List your company

Question Answering on HotpotQA (Accuracy and Efficiency)

84.6Accuracyexpert_synthesis_prompting
78.25679.90381.5583.197Feb 14, 2026

Evaluation results

MethodLinks
expert_synthesis_promptingBackbone=Gemma-3 4B2026.0284.66.980.121
multi_hop_compact_promptingBackbone=Gemma-3 4B2026.0284.18.370.1
multi_hop_decomposition_promptingBackbone=Gemma-3 4B2026.0283.28.010.104
optimized_multihop_3b_promptingBackbone=Gemma-3 4B2026.0283.1——
optimized_multi_hop_promptingBackbone=Gemma-3 4B2026.0280.9——
Standard_context_aware (baseline)Backbone=Gemma-3 4B2026.0278.50.7091.107
slm_hotpot_smecBackbone=Gemma-3 4B2026.02—7.080.11