Enregistré dans:
Détails bibliographiques
Auteurs principaux: Langner, Mikołaj, Eliasz, Jan, Rudnicka, Ewa, Kocoń, Jan
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2511.03830
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866911250961661952
author Langner, Mikołaj
Eliasz, Jan
Rudnicka, Ewa
Kocoń, Jan
author_facet Langner, Mikołaj
Eliasz, Jan
Rudnicka, Ewa
Kocoń, Jan
contents We introduce a method for efficient multi-label text classification with large language models (LLMs), built on reformulating classification tasks as sequences of dichotomic (yes/no) decisions. Instead of generating all labels in a single structured response, each target dimension is queried independently, which, combined with a prefix caching mechanism, yields substantial efficiency gains for short-text inference without loss of accuracy. To demonstrate the approach, we focus on affective text analysis, covering 24 dimensions including emotions and sentiment. Using LLM-to-SLM distillation, a powerful annotator model (DeepSeek-V3) provides multiple annotations per text, which are aggregated to fine-tune smaller models (HerBERT-Large, CLARIN-1B, PLLuM-8B, Gemma3-1B). The fine-tuned models show significant improvements over zero-shot baselines, particularly on the dimensions seen during training. Our findings suggest that decomposing multi-label classification into dichotomic queries, combined with distillation and cache-aware inference, offers a scalable and effective framework for LLM-based classification. While we validate the method on affective states, the approach is general and applicable across domains.
format Preprint
id arxiv_https___arxiv_org_abs_2511_03830
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Divide, Cache, Conquer: Dichotomic Prompting for Efficient Multi-Label LLM-Based Classification
Langner, Mikołaj
Eliasz, Jan
Rudnicka, Ewa
Kocoń, Jan
Computation and Language
We introduce a method for efficient multi-label text classification with large language models (LLMs), built on reformulating classification tasks as sequences of dichotomic (yes/no) decisions. Instead of generating all labels in a single structured response, each target dimension is queried independently, which, combined with a prefix caching mechanism, yields substantial efficiency gains for short-text inference without loss of accuracy. To demonstrate the approach, we focus on affective text analysis, covering 24 dimensions including emotions and sentiment. Using LLM-to-SLM distillation, a powerful annotator model (DeepSeek-V3) provides multiple annotations per text, which are aggregated to fine-tune smaller models (HerBERT-Large, CLARIN-1B, PLLuM-8B, Gemma3-1B). The fine-tuned models show significant improvements over zero-shot baselines, particularly on the dimensions seen during training. Our findings suggest that decomposing multi-label classification into dichotomic queries, combined with distillation and cache-aware inference, offers a scalable and effective framework for LLM-based classification. While we validate the method on affective states, the approach is general and applicable across domains.
title Divide, Cache, Conquer: Dichotomic Prompting for Efficient Multi-Label LLM-Based Classification
topic Computation and Language
url https://arxiv.org/abs/2511.03830