List your company
ResearchFormerly WizWand

Classification on JAILBREAK (test)

98.8AccuracyACT
56.99267.84678.789.554Sep 30, 2025
Updated 6mo ago

Evaluation results

MethodLinks
ACTTree Depth (d)=4, Optimization Steps (k)=20, LLM Backbone=GPT-4.1 Mini2025.0998.898.8
Rule FitLLM Backbone=Avg2025.0996.896.8
ACTTree Depth (d)=3, Optimization Steps (k)=10, LLM Backbone=GPT-4.1 Mini2025.0995.295.5
DSPyEvaluation Protocol=BFSR, 8 demos, LLM Backbone=GPT-4.1 Mini2025.0994.894.8
TEXTGRADLLM Backbone=GPT-4.1 Mini2025.0994.895.1
CoTEvaluation Protocol=0-shot, LLM Backbone=GPT-4.1 Mini2025.0993.693.3
ACTTree Depth (d)=4, Optimization Steps (k)=20, LLM Backbone=Avg2025.0993.493.3
TF-IDF + CARTTree Depth (d)=4, LLM Backbone=Avg2025.0992.892.4
ACTTree Depth (d)=4, Optimization Steps (k)=20, LLM Backbone=GPT-4.1 Nano2025.099291.5
ACTTree Depth (d)=4, Optimization Steps (k)=20, LLM Backbone=Qwen3 4b2025.099292.7
DSPyEvaluation Protocol=BFSR, 8 demos, LLM Backbone=GPT-4.1 Nano2025.0991.691.6
TEXTGRADLLM Backbone=Gemma3 4b2025.0991.691.3
TEXTGRADLLM Backbone=Avg2025.0991.491.7
ACTTree Depth (d)=4, Optimization Steps (k)=20, LLM Backbone=Gemma3 4b2025.0990.890
DSPyEvaluation Protocol=BFSR, 8 demos, LLM Backbone=Avg2025.0990.489.9
TEXTGRADLLM Backbone=GPT-4.1 Nano2025.0990.490.3
DSPyEvaluation Protocol=BFSR, 8 demos, LLM Backbone=Gemma3 4b2025.0989.689.3
TEXTGRADLLM Backbone=Qwen3 4b2025.0988.890
ACTTree Depth (d)=3, Optimization Steps (k)=10, LLM Backbone=Avg2025.0986.486.5
CoTEvaluation Protocol=0-shot, LLM Backbone=GPT-4.1 Nano2025.0985.583.5
DSPyEvaluation Protocol=BFSR, 8 demos, LLM Backbone=Qwen3 4b2025.0985.583.9
ACTTree Depth (d)=3, Optimization Steps (k)=10, LLM Backbone=GPT-4.1 Nano2025.0985.585.1
CoTEvaluation Protocol=0-shot, LLM Backbone=Avg2025.0984.584.2
Con. tag.LLM Backbone=Gemma3 4b2025.0982.782.2
Con. tag.LLM Backbone=GPT-4.1 Mini2025.0982.779.8
ACTTree Depth (d)=3, Optimization Steps (k)=10, LLM Backbone=Qwen3 4b2025.0982.782.6
ACTTree Depth (d)=3, Optimization Steps (k)=10, LLM Backbone=Gemma3 4b2025.0982.382.7
CoTEvaluation Protocol=0-shot, LLM Backbone=Qwen3 4b2025.0981.177.7
CoTEvaluation Protocol=0-shot, LLM Backbone=Gemma3 4b2025.0977.982.2
Con. tag.LLM Backbone=Qwen3 4b2025.0976.370.1
Con. tag.LLM Backbone=Avg2025.0975.166
Con. tag.LLM Backbone=GPT-4.1 Nano2025.0958.631.8