Helpfulness Evaluation on DNA
70.1Win RateSHARD
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SHARDReframing condition=SHARD, Backbone=Qwen3.5 27B2026.06 | 70.1 | 3.7 | 66.4 | |
| SHARDReframing condition=SHARD, Backbone=Qwen3.5 9B2026.06 | 63.3 | 11.8 | 51.5 | |
| SHARDReframing condition=SHARD, Backbone=Llama 3.3 70B2026.06 | 59.7 | 18.2 | 41.5 | |
| SHARDReframing condition=SHARD, Backbone=Llama 3.1 8B2026.06 | 54 | 37.2 | 16.8 | |
| SHARDReframing condition=SHARD, Backbone=Mistral 24B2026.06 | 52.6 | 30.4 | 22.2 | |
| SHARDReframing condition=SHARD, Backbone=Phi-42026.06 | 50.3 | 20.6 | 29.7 | |
| Benign IntentReframing condition=Benign Intent, Backbone=Phi-42026.06 | 47.9 | 25.3 | 22.6 | |
| Benign IntentReframing condition=Benign Intent, Backbone=Qwen3.5 9B2026.06 | 44.8 | 34 | 10.8 | |
| Benign IntentReframing condition=Benign Intent, Backbone=Qwen3.5 27B2026.06 | 44.6 | 34 | 10.6 | |
| Benign IntentReframing condition=Benign Intent, Backbone=Llama 3.3 70B2026.06 | 39.8 | 34.1 | 5.7 | |
| Benign IntentReframing condition=Benign Intent, Backbone=Mistral 7B2026.06 | 37.5 | 43.7 | -6.2 | |
| Benign IntentReframing condition=Benign Intent, Backbone=Mistral 24B2026.06 | 32.7 | 42.2 | -9.5 | |
| Benign IntentReframing condition=Benign Intent, Backbone=Llama 3.1 8B2026.06 | 30.2 | 37 | -6.8 | |
| SHARDReframing condition=SHARD, Backbone=Mistral 7B2026.06 | 22.9 | 55.2 | -32.3 |