List your company
ResearchFormerly WizWand

Conversational response generation on FaithDial

19.2F1 ScoreChatR1-3b
9.21611.80814.416.992Oct 15, 2025
Updated 6mo ago

Evaluation results

MethodLinks
ChatR1-3bRAG=RAG, LLM=Qwen-3b, Training Strategy=Supervised Fine-tuning or RL Training2025.1019.284
SFTRAG=No, LLM=Qwen-3b, Training Strategy=Supervised Fine-tuning or RL Training2025.1018.683.8
ChatR1-7bRAG=RAG, LLM=Qwen-7b, Training Strategy=Supervised Fine-tuning or RL Training2025.1018.184.8
ChatR1 (w/o Rint.)RAG=RAG, LLM=Qwen-3b, Training Strategy=Supervised Fine-tuning or RL Training2025.1015.581.3
CoT R1RAG=No, LLM=Qwen-3b, Training Strategy=Supervised Fine-tuning or RL Training2025.1014.580.9
QR Search R1RAG=RAG, LLM=Qwen-3b, Training Strategy=Supervised Fine-tuning or RL Training2025.1014.482.2
ChatGPT (DI)RAG=No, LLM=GPT-3.5, Training Strategy=Zero-Shot2025.1012.980.9
Qwen-Instr. (RAG)RAG=RAG, LLM=Qwen-3b, Training Strategy=Zero-Shot2025.1012.373.8
UniConvRAG=RAG, LLM=Mistral 7b, Training Strategy=Supervised Fine-tuning or RL Training2025.1011.6—
Qwen-Instr. (DI)RAG=No, LLM=Qwen-3b, Training Strategy=Zero-Shot2025.1010.964.8
Qwen-Instr. (CoT)RAG=No, LLM=Qwen-3b, Training Strategy=Zero-Shot2025.109.664.8
IRCoTRAG=RAG, LLM=Qwen-3b, Training Strategy=Zero-Shot2025.109.664.7