Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Knowledge Retention on WMDP-chem (retain)
Loading...
56
Rt (Knowledge Retention)
Zero-shot
21.784
30.667
39.55
48.433
May 14, 2026
Rt (Knowledge Retention)
Updated 19d ago
Evaluation Results
Method
Method
Links
Rt (Knowledge Retention)
Zero-shot
Backbone=Qwen-3-8B
2026.05
56
Zero-shot
Backbone=Llama-3.1-8B-...
2026.05
53.3
LUNAR
Backbone=Qwen-3-8B
2026.05
51.8
Global GA
Backbone=Llama-3.1-8B-...
2026.05
49.1
LUNAR
Backbone=Llama-3.1-8B-...
2026.05
47.9
Surgical GA
Backbone=Llama-3.1-8B-...
2026.05
42.6
MANSU
Backbone=Llama-3.1-8B-...
2026.05
39.8
Surgical GA
Backbone=Qwen-3-8B
2026.05
39.8
MANSU
Backbone=Qwen-3-8B
2026.05
36.4
NPO
Backbone=Qwen-3-8B
2026.05
29.6
NPO
Backbone=Llama-3.1-8B-...
2026.05
26.9
Global GA
Backbone=Qwen-3-8B
2026.05
26.9
SimNPO
Backbone=Qwen-3-8B
2026.05
25.9
GU+SimNPO
Backbone=Qwen-3-8B
2026.05
25
SimNPO
Backbone=Llama-3.1-8B-...
2026.05
24.1
GU+SimNPO
Backbone=Llama-3.1-8B-...
2026.05
23.1
Feedback
Search any
task
Search any
task