# Multilingual Cognitive Impairment Detection in the Era of Foundation Models

arXiv: https://arxiv.org/abs/2604.06758 · HTML: https://mlsanity.com/paper/2604.06758/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 9 of its 54 references that are in the map.
- First posted on arXiv 8 Apr 2026.

## Builds on (most cited first, 9 of 9)

- [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://mlsanity.com/paper/1908.10084/) (2019)
- [Stanza: A Python Natural Language Processing Toolkit for Many Human Languages](https://mlsanity.com/paper/2003.07082/) (2020)
- [Generalized Shape Metrics on Neural Representations](https://mlsanity.com/paper/2110.14739/) (2021)
- [Alzheimer's Dementia Recognition through Spontaneous Speech: The ADReSS Challenge](https://mlsanity.com/paper/2004.06833/) (2020)
- [Artificial Intelligence, speech and language processing approaches to monitoring Alzheimer's Disease: a systematic review](https://mlsanity.com/paper/2010.06047/) (2020)
- [Explainable Identification of Dementia from Transcripts using Transformer Networks](https://mlsanity.com/paper/2109.06980/) (2021)
- [Alzheimer's disease detection based on large language model prompt engineering](https://mlsanity.com/paper/2501.00861/) (2025)
- [CognoSpeak: an automatic, remote assessment of early cognitive decline in real-world conversational speech](https://mlsanity.com/paper/2501.05755/) (2025)
- [Longitudinal Progression Prediction of Alzheimer's Disease with Tabular Foundation Model](https://mlsanity.com/paper/2508.17649/) (2025)

## Abstract (arXiv)

We evaluate cognitive impairment (CI) classification from transcripts of speech in English, Slovene, and Korean. We compare zero-shot large language models (LLMs) used as direct classifiers under three input settings -- transcript-only, linguistic-features-only, and combined -- with supervised tabular approaches trained under a leave-one-out protocol. The tabular models operate on engineered linguistic features, transcript embeddings, and early or late fusion of both modalities. Across languages, zero-shot LLMs provide competitive no-training baselines, but supervised tabular models generally perform better, particularly when engineered linguistic features are included and combined with embeddings. Few-shot experiments focusing on embeddings indicate that the value of limited supervision is language-dependent, with some languages benefiting substantially from additional labelled examples while others remain constrained without richer feature representations. Overall, the results suggest that, in small-data CI detection, structured linguistic signals and simple fusion-based classifiers remain strong and reliable signals.