4 | 75.9 | |
| AlignerAligner size=7B, Upstream LLM=Vicuna-33B, Zero-shot=true2024.02 | -1 | 33.6 | |
| AlignerAligner size=7B, Upstream LLM=Llama2-7B-Chat, Zero-shot=true2024.02 | -5.7 | 22.1 | |
| AlignerAligner size=7B, Upstream LLM=Llama2-70B-Chat, Zero-shot=true2024.02 | -6.6 | 4.1 |