{"slug": "dualview-leveraging-broad-and-output-specific-evidence-for-llm-based-multi-text", "title": "DUALVIEW: Leveraging Broad and Output-Specific Evidence for LLM-based Multi-Output Text Regression", "summary": "DualView, a parameter-efficient method for LLM-based multi-output text regression, outperformed seven supervised regression baselines across five heterogeneous regression tasks and eight LLMs up to 122B parameters while using 82 to 1842 times fewer fitted parameters than the three strongest baselines, according to the arXiv paper 2610.19206v1. DualView extracts broad and output-specific representations from a frozen LLM and feeds them to a dual-branch architecture that produces the full score vector, with gains holding across individual output dimensions and under limited training data. The authors report that DualView with a smaller open-weight LLM can outperform unsupervised prediction with Gemini 3.1 Pro, and ablations show broad and output-specific evidence provide complementary predictive benefits.", "body_md": "arXiv:2610.19206v1 Announce Type: new \nAbstract: While work on LLM-based regression has largely focused on scalar prediction, multi-output text regression remains less explored. We present DualView, a parameter-efficient approach to leverage LLMs for multi-output text regression. \\method{} first extracts broad and output-specific representations from a frozen LLM, then uses a dual-branch architecture to produce the full score vector. Across five heterogeneous regression tasks and eight LLMs (up to 122B), DualView outperforms seven supervised regression baselines while requiring 82--1842 times fewer fitted parameters than the three strongest baselines. These improvements extend across individual output dimensions and persist even with limited training data. Notably, DualView using a smaller open-weight LLM can outperform unsupervised prediction with Gemini 3.1 Pro. Controlled ablations show that broad and output-specific evidence provide complementary predictive benefits, while lightweight shared predictors achieve strong performance with fewer trainable parameters.", "url": "https://wpnews.pro/news/dualview-leveraging-broad-and-output-specific-evidence-for-llm-based-multi-text", "canonical_source": "https://www.machinebrief.com/news/dualview-leveraging-broad-and-output-specific-evidence-for-l-h8vj", "published_at": "2026-10-15 04:00:00+00:00", "updated_at": "2026-10-15 05:52:42.573566+00:00", "lang": "en", "topics": ["large-language-models", "machine-learning", "ai-research", "natural-language-processing"], "entities": ["DualView", "Gemini 3.1 Pro", "arXiv"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/dualview-leveraging-broad-and-output-specific-evidence-for-llm-based-multi-text", "markdown": "https://wpnews.pro/news/dualview-leveraging-broad-and-output-specific-evidence-for-llm-based-multi-text.md", "text": "https://wpnews.pro/news/dualview-leveraging-broad-and-output-specific-evidence-for-llm-based-multi-text.txt", "jsonld": "https://wpnews.pro/news/dualview-leveraging-broad-and-output-specific-evidence-for-llm-based-multi-text.jsonld"}}