List your company
ResearchFormerly WizWand

Web navigation on WebShop

90.5Average ScoreSAPO
1.47624.58847.770.812Nov 27, 2025Dec 29, 2025Jan 30, 2026Mar 3, 2026Apr 4, 2026May 6, 2026Jun 7, 2026
Updated 4mo ago

Evaluation results

MethodLinks
SAPOMethod Category=Memory-Augmented RL-based Methods2026.0690.5—78.1
SKILLRLMethod Category=Memory-Augmented RL-based Methods2026.0685.2—72.7
D2SkillMethod Category=Memory-Augmented RL-based Methods2026.0683.4—73.4
Skill0Method Category=Memory-Augmented RL-based Methods2026.0683.2—71.9
RLOO*Method Category=RL-based Methods2026.0680.3—65.7
GRPO*Method Category=RL-based Methods2026.0679.3—66.1
IPRRefinement Category=Process Refinement, Backbone=Llama-2-7B2025.1271.3——
Q-EvolveBackbone=Llama-2-7B-Chat2026.0670.5——
QLASSBackbone=Llama-2-7B-Chat2026.0670.3——
MACLARefinement Category=Process Refinement, Backbone=Llama-2-7B2025.1270.2——
DMPOBackbone=Llama-2-7B-Chat2026.0670.1——
Best-of-NBackbone=Llama-2-7B-Chat, N=62026.0667.9——
SimpleMem+GRPOMethod Category=Memory-Augmented RL-based Methods2026.0667.8—46.9
ETORefinement Category=Outcome Refinement, Backbone=Llama-2-7B2025.1267.4——
ETOBackbone=Llama-2-7B-Chat2026.0667.4——
Co-Evolving AgentsBackbone=Qwen3-4B-Instruct-25072025.1166.372.5—
RFT-PPORefinement Category=Outcome Refinement, Backbone=Llama-2-7B2025.1264.2——
ReflexionPrompting=ReAct2026.0664.2——
PPOBackbone=Llama-2-7B-Chat2026.0664.2——
Step-PPORefinement Category=Process Refinement, Backbone=Llama-2-7B2025.1264——
RFT-CRRefinement Category=Outcome Refinement, Backbone=Llama-2-7B2025.1263.6——
RFTBackbone=Llama-2-7B-Chat2026.0663.6——
GPT-4Refinement Category=Prompt-based, Backbone=GPT-42025.1263.2——
GPT-4Prompting=ReAct2026.0663.2——
SFTBackbone=Llama-2-7B-Chat2026.0663.1——
GPT-3.5-TurboRefinement Category=Prompt-based, Backbone=GPT-3.5-Turbo2025.1262.4——
GPT-3.5-TurboPrompting=ReAct2026.0662.4——
SFTRefinement Category=Outcome Refinement, Backbone=Llama-2-7B2025.1260.2——
ETOBackbone=Qwen3-4B-Instruct-25072025.1159.565.7—
Reflexion*Method Category=Prompt-based Agentic or Memory-based Methods2026.0658.1—28.8
Mem0+GRPOMethod Category=Memory-Augmented RL-based Methods2026.0658.1—37.5
ReAct*Method Category=Prompt-based Agentic or Memory-based Methods2026.0646.2—19.5
Gemini-2.5-ProMethod Category=Closed-source LLMs2026.0642.5—35.9
EvolveRMethod Category=Memory-Augmented RL-based Methods2026.0642.5—17.6
SFTBackbone=Qwen3-4B-Instruct-25072025.1140.363.9—
SimpleMemMethod Category=Prompt-based Agentic or Memory-based Methods2026.0633.2—8.59
GPT-4oMethod Category=Closed-source LLMs2026.0631.8—23.7
ExpeLMethod Category=Prompt-based Agentic or Memory-based Methods2026.0630.9—11.2
MemRLMethod Category=Memory-Augmented RL-based Methods2026.0629.5—9.2
Qwen2.5Method Category=Qwen2.5-7B-Instruct2026.0626.4—7.8
MemPMethod Category=Prompt-based Agentic or Memory-based Methods2026.0625.3—6.4
Mem0Method Category=Prompt-based Agentic or Memory-based Methods2026.0623.9—2
InternVL2.5-8BModel=InternVL2.5-8B, Strategy=Zero-Shot2026.0523.5—3.9
Qwen2.5-VL-7BModel=Qwen2.5-VL-7B, Strategy=Zero-Shot2026.0519.4—4.1
Llama-2-7BRefinement Category=Prompt-based, Backbone=Llama-2-7B2025.1217.9——
Base AgentBackbone=Llama-2-7B-Chat2026.0617.9——
InternVL2.5-8BModel=InternVL2.5-8B, Strategy=GPT Trajectory Imitation2026.0516.5—8.8
GPT-4oModel=GPT-4o, Strategy=Zero-Shot2026.0515.5—17.2
ViGoRLModel=ViGoRL, Strategy=Zero-Shot2026.0513.4—5.6
Qwen2.5-VL-7BModel=Qwen2.5-VL-7B, Strategy=OS-Genesis2026.0512.7—11.2
SCALE-20kModel=LLaVA-NeXT-8B, Strategy=SCALE-20k2026.0510.1—2.1
Qwen2.5-VL-7BModel=Qwen2.5-VL-7B, Strategy=GPT Trajectory Imitation2026.058.5—18.3
InternVL2.5-8BModel=InternVL2.5-8B, Strategy=OS-Genesis2026.056.6—11.6
SCALEModel=Qwen2.5-VL-7B, Strategy=SCALE(ours)2026.055.7—14.4
SCALEModel=InternVL2.5-8B, Strategy=SCALE(ours)2026.054.9—11
AugvisModel=Augvis, Strategy=Zero-Shot2026.05——4.3
Critique-GRPOBackbone=Qwen3-4B2026.05—87.172.5
Critique-GRPOBackbone=Qwen3-8B2026.05—86.873.5
Gemini-2.5-flashProtocol=Prompting2026.05—48.741.2
Gemini-3-FlashProtocol=Prompting2026.05—57.353
GRPOBackbone=Qwen3-4B2026.05—81.863.5
GRPOBackbone=Qwen3-8B2026.05—83.771.5
GSPOBackbone=Qwen3-4B2026.05—82.165
GSPOBackbone=Qwen3-8B2026.05—83.471
ICRLBackbone=Qwen3-4B2026.05—88.974.5
ICRLBackbone=Qwen3-8B2026.05—88.376
InternVL2.5-8BModel=InternVL2.5-8B, Strategy=Tree Search2026.05——3
LLaVA-NeXT-8BModel=LLaVA-NeXT-8B, Strategy=Zero-Shot2026.05——0
MATPOBackbone=Qwen3-4B2026.05—82.465
MATPOBackbone=Qwen3-8B2026.05—81.668
Qwen2.5-VL-7BModel=Qwen2.5-VL-7B, Strategy=Tree Search2026.05——4.1
Qwen3-30B-A3BProtocol=Prompting2026.05—25.63
Qwen3-4BProtocol=Prompting2026.05—9.61.5
Qwen3-8BProtocol=Prompting2026.05—10.32
ScalingInterBackbone=Qwen3-4B2026.05—84.766.5
ScalingInterBackbone=Qwen3-8B2026.05—86.273