Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Prompt Hijacking on IHEval Prompt Hijacking - Alignment 1.0
Loading...
82.5
Accuracy
Llama3.1-8B-NSHA-DPO
33.204
46.002
58.8
71.598
Apr 10, 2026
Accuracy
Updated 6d ago
Evaluation Results
Method
Method
Links
Accuracy
Llama3.1-8B-NSHA-DPO
Backbone=Llama3.1-8B,...
2026.04
82.5
Llama3.1-8B-NS
Backbone=Llama3.1-8B,...
2026.04
70.4
Llama3.1-8B-NSHA-HCAL
Backbone=Llama3.1-8B,...
2026.04
68.5
Llama3.1-8B
Backbone=Llama3.1-8B,...
2026.04
66.3
Qwen3-4B-it-NSHA-DPO
Backbone=Qwen3-4B, Met...
2026.04
63.7
Qwen3-4B-it
Backbone=Qwen3-4B, Met...
2026.04
62.6
Qwen3-4B-it-NS
Backbone=Qwen3-4B, Met...
2026.04
61.9
Llama3.1-8B-CoT
Backbone=Llama3.1-8B,...
2026.04
59.8
Qwen3-4B-it-NSHA-HCAL
Backbone=Qwen3-4B, Met...
2026.04
58.7
Qwen3-4B-it-CoT
Backbone=Qwen3-4B, Met...
2026.04
58.2
Qwen3-4B-it-NSHA-SFT
Backbone=Qwen3-4B, Met...
2026.04
50.9
Llama3.1-8B-NSHA-SFT
Backbone=Llama3.1-8B,...
2026.04
35.1
Feedback
Search any
task
Search any
task