Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

R-Debater: Retrieval-Augmented Debate Generation through Argumentative Memory

Maoyuan Li, Zhongsheng Wang, Haoyuan Li, Jiamou Liu

TL;DR

R-Debater introduces a retrieval-augmented, memory-based framework for multi-turn debate generation that integrates an argumentative memory knowledge base with role-based planning. It comprises three pipelines—Database Construction, Debate Data Retrieval, and Debate Generation—that jointly reconstruct argumentative memory to produce stance-consistent utterances. Empirical evaluation on the ORCHID dataset shows that R-Debater outperforms strong LLM and RAG baselines in factual grounding, logical coherence, and persuasive quality, with expert alignment confirming interpretability. The work advances grounded, transparent, and scalable debate generation with potential applications in automatic argumentation systems and interactive AI assistants.

Abstract

We present R-Debater, an agentic framework for generating multi-turn debates built on argumentative memory. Grounded in rhetoric and memory studies, the system views debate as a process of recalling and adapting prior arguments to maintain stance consistency, respond to opponents, and support claims with evidence. Specifically, R-Debater integrates a debate knowledge base for retrieving case-like evidence and prior debate moves with a role-based agent that composes coherent utterances across turns. We evaluate on standardized ORCHID debates, constructing a 1,000-item retrieval corpus and a held-out set of 32 debates across seven domains. Two tasks are evaluated: next-utterance generation, assessed by InspireScore (subjective, logical, and factual), and adversarial multi-turn simulations, judged by Debatrix (argument, source, language, and overall). Compared with strong LLM baselines, R-Debater achieves higher single-turn and multi-turn scores. Human evaluation with 20 experienced debaters further confirms its consistency and evidence use, showing that combining retrieval grounding with structured planning yields more faithful, stance-aligned, and coherent debates across turns.

R-Debater: Retrieval-Augmented Debate Generation through Argumentative Memory

TL;DR

R-Debater introduces a retrieval-augmented, memory-based framework for multi-turn debate generation that integrates an argumentative memory knowledge base with role-based planning. It comprises three pipelines—Database Construction, Debate Data Retrieval, and Debate Generation—that jointly reconstruct argumentative memory to produce stance-consistent utterances. Empirical evaluation on the ORCHID dataset shows that R-Debater outperforms strong LLM and RAG baselines in factual grounding, logical coherence, and persuasive quality, with expert alignment confirming interpretability. The work advances grounded, transparent, and scalable debate generation with potential applications in automatic argumentation systems and interactive AI assistants.

Abstract

We present R-Debater, an agentic framework for generating multi-turn debates built on argumentative memory. Grounded in rhetoric and memory studies, the system views debate as a process of recalling and adapting prior arguments to maintain stance consistency, respond to opponents, and support claims with evidence. Specifically, R-Debater integrates a debate knowledge base for retrieving case-like evidence and prior debate moves with a role-based agent that composes coherent utterances across turns. We evaluate on standardized ORCHID debates, constructing a 1,000-item retrieval corpus and a held-out set of 32 debates across seven domains. Two tasks are evaluated: next-utterance generation, assessed by InspireScore (subjective, logical, and factual), and adversarial multi-turn simulations, judged by Debatrix (argument, source, language, and overall). Compared with strong LLM baselines, R-Debater achieves higher single-turn and multi-turn scores. Human evaluation with 20 experienced debaters further confirms its consistency and evidence use, showing that combining retrieval grounding with structured planning yields more faithful, stance-aligned, and coherent debates across turns.
Paper Structure (25 sections, 8 equations, 3 figures, 6 tables)

This paper contains 25 sections, 8 equations, 3 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Computational Argumentation
  4. Retrieval-Augmented and Agentic AI
  5. Task Formulation
  6. Methodology
  7. Database Construction
  8. Debate Data Retrieval
  9. Debate Generation
  10. Experiments
  11. Experimental Setup
  12. Results and Analysis
  13. Single-turn Debate Evaluation
  14. Multi-turn Competitive Debate Evaluation
  15. Expert Alignment Evaluation.
  16. ...and 10 more sections

Figures (3)

  • Figure 1: Humans debate by integrating external knowledge and debate strategies, while LLMs often fail to connect retrieval with argumentation. Our task aims to bridge this gap by enhancing the ability of LLMs to utilize materials grounded in retrieved debate history and strategies more effectively.
  • Figure 2: R-Debater consists of three modules: (1) Database Construction, (2) Debate Data Retrieval, and (3) Debate Generation. Together, they enable the dynamic reconstruction of argumentative memory for generating next-turn debates.
  • Figure 3: Human evaluation results on the six InspireScore dimensions. Each spoke shows the macro-averaged Likert rating (1-10) for a system; shaded bands indicate 95% bootstrap CIs.