List your company
ResearchFormerly WizWand

Question Answering on DROP (F1 score)

87.5F1 ScoreOneFlow
6.58827.59448.669.606Jun 18, 2025Aug 12, 2025Oct 7, 2025Dec 2, 2025Jan 27, 2026Mar 24, 2026May 19, 2026
Updated 5mo ago

Evaluation results

MethodLinks
OneFlowModel=Claude 3.5 Haiku2026.0187.5
AFlowModel=Claude 3.5 Haiku2026.0186.8
AFlowMethod category=Automatically designed multi-agent frameworks, Executor LLM=GPT-4o-mini, Execution protocol=multi-agent2026.0183.1
ReConcileBackbone=GPT-4o-mini2026.0582.1
OneFlowMethod category=Single-LLM implementation, Executor LLM=GPT-4o-mini, Execution protocol=single-agent execution2026.0181.7
LLM-DebateBackbone=GPT-4o-mini2026.0581.4
OneFlowMethod category=Automatically designed multi-agent frameworks, Executor LLM=GPT-4o-mini, Execution protocol=multi-agent2026.0181.1
AFlowMethod category=Single-LLM implementation, Executor LLM=GPT-4o-mini, Execution protocol=single-agent execution2026.0181.1
AFlow*Backbone=Mixed2026.0580.6
CoT SCMethod category=Manual baselines, Executor LLM=GPT-4o-mini, Execution protocol=standard, shots=5-shot2026.0179.4
CoTMethod category=Manual baselines, Executor LLM=GPT-4o-mini, Execution protocol=standard2026.0178.9
MultiPersonaMethod category=Manual baselines, Executor LLM=GPT-4o-mini, Execution protocol=standard2026.0178.9
CoT SCBackbone=GPT-4o-mini, Prompting Setting=5-shot2026.0578.8
CoTBackbone=GPT-4o-mini2026.0578.5
MedPromptBackbone=GPT-4o-mini2026.0578
AGENTCO-OPBackbone=GPT-4o-mini2026.0577.2
ADASBackbone=GPT-4o-mini2026.0576.6
MultiPersonaBackbone=GPT-4o-mini2026.0574.4
Self RefineBackbone=GPT-4o-mini2026.0570.2
AFlowBackbone=GPT-4o-mini2026.0568.9
IOBackbone=GPT-4o-mini2026.0568.3
IOMethod category=Manual baselines, Executor LLM=GPT-4o-mini, Execution protocol=standard2026.0166.2
IOModel=Claude 3.5 Haiku2026.0164
MeZOModel Scale=OPT-6.7B, Privacy Level=non-private, Few-shot shots (k)=10002025.0628.8
DPZeroModel Scale=OPT-6.7B, Privacy Budget (epsilon)=6, Few-shot shots (k)=10002025.0628.4
DP-GRAPEModel Scale=OPT-6.7B, Privacy Budget (epsilon)=6, Few-shot shots (k)=10002025.0628.2
DP-GRAPEModel Scale=OPT-6.7B, Privacy Budget (epsilon)=2, Few-shot shots (k)=10002025.0627.8
DPZeroModel Scale=OPT-6.7B, Privacy Budget (epsilon)=2, Few-shot shots (k)=10002025.0627.6
DP-AdamModel Scale=OPT-2.7B, Privacy Budget (epsilon)=6, Few-shot shots (k)=10002025.0626.3
DP-GRAPEModel Scale=OPT-1.3B, Privacy Budget (epsilon)=6, Few-shot shots (k)=10002025.0626.1
DP-AdamModel Scale=OPT-1.3B, Privacy Budget (epsilon)=6, Few-shot shots (k)=10002025.0625.9
MeZOModel Scale=OPT-2.7B, Privacy Level=non-private, Few-shot shots (k)=10002025.0625.5
DP-GRAPEModel Scale=OPT-2.7B, Privacy Budget (epsilon)=6, Few-shot shots (k)=10002025.0625.5
DP-AdamModel Scale=OPT-1.3B, Privacy Budget (epsilon)=2, Few-shot shots (k)=10002025.0625.2
DP-GRAPEModel Scale=OPT-1.3B, Privacy Budget (epsilon)=2, Few-shot shots (k)=10002025.0625
DP-AdamModel Scale=OPT-2.7B, Privacy Budget (epsilon)=2, Few-shot shots (k)=10002025.0624.9
DPZeroModel Scale=OPT-1.3B, Privacy Budget (epsilon)=6, Few-shot shots (k)=10002025.0624.7
DPZeroModel Scale=OPT-2.7B, Privacy Budget (epsilon)=6, Few-shot shots (k)=10002025.0624.6
MeZOModel Scale=OPT-1.3B, Privacy Level=non-private, Few-shot shots (k)=10002025.0624.4
DP-GRAPEModel Scale=OPT-2.7B, Privacy Budget (epsilon)=2, Few-shot shots (k)=10002025.0624.1
DPZeroModel Scale=OPT-1.3B, Privacy Budget (epsilon)=2, Few-shot shots (k)=10002025.0623.9
DPZeroModel Scale=OPT-2.7B, Privacy Budget (epsilon)=2, Few-shot shots (k)=10002025.0623.1
Zero-ShotModel Scale=OPT-6.7B, Evaluation Protocol=zero-shot2025.0617.8
Zero-ShotModel Scale=OPT-1.3B, Evaluation Protocol=zero-shot2025.0611.1
Zero-ShotModel Scale=OPT-2.7B, Evaluation Protocol=zero-shot2025.069.7