List your company
ResearchFormerly WizWand

tracking shuffled objects seven objects on BBH (test)

92.8AccuracyTextGrad
30.71246.83162.9579.069Mar 3, 2025May 15, 2025Jul 28, 2025Oct 10, 2025Dec 23, 2025Mar 7, 2026May 20, 2026
Updated 5mo ago

Evaluation results

MethodLinks
TextGradTest Engine=Phi-3.5-Mini-Instruct2026.0592.8
TextRegTest Engine=Phi-3.5-Mini-Instruct2026.0592.2
CoTTest Engine=Phi-3.5-Mini-Instruct2026.0589.1
EvoPrompt(GA)-OPTS(TS)Task-solving LLM=GPT-4o mini, Number of trials=32025.0388.33
REVOLVETest Engine=Phi-3.5-Mini-Instruct2026.0587.7
EvoPrompt(DE)-OPTS(TS)Task-solving LLM=GPT-4o mini2025.0381.83
EvoPrompt(DE)Task-solving LLM=GPT-4o mini2025.0380.67
TextRegTest Engine=Llama-3.1-8B-Instruct2026.0576.6
EvoPrompt(GA)Task-solving LLM=GPT-4o mini, Number of trials=32025.0375.67
TextGradTest Engine=Llama-3.1-8B-Instruct2026.0566.7
REVOLVETest Engine=Llama-3.1-8B-Instruct2026.0565.4
CoTTest Engine=Llama-3.1-8B-Instruct2026.0548.5
TextRegTest Engine=Llama-3-8B-Instruct2026.0548.3
REVOLVETest Engine=Llama-3-8B-Instruct2026.0541.8
REVOLVETest Engine=Qwen2-7B-Instruct2026.0540.2
TextRegTest Engine=Qwen2-7B-Instruct2026.0539.1
TextGradTest Engine=Llama-3-8B-Instruct2026.0538
CoTTest Engine=Llama-3-8B-Instruct2026.0536.5
CoTTest Engine=Qwen2-7B-Instruct2026.0533.9
TextGradTest Engine=Qwen2-7B-Instruct2026.0533.1