Question Answering on HotpotQA (Accuracy and Efficiency)
84.6Accuracyexpert_synthesis_prompting
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| expert_synthesis_promptingBackbone=Gemma-3 4B2026.02 | 84.6 | 6.98 | 0.121 | |
| multi_hop_compact_promptingBackbone=Gemma-3 4B2026.02 | 84.1 | 8.37 | 0.1 | |
| multi_hop_decomposition_promptingBackbone=Gemma-3 4B2026.02 | 83.2 | 8.01 | 0.104 | |
| optimized_multihop_3b_promptingBackbone=Gemma-3 4B2026.02 | 83.1 | — | — | |
| optimized_multi_hop_promptingBackbone=Gemma-3 4B2026.02 | 80.9 | — | — | |
| Standard_context_aware (baseline)Backbone=Gemma-3 4B2026.02 | 78.5 | 0.709 | 1.107 | |
| slm_hotpot_smecBackbone=Gemma-3 4B2026.02 | — | 7.08 | 0.11 |