List your company
ResearchFormerly WizWand

Interactive Decision Making on ALFWorld Seen

97.86Success RateSTEP-HRL
-0.9424.7150.3676.01Aug 5, 2025Sep 29, 2025Nov 24, 2025Jan 19, 2026Mar 15, 2026May 10, 2026Jul 5, 2026
Updated 3mo ago

Evaluation results

MethodLinks
STEP-HRLBackbone=Gemma-7B, Evaluation Protocol=Fine-tuning (LoRA)2026.0497.86
STEP-HRLBackbone=Llama-3-8B, Evaluation Protocol=Fine-tuning (LoRA)2026.0497.14
STEP-HRLBackbone=Mistral-7B, Evaluation Protocol=Fine-tuning (LoRA)2026.0496.43
SFT+HSLBackbone=Llama3.2-1b2026.0793.57
DPO+HSLBackbone=Llama3.2-1b2026.0792.86
BAGELBackbone=Llama3.2-1b2026.0788.09
BPOApproach=System-2, Evaluation Protocol=Fine-Tuned, Base Model=Llama-3.1-8B-Instruct2025.0887.9
DPOBackbone=Llama3.2-1b2026.0784.52
MPOApproach=System-2, Evaluation Protocol=Fine-Tuned, Base Model=Llama-3.1-8B-Instruct2025.0882.9
BEHAVIORCLONEBackbone=Llama3.2-1b2026.0780.72
SFTApproach=System-2, Evaluation Protocol=Fine-Tuned, Base Model=Llama-3.1-8B-Instruct2025.0880
SFTBackbone=Llama3.2-1b2026.0779.52
ETOApproach=System-2, Evaluation Protocol=Fine-Tuned, Base Model=Llama-3.1-8B-Instruct2025.0878.6
WKMBackbone=Mistral-7B, Evaluation Protocol=Fine-tuning (LoRA)2026.0473.57
GLIDERBackbone=Gemma-7B, Evaluation Protocol=Fine-tuning (LoRA)2026.0472.12
Qwen-2.5-7B-InstructApproach=System-1, Evaluation Protocol=Zero-shot, Model Category=Base Models2025.0872.1
GLIDERBackbone=Llama-3-8B, Evaluation Protocol=Fine-tuning (LoRA)2026.0471.56
WKMBackbone=Gemma-7B, Evaluation Protocol=Fine-tuning (LoRA)2026.0470.71
GLIDERBackbone=Mistral-7B, Evaluation Protocol=Fine-tuning (LoRA)2026.0470.02
SELFIMITBackbone=Llama3.2-1b2026.0769.05
WKMBackbone=Llama-3-8B, Evaluation Protocol=Fine-tuning (LoRA)2026.0468.57
REACTBackbone=GPT-4, Evaluation Protocol=Prompt-based2026.0467.32
ETOBackbone=Mistral-7B, Evaluation Protocol=Fine-tuning (LoRA)2026.0466.84
ETOBackbone=Gemma-7B, Evaluation Protocol=Fine-tuning (LoRA)2026.0466.43
ETOBackbone=Llama-3-8B, Evaluation Protocol=Fine-tuning (LoRA)2026.0464.29
Deepseek-R1Approach=System-2, Evaluation Protocol=Zero-shot, Model Category=Reasoning LLMs2025.0861.4
o3-miniApproach=System-2, Evaluation Protocol=Zero-shot, Model Category=Reasoning LLMs2025.0858.9
ADAMEMMemory Policy Setup=On-Policy, Backbone=Qwen3-4B-Instruct, Long Term=Traj., Short Term=Strat.2026.0654
SynapseMemory Policy Setup=On-Policy, Backbone=Qwen3-4B-Instruct, Long Term=Traj., Short Term=Traj.2026.0652.1
Qwen-3-8B-ThinkingApproach=System-2, Evaluation Protocol=Zero-shot, Model Category=Reasoning LLMs2025.0849.3
ReasoningBankMemory Policy Setup=On-Policy, Backbone=Qwen3-4B-Instruct, Long Term=Strat., Short Term=Strat.2026.0649.3
ADAMEMMemory Policy Setup=Off-Policy, Backbone=Gemma-3-27b-it, Long Term=Traj., Short Term=Strat.2026.0647.4
No MemoryMemory Policy Setup=On-Policy, Backbone=Qwen3-4B-Instruct, Long Term=-, Short Term=-2026.0645.2
ReasoningBankMemory Policy Setup=Off-Policy, Backbone=Gemma-3-27b-it, Long Term=Strat., Short Term=Strat.2026.0645.2
No MemoryMemory Policy Setup=Off-Policy, Backbone=Gemma-3-27b-it, Long Term=-, Short Term=-2026.0636.7
SynapseMemory Policy Setup=Off-Policy, Backbone=Gemma-3-27b-it, Long Term=Traj., Short Term=Traj.2026.0635
Llama-3.1-8B-InstructApproach=System-1, Evaluation Protocol=Zero-shot, Model Category=Base Models2025.0832.9
SwitchSageBackbone=Mistral-7B, Evaluation Protocol=Prompt-based2026.0430.29
SwitchSageBackbone=Llama-3-8B, Evaluation Protocol=Prompt-based2026.0420.39
REACTBackbone=GPT-3.5-Turbo, Evaluation Protocol=Prompt-based2026.0415.41
ReflexionBackbone=Mistral-7B, Evaluation Protocol=Prompt-based2026.0411.56
SwitchSageBackbone=Gemma-7B, Evaluation Protocol=Prompt-based2026.048.23
ReActBackbone=Mistral-7B, Evaluation Protocol=Prompt-based2026.047.86
ReflexionBackbone=Gemma-7B, Evaluation Protocol=Prompt-based2026.047.14
ReActBackbone=Gemma-7B, Evaluation Protocol=Prompt-based2026.046.43
ReflexionBackbone=Llama-3-8B, Evaluation Protocol=Prompt-based2026.044.29
ReActBackbone=Llama-3-8B, Evaluation Protocol=Prompt-based2026.042.86