Argchestrators at UZH Shared Task 2026: Efficient Argument Mining in UN Resolutions: A Sub-8B Pipeline using Agentic Debate and Heuristic Retrieval

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Advanced, quick

Summary

The Argchestrators system, submitted to the ArgMining 2026 Shared Task, addresses the challenge of efficient argument mining within the highly formal language of United Nations resolutions. This system adheres to a strict constraint, utilizing open-weight models with a maximum of 8 billion parameters, specifically powered by Qwen3-8B. It employs a hybrid, compute-efficient architecture that integrates several components. For classifying preambular-operative sections, it uses deterministic rules tailored to UN documents. The system further incorporates an LLM-based multi-label classifier to identify thematic dimensions and a directed-graph extraction approach for predicting argumentative relations. This pipeline aims to reconstruct reasoning in UN resolutions effectively under tight computational limits.

Key takeaway

For NLP Engineers developing argument mining solutions for highly formal texts like UN resolutions, this work demonstrates that effective systems can be built within strict computational constraints. You should consider hybrid architectures combining smaller LLMs, such as Qwen3-8B, with deterministic rules and graph-based methods. This approach allows for robust performance without requiring large-scale models, optimizing resource use for specialized tasks.

Key insights

Efficient argument mining in UN resolutions is achievable with a sub-8B parameter hybrid LLM pipeline using deterministic rules and graph extraction.

Principles

• Hybrid architectures can meet strict compute limits.
• Deterministic rules enhance LLM performance on formal texts.
• Graph-based methods aid argumentative relation extraction.

Method

The system classifies preambular-operative sections via deterministic rules, uses an LLM for thematic dimensions, and extracts argumentative relations with a directed-graph approach, all powered by Qwen3-8B.

In practice

• Use Qwen3-8B for sub-8B argument mining.
• Apply deterministic rules for formal document classification.
• Employ graph extraction for complex argumentative relations.

Topics

• Argument Mining
• UN Resolutions
• Qwen3-8B
• Large Language Models
• Hybrid Architectures
• Natural Language Processing

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• From Argument Components to Graphs: A Multi-Agent Debate with Confidence Gating for Argument Relations — Multi-agent debate with confidence gating improves training-free argument relation identification and offers interpretability.
• Modeling Co-Pilots for Text-to-Model Translation — A unified, solver-agnostic approach using LLMs and \textsc{MiniZinc} translates natural language to formal optimization models.
• CAF-Gen: A Multi-Agent System for Enriching Argumentation Structures — Multi-agent LLM systems with iterative Creator-Reviewer pipelines reliably enrich shallow argument structures into complex formal models.
• Beyond Vector Similarity: A Structural Analysis of Graph-Augmented Retrieval for Industrial Knowledge Graphs — RAG struggles with structural reasoning; LLMs require explicit graph computational operators.
• InfiniteScienceGym: An Unbounded, Procedurally-Generated Benchmark for Scientific Analysis — Procedurally generated benchmarks offer controlled, scalable evaluation for LLM scientific reasoning and abstention.
• Evaluating LLM-Driven Summarisation of Parliamentary Debates with Computational Argumentation — A new framework evaluates LLM summaries of parliamentary debates by preserving argumentative reasoning.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.