List your company

Multi-Hop Question Answering on HotpotQA (EM, FBERT)

30.81Exact Match (EM)CoT
8.564414.339720.11525.8903Apr 20, 2026

Evaluation results

MethodLinks
CoTSLM=Qwen3-235B2026.0430.8169.26
Dynamic Collaboration FrameworkSLM=Qwen3-8B2026.0429.6669.21
Dynamic Collaboration FrameworkSLM=Llama-3.1-8B2026.0425.5467.06
Dynamic Collaboration FrameworkSLM=Qwen3-4B2026.0425.5167.3
Direct InferSLM=Qwen3-235B2026.042565.35
Dynamic Collaboration FrameworkSLM=Llama-3.2-3B2026.0422.7764.96
CoTSLM=Qwen3-8B2026.0415.1559.59
CoTSLM=Llama-3.1-8B2026.0414.8255.31
CoTSLM=Qwen3-4B2026.0410.9857.41
CoTSLM=Llama-3.2-3B2026.049.4251.36