List your company

Multi-hop Question Answering on MuSiQue (EM, F1, ∆ EM)

30EM3-Agent Baseline
11.2816.142125.86Jun 23, 2026

Evaluation results

MethodLinks
3-Agent BaselineBackbone=Qwen2.5-7B-Instruct, Number of queries=100, Context=20 mixed supporting/distractor paragraphs per query2026.063039.616
SDA (rank)Backbone=Qwen2.5-7B-Instruct, Number of queries=100, Context=20 mixed supporting/distractor paragraphs per query2026.063038.716
PDEBackbone=Qwen2.5-7B-Instruct, Number of queries=100, Context=20 mixed supporting/distractor paragraphs per query2026.062434.410
CoT De-PolarizationBackbone=Qwen2.5-7B-Instruct, Number of queries=100, Context=20 mixed supporting/distractor paragraphs per query2026.062130.97
NF (No-Filter)Backbone=Qwen2.5-7B-Instruct, Number of queries=100, Context=20 mixed supporting/distractor paragraphs per query2026.061421.3—
PDE-RandomBackbone=Qwen2.5-7B-Instruct, Number of queries=100, Context=20 mixed supporting/distractor paragraphs per query2026.061219.62