RESOLVENOW at UZH Shared Task 2026: Rule-Based Type Classification with LLM-Driven Multi-Label Tagging for UN Resolutions

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

Summary

The RESOLVENOW system, submitted for Subtask 1 of the UZH Shared Task 2026, addresses paragraph-level classification and multi-label tagging of UN resolutions. This system classifies paragraphs as preambular or operative and assigns tags from a 141-code, 15-dimension taxonomy. It achieves a type macro-F1 of 0.910 on the official silver standard for paragraph type classification using a deterministic French-English lexical classifier, requiring no LLM calls for this step. For multi-label tagging, the system employs DeepSeek-R1-0528-Qwen3-8B, which predicts tags via a single merged prompt exposing the full taxonomy. This hybrid approach resolves issues faced by earlier pipelines, which were either too slow or dropped relevant tags.

Key takeaway

For NLP Engineers developing document classification systems for complex, multi-task scenarios, consider a hybrid architecture. Your team should integrate deterministic, rule-based classifiers for high-confidence, low-latency subtasks, like initial paragraph typing. Then, deploy a powerful LLM, such as DeepSeek-R1-0528-Qwen3-8B, for intricate multi-label tagging, ensuring the prompt exposes the full taxonomy to maximize accuracy and context. This approach can significantly improve both efficiency and performance.

Key insights

Combining deterministic rule-based classification with LLM-driven multi-label tagging effectively processes complex document structures.

Principles

• Hybrid systems can overcome individual model limitations.
• Deterministic rules offer speed and accuracy for specific subtasks.
• LLMs can handle complex, high-dimensional tagging with full context.

Method

First, a deterministic French-English lexical classifier assigns paragraph types. Second, DeepSeek-R1-0528-Qwen3-8B predicts multi-labels using a single merged prompt that exposes the entire 141-code taxonomy.

In practice

• Use rule-based methods for high-precision, low-cost classification.
• Design LLM prompts to expose full taxonomies for comprehensive tagging.

Topics

• UN Resolutions
• Multi-label Tagging
• Text Classification
• LLM-as-Judge
• Rule-Based Systems
• DeepSeek-R1-0528-Qwen3-8B

Best for: AI Scientist, NLP Engineer, Research Scientist

Related on AIssential

• Multi-Label Text Classification with Scikit-LLM — Scikit-LLM enables zero-shot multi-label text classification using LLMs with a scikit-learn-like workflow.
• Tagging my blog posts with BERTopic and LLMs — Blending BERTopic with LLMs and agentic harnesses efficiently creates robust content tagging systems for personal projects.
• Comparing LLM and Fine-Tuned Model Performance on NVDRS Circumstance Extraction with Varying Prompt Complexity — LLMs, guided by a context-aware "Complexity Score" for prompt selection, significantly outperform fine-tuned models on inferentially complex, low-prevalence data extraction.
• Indigenous Letters to Brazil: Multi-Label Classification — Multi-label classification of indigenous letters presents challenges due to corpus imbalance and semantic overlap.
• Text Classification at Scale: Why It’s Harder Than It Looks, and How to Think About It — Text classification at scale requires a systems approach focused on understanding, not just pattern matching, to overcome LLM limitations and high-cardinality challenges.
• Beyond Surface Forms: A Comprehensive, Mechanism-Oriented Taxonomy of Indirect Linguistic Encoding for LLM-Based Coded Language Detection — A mechanism-oriented taxonomy of indirect linguistic encoding significantly enhances LLM-based coded language detection.

Open in AIssential →

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