List your company
ResearchFormerly WizWand

Multi-hop Question Answering on HotpotQA (Avg@8 accuracy)

92.52Avg@8 AccuracyST-EVO
31.201647.120863.0478.9592Dec 31, 2025Jan 7, 2026Jan 15, 2026Jan 23, 2026Jan 31, 2026Feb 8, 2026Feb 16, 2026
Updated 8mo ago

Evaluation results

MethodLinks
ST-EVOBackbone=gpt-oss-120b2026.0292.52
AFlowBackbone=gpt-oss-120b2026.0290.04
STEERBackbone=gpt-oss-120b2026.0289.35
SpecReasonBackbone=gpt-oss-120b2026.0288.29
G-DesignerBackbone=gpt-oss-120b2026.0287.66
SafeSieveBackbone=gpt-oss-120b2026.0287.22
ARG-DesignerBackbone=gpt-oss-120b2026.0286.45
GPTSwarmBackbone=gpt-oss-120b2026.0285.4
LangGraphBackbone=gpt-oss-120b2026.0284.92
LLM-DebateBackbone=gpt-oss-120b2026.0284.55
Random GraphBackbone=gpt-oss-120b2026.0284.28
StarBackbone=gpt-oss-120b2026.0283.52
Complete GraphBackbone=gpt-oss-120b2026.0283.35
AutoGenBackbone=gpt-oss-120b2026.0283.25
VanillaBackbone=gpt-oss-120b2026.0282.35
SCBackbone=gpt-oss-120b2026.0282.24
TreeBackbone=gpt-oss-120b2026.0281.5
CoTBackbone=gpt-oss-120b2026.0281.04
ChainBackbone=gpt-oss-120b2026.0280.77
ADOPT-JointPrompt Optimization Strategy=ADOPT-Joint2025.1268
ADOPT-InstructPrompt Optimization Strategy=ADOPT-Instruct2025.1267
GEPAPrompt Optimization Strategy=GEPA2025.1263
MIPROPrompt Optimization Strategy=MIPRO2025.1262
COTPrompt Optimization Strategy=Chain-of-Thought2025.1258
No-COTPrompt Optimization Strategy=No-COT2025.1252
MAS-OrchestraOrchestration Type=Training-time Orchestration, Orchestrator LLM=Qwen-7b, Sub-agent Backbone=GPT-120b (low)2026.0149
SearchAgentOrchestration Type=Standalone Agent, Sub-agent Backbone=GPT-120b (low)2026.0146.44
ToolOrchestraOrchestration Type=Public Training-time Orchestration, Sub-agent Backbone=GPT-120b (low)2026.0137.44
DebateAgentOrchestration Type=Standalone Agent, Sub-agent Backbone=GPT-120b (low)2026.0136.88
ReflexionAgentOrchestration Type=Standalone Agent, Sub-agent Backbone=GPT-120b (low)2026.0136.63
SCAgentOrchestration Type=Standalone Agent, Sub-agent Backbone=GPT-120b (low)2026.0135.5
CoTAgentOrchestration Type=Standalone Agent, Sub-agent Backbone=GPT-120b (low)2026.0133.56