List your company

Multi-hop Question Answering on HotpotQA (SR)

52Success Rate (SR)Gemini-3-Flash
22.8830.443845.56May 13, 2026

Evaluation results

MethodLinks
Gemini-3-FlashProtocol=Prompting2026.0552
Gemini-2.5-flashProtocol=Prompting2026.0547
MATPOBackbone=Qwen3-4B2026.0546
Critique-GRPOBackbone=Qwen3-8B2026.0546
MATPOBackbone=Qwen3-8B2026.0545
ScalingInterBackbone=Qwen3-4B2026.0544
ICRLBackbone=Qwen3-4B2026.0544
GSPOBackbone=Qwen3-8B2026.0544
ICRLBackbone=Qwen3-8B2026.0544
Critique-GRPOBackbone=Qwen3-4B2026.0543.5
GSPOBackbone=Qwen3-4B2026.0542
ScalingInterBackbone=Qwen3-8B2026.0541.5
Qwen3-30B-A3BProtocol=Prompting2026.0540
GRPOBackbone=Qwen3-4B2026.0538
GRPOBackbone=Qwen3-8B2026.0538
Qwen3-8BProtocol=Prompting2026.0532
Qwen3-4BProtocol=Prompting2026.0524