List your company

Multi-hop Question Answering on HotpotQA (F1, Tokens)

69.9F1 ScorePACT
44.83651.34357.8564.357Jun 3, 2026

Evaluation results

MethodLinks
PACTBackbone=Qwen3-8B2026.0669.96,704
Multi-Agent DebateBackbone=Qwen3-8B2026.0669.613,520
CoABackbone=Qwen3-8B2026.0668.67,281
PACTBackbone=Qwen3-32B2026.0661.54,821
Multi-Agent DebateBackbone=Qwen3-32B2026.0660.312,258
TextMASBackbone=Qwen3-8B2026.0659.59,384
PACTBackbone=Qwen3-14B2026.0656.85,996
CoABackbone=Qwen3-14B2026.0654.86,782
CoABackbone=Qwen3-32B2026.0654.25,871
TextMASBackbone=Qwen3-32B2026.0650.18,857
Multi-Agent DebateBackbone=Qwen3-14B2026.0648.412,133
TextMASBackbone=Qwen3-14B2026.0645.88,652