Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Utility Preservation on User Fine-tuning Dataset
Loading...
77.5
Final FA
Buffer-and-Reinforce (ours)
26.852
40.001
53.15
66.299
May 23, 2026
Final FA
Updated 8d ago
Evaluation Results
Method
Method
Links
Final FA
Buffer-and-Reinforce (ours)
Model=LLaMA3-8B-Instruct
2026.05
77.5
Buffer-and-Reinforce (ours)
Model=LLaMA3-8B-LAT
2026.05
75.2
Base
Model=LLaMA3-8B-LAT
2026.05
73.5
Buffer-and-Reinforce (ours)
Model=LLaMA3-8B-ReFAT
2026.05
71.1
SFT
Model=LLaMA3-8B-LAT
2026.05
70.6
SFT
Model=LLaMA3-8B-ReFAT
2026.05
68.9
SFT
Model=LLaMA3-8B-Instruct
2026.05
68.4
Base
Model=LLaMA3-8B-Instruct
2026.05
62.8
Base
Model=LLaMA3-8B-ReFAT
2026.05
40.2
Buffer-and-Reinforce (ours)
Model=LLaMA2-13B-Chat
2026.05
35.4
SFT
Model=LLaMA2-13B-Chat
2026.05
33.1
Base
Model=LLaMA2-13B-Chat
2026.05
28.8
Feedback
Search any
task
Search any
task