Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Injected triggers

Benchmarks

Task NameDataset NameSOTA ResultTrend
Exact recovery of injected triggersInjected triggers Llama-3.1 70B
Trigger Count2,462
6
Exact recovery of injected triggersInjected triggers Llama-3.1 8B
Rule Adherence Count3,428
6
Exact recovery of injected triggersInjected triggers GPT-4o
Trigger Count (Rules)4,090
6
Exact recovery of injected triggersInjected triggers GPT-4o mini
Rule Adherence Count2,251
6
Exact recovery of injected triggersInjected triggers GPT-3.5 Turbo
Test Rule Count1,429
6
Showing 5 of 5 rows