Research map / Large language models / Agents, code and math
Truth-Tracking Evaluation in Opinion-Based Argumentation
Juliete Rossie, Jérôme Delobelle, Sébastien Konieczny, Srdjan Vesic
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| switch from preferred to complete semantics (ablation) | COS AB and COS AR · VAST, reliability 100% | accuracy (AM): around 100% -> approximately 70% | Hurt |
| switch from preferred to complete semantics (ablation) | CSS U,Sum and CSS U,leximin · VAST, high reliability | accuracy (AM): stabilizing at 98 % rather than 100% | No clear effect |
| combine argumentation framework with votes instead of majority rule alone | CSS U,Sum · VAST, preferred, reliability 60% | accuracy (AM): 25% -> exceeds 95% | Improved |
| increase number of votes from 5 to 20 (ablation) | COS AR and CSS U,Sum · VAST, preferred, reliability 60% | accuracy (AM): 50% -> nearly 90% | Improved |
| increase number of votes from 5 to 100 (ablation) | Majority rule · VAST, preferred, reliability 60% | accuracy (AM): 0% -> approximately 70% | Improved |
| increase number of votes (ablation) | COS AB and CSS U,leximin · VAST, preferred, reliability 60% | accuracy (AM) | No clear effect |
| increase number of extensions and arguments (ablation) | CSS U,Sum and COS AR · VAST, preferred, reliability 60% | accuracy (AM): maximum decrease of 10 points | No clear effect |
| increase number of extensions and arguments (ablation) | other COS (egalitarian, AB, labelling) · VAST, preferred, reliability 60% | accuracy (AM): average decline of 40 points | Hurt |
| use leximin instead of sum aggregation at low reliability (ablation) | CSS with utility score · VAST, preferred, reliability at most 50% | accuracy (AM) | Improved |
| use utilitarian instead of egalitarian OWA aggregation (ablation) | COS AB (approval ballots) · VAST, preferred, reliability at least 50% | accuracy (AM) | Improved |
| evaluate with similarity metric SM instead of AM (ablation) | egalitarian methods · VAST, preferred | truth-tracking score | Improved |
| use labelling aggregation instead of other COS | COS sco · VAST, preferred | accuracy (AM) | Hurt |
Closest papers by meaning
- Debate or Vote: Which Yields Better Decisions in Multi-Agent Large Language Models? (NeurIPS 2025 · similarity 0.85)
- When Debate Helps: Proposal Supply and Verification-Aware Readout in Multi-Agent Reasoning (NeurIPS 2026 · similarity 0.84)
- Collaborative Disagreement Resolution for Scalable Oversight (ICML 2026 · similarity 0.84)
- Efficient Multi-Agent Reasoning via Confidence-Guided Adaptive Debate (ICML 2026 · similarity 0.84)
- DEBATE: A Large-Scale Benchmark for Evaluating Opinion Dynamics in Role-Playing LLM Agents (NeurIPS 2026 · similarity 0.84)
- Multi-Agent Debate for LLM Judges with Adaptive Stability Detection (NeurIPS 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
