List your company

Multi-Hop Question Answering on HotpotQA (Accuracy)

44.2AccuracySkillOrchestra+
12.79220.94629.137.254Feb 23, 2026Feb 25, 2026Feb 27, 2026Mar 1, 2026Mar 3, 2026Mar 5, 2026Mar 7, 2026

Evaluation results

MethodLinks
SkillOrchestra+Routing Strategy=Skill-based Routing, Orchestrator Configuration=Best/Multiple2026.0244.2
HCAPO (Ours)Type=RL Training, Base Model=Qwen2.5-7B-Instruct2026.0342.1
GiGPOType=RL Training, Base Model=Qwen2.5-7B-Instruct2026.0341.6
SkillOrchestraRouting Strategy=Skill-based Routing, Orchestrator Configuration=Single2026.0239
HCAPO (Ours)Type=RL Training, Base Model=Qwen2.5-3B-Instruct2026.0338.6
StepSearchType=RL Training, Base Model=Qwen2.5-7B-Instruct2026.0338.6
Search-R1Type=RL Training, Base Model=Qwen2.5-7B-Instruct2026.0337
GiGPOType=RL Training, Base Model=Qwen2.5-3B-Instruct2026.0336.9
Router-R1Routing Strategy=RL-based Routing2026.0235.2
ZeroSearchType=RL Training, Base Model=Qwen2.5-7B-Instruct2026.0334.6
StepSearchType=RL Training, Base Model=Qwen2.5-3B-Instruct2026.0334.5
Search-R1Type=RL Training, Base Model=Qwen2.5-3B-Instruct2026.0332.4
Largest LLMRouting Strategy=Heuristic & Discriminative Routing2026.0227.8
ZeroSearchType=RL Training, Base Model=Qwen2.5-3B-Instruct2026.0327.4
Prompt LLMRouting Strategy=Heuristic & Discriminative Routing2026.0226.8
RouterDCRouting Strategy=Heuristic & Discriminative Routing2026.0224.4
R1-InstructType=RL Training, Base Model=Qwen2.5-3B-Instruct2026.0323.7
Search-R1Routing Strategy=No Routing2026.0223.6
GraphRouterRouting Strategy=Heuristic & Discriminative Routing2026.0223.4
FrugalGPTRouting Strategy=Heuristic & Discriminative Routing2026.0223.4
KNN RouterRouting Strategy=Heuristic & Discriminative Routing2026.0222.4
RAGRouting Strategy=No Routing2026.0221.6
BERT RouterRouting Strategy=Heuristic & Discriminative Routing2026.0221.6
R1-InstructType=RL Training, Base Model=Qwen2.5-7B-Instruct2026.0320.8
Prompt LLM+Routing Strategy=Heuristic & Discriminative Routing, Turn Protocol=multi turn2026.0220.6
SFTRouting Strategy=No Routing2026.0219.8
MLP RouterRouting Strategy=Heuristic & Discriminative Routing2026.0219.8
CoTRouting Strategy=No Routing2026.0216.8
KNN Router+Routing Strategy=Heuristic & Discriminative Routing, Turn Protocol=multi turn2026.0215.4
VanillaRouting Strategy=No Routing2026.0214