Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Runtime monitor execution on LabUtopia L1-L4 grounding 1.0 (source-held-out)
Loading...
23.8
Unsafe Events
LabGuard-Hybrid
23.084
27.917
32.75
37.583
Jun 30, 2026
Unsafe Events
False Interventions
Task Success Rate
Updated 25d ago
Evaluation Results
Method
Method
Links
Unsafe Events
False Interventions
Task Success Rate
LabGuard-Hybrid
2026.06
23.8
6.2
81.2
Base Grounder + Constraint decoding
Backbone=Qwen3-8B, Dec...
2026.06
26.3
6.9
79.8
Base Grounder + LoRA fine-tuning
Backbone=Qwen3-8B, Fin...
2026.06
28.7
7.5
78.4
Qwen2.5-72B
Evaluation Protocol=0-...
2026.06
32.4
8.7
75.8
DeepSeek-V3.1
Evaluation Protocol=0-...
2026.06
33.8
9.2
74.6
Regex BL
2026.06
34.2
8.3
72.4
Qwen3-32B
Evaluation Protocol=0-...
2026.06
35.6
9.8
73.2
Qwen-7B
Evaluation Protocol=0-...
2026.06
38.9
10.7
70.8
Base Grounder
Backbone=Qwen3-8B
2026.06
39.5
11.3
69.7
SciBERT
Backbone=SciBERT
2026.06
41.7
12.5
68.3
Feedback
Search any
task
Search any
task