RAM3C: Enhancing Goal-oriented Open-ended Dialogue-based Educational System by Retrieval-augmented Multi-agent Collaboration

Anonymous

Abstract: This study presents the Retrieval-Augmented Multi-role Multi-agent Multi-round Collaboration (RAM3C) system, designed to improve the overall effectiveness of open-ended dialogue-based educational systems. Focusing on aspects of Humanlikeness, Individualization, Teaching expertise and Safety (HITS), RAM3C utilizes a dynamic framework that incorporates multi-agent, multi-round collaboration with multiple roles to harness collective expertise. RAM3C equips agents with tailored, multi-source knowledge bases and implements a history-sampling weighted retrieval-fusion approach to generate diverse, accurate, and safe educational dialogues. Our evaluation on a scenario of "Literature Discussion Class" by human volunteers and a decentralized, LLM-emulated expert group, confirms RAM3C's capability to deliver high-quality educational experiences, underscoring its substantial potential to elevate educational quality.

Paper Type: short

Research Area: Dialogue and Interactive Systems

Contribution Types: NLP engineering experiment, Publicly available software and/or pre-trained models, Data resources

Languages Studied: Chinese, English

Preprint Status: We plan to release a non-anonymous preprint in the next two months (i.e., during the reviewing process).

A1: yes

A1 Elaboration For Yes Or No: 6. Limitations

A2: yes

A2 Elaboration For Yes Or No: 5. conclusion

A3: yes

B: no

C: yes

C1: no

C1 Elaboration For Yes Or No: I mainly use the OpenAI API to do my experiments. It doesn't take too much resources.

C2: n/a

C3: yes

C3 Elaboration For Yes Or No: Table 1

C4: yes

C4 Elaboration For Yes Or No: I use LangChain and Chromadb.

D: no

D1: n/a

E: no