Data for MultiTrait Subspace Steering
收藏官方服务:
资源简介:
Dataset consisting of model responses and evaluation outcome (2 evaluation outcome) for 3 models Llama-3.1-8B Qwen2.5-1.5B Qwen3-14B consolidated_models/├── singleturn/│ ├── Llama-3.1-8B-Instruct/│ │ ├── baseline/│ │ ├── dark_safe_d256_a1.0/│ │ └── dark_aggr_d128_a1.5/│ ├── Qwen2.5-1.5B-Instruct/│ │ ├── baseline/│ │ ├── dark_safe_d32_a1.0/│ │ └── dark_safe_d64_a1.0/│ └── Qwen3-14B/│ ├── baseline/│ ├── dark_safe_d128_a1.0/│ └── dark_safe_d256_a1.0/└── multiturn/ ├── Llama-3.1-8B-Instruct/ │ ├── baseline/ │ ├── dark_safe_d256_a1.0/ │ └── dark_aggr_d128_a1.5/ ├── Qwen2.5-1.5B-Instruct/ │ ├── baseline/ │ ├── dark_safe_d32_a1.0/ │ └── dark_safe_d64_a1.0/ └── Qwen3-14B/ ├── baseline/ ├── dark_safe_d128_a1.0/ └── dark_safe_d256_a1.0/
提供机构:
Zenodo创建时间:
2026-02-19



