Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Instruction-data separation on SEP
Loading...
85
Probe Execution Rate
SFT
1.488
23.169
44.85
66.531
Jun 9, 2026
Probe Execution Rate
Overall Response Rate
Non-Trivial Response Rate
Updated 1mo ago
Evaluation Results
Method
Method
Links
Probe Execution Rate
Overall Response Rate
Non-Trivial Response Rate
SFT
Configuration=trained-...
2026.06
85
93.6
28.9
Llama-3.1-Inst.
Configuration=off-the-...
2026.06
69.3
67.5
93.3
baseline
Configuration=trained-...
2026.06
47.6
57.7
72.5
3-level (lin.)
Configuration=trained-...
2026.06
47.2
47.1
98.8
GW-DPO (lin.)
Configuration=trained-...
2026.06
13.9
13.9
98.5
DPO
Configuration=trained-...
2026.06
13.3
13.1
94.8
GW-DPO (bi.)
Configuration=trained-...
2026.06
11.8
11.7
98.3
No-ISE
Configuration=trained-...
2026.06
4.7
2.9
45.7
Feedback
Search any
task
Search any
task