Newest papers

Asking a Multimodal LLM Whether Someone Fell: Zero-Shot Screening of Low-Resolution Thermal Video in a Laboratory and an Occupied Home

Marcos Junca, Mauro Junca · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.29913 · cited by 0 works (OpenAlex) · open-access PDF

Spanish

Las dos cámaras térmicas de lanzamiento, pero las cámaras de radiodifusión no pueden vigilar los datos convencionales de las personas mayores que viven solas sin capturar caras, y los modelos multimodales de gran lenguaje en tierra (LLMs) pueden juzgar un vídeo corto que cae dentro de una formación específica de tareas. Evaluaremos una producción multimodal LLM, Gemini 3.8 Flash en vivo a temperatura 0, como la fase de decisión de una pantalla de caída sin caídas, usada exactamente como un prototipo de tierra desplegado: una solicitud por cortos clips térmicos, con su pista de caída de sonido cuando hay uno, y su predecesor Gemini 3.7 Flash en las mismas solicitudes. En el 438 el 99% de las personas que se encuentran en el suelo y salen de la

Machine translation (Meta M2M100); the English original follows.

Abstract

Thermal cameras can watch over older people living alone without capturing faces, and multimodal large language models (LLMs) can judge a short video without task-specific training. We evaluate a production multimodal LLM, Gemini 3.8 Flash at temperature 0, as the decision stage of an event-triggered fall screener, used exactly as a deployed prototype uses it: one request per short thermal clip, with its sound track when there is one, and its predecessor Gemini 3.7 Flash on the same requests. On 438 staged falls and 563 non-fall windows that we located by hand in a public 32x24 thermal dataset, the deployed request called 69% of the held-out subjects' falls (67% over all subjects; bed falls excluded) and none of 499 non-fall windows, including people lying on the floor and getting up. Over 76 h of ordinary life in an occupied room, recorded at 160x120 with sound, it called none of 395 clips a fall (95% upper bound 1.2 false alarms per day) and caught 4 of the 9 staged falls the camera could see. One cautionary sentence in the prompt ("a person who goes out of view ... is not a fall") accounted for many of the misses: removing it raised sensitivity on the held-out subjects from 69% to 83% on 3.8 Flash and from 80% to 92% on Gemini 3.7 Flash, and in the home from 4 to 9 of the 9 visible falls, with no false call in ordinary living but at the price of calling deliberate lie-downs a fall. Sensitivity collapsed at 8x6 pixels (34%); identical requests disagreed on 4% of windows; and when the model called a fall, its explanation placed it within a second of the annotated landing 99% of the time. Two conventional detectors trained on the laboratory data caught nearly every held-out laboratory fall but raised 14.2 and 23.9 false alarms per day in the home. We release the frame-level fall annotations of the public dataset and every model answer on it.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: CLIPS, Flash (photography), Computer science, Sentence, Artificial intelligence, Sensitivity (control systems)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange