Search results

Tearing us Apart

Book review of: N. Carr (2025). Superbloom. How Technologies of Connection Tear Us Apart, W. W. Norton & Company, 272 pages.

The Impact of Quantization on Retrieval-Augmented Generation: An Analysis of Small LLMs

Post-training quantization reduces the computational demand of Large Language Models (LLMs) but can weaken some of their capabilities. Since LLM abilities emerge with scale, smaller LLMs are more sensitive to quantization. In this paper, we explore how quantization affects smaller LLMs’ ability to perform retrieval-augmented generation (RAG), specifically in longer contexts. We chose personalization for evaluation because it is a challenging domain to perform using RAG as it requires long-context reasoning over multiple documents. We compare the original FP16 and the quantized INT4 performance of multiple 7B and 8B LLMs on two tasks while progressively increasing the number of retrieved documents to test how quantized models fare against longer contexts. To better understand the effect of retrieval, we evaluate three retrieval models in our experiments. Our findings reveal that if a 7B LLM performs the task well, quantization does not impair its performance and long-context reasoning capabilities. We conclude that it is possible to utilize RAG with quantized smaller LLMs.

Informal care and diversity in OT education

Integration of Large Language Models in the Public Sector

Kognitiv verhaltenstherapeutisches Behandlungsprogramm zur Verbesserung des Selbstbildes bei forensisch psychiatrischen Patienten mit einer psychotischen Störung und Paranoia. Eine randomisierte kontrollierte Studie

Menschen mit einer psychotischen Störung haben häufig ein negatives Selbstbild und damit ein geringes Selbstwertgefühl. Untersuchungen haben auch gezeigt, dass Paranoia direkt mit negativen Vorstellungen über das Selbst zusammenhängen kann. Sich minderwertig zu fühlen und an Paranoia zu leiden, kann zu Verletzlichkeit und Aggression führen. In der forensischen Psychiatrie gibt es eine Kategorie stationärer Patienten mit einer psychotischen Erkrankung und Paranoia. Diese Patienten leiden schwer unter ihrer Störung, die obendrein ein Prädiktor für einen ungünstigen Behandlungsverlauf ist und eine lange Verweildauer innerhalb der (forensischen) Psychiatrie nach sich zieht. In einer randomisierten kontrollierten Studie wird die Wirksamkeit des Competitive Memory Training (COMET), eines kognitiv-verhaltenstherapeutischen Behandlungsprogramms zur Verbesserung des Selbstbildes bei in der Forensik untergebrachten psychotischen Patienten, die ebenfalls an Paranoia leiden, evaluiert und untersucht, ob ein positiveres Selbstbild auch zur Verringerung der Paranoia dieser Patienten führt. Die an einer kleinen Stichprobe durchgeführte Studie ergab eine signifikante Verbesserung des Selbstwertgefühls, aber keine Verringerung der Paranoia bei den Patienten, die an COMET teilnahmen.

Discomfort and factual recollection in intensive care unit patients

INTRODUCTION: A stay in the intensive care unit (ICU), although potentially life-saving, may cause considerable discomfort to patients. However, retrospective assessment of discomfort is difficult because recollection of stressful events may be impaired by sedation and severe illness during the ICU stay. This study addresses the following questions. What is the incidence of discomfort reported by patients recently discharged from an ICU? What were the sources of discomfort reported? What was the degree of factual recollection during patients' stay in the ICU? Finally, was discomfort reported more often in patients with good factual recollection?METHODS: All ICU patients older than 18 years who had needed prolonged (>24 hour) admission with tracheal intubation and mechanical ventilation were consecutively included. Within three days after discharge from the ICU, a structured, in-person interview was conducted with each individual patient. All patients were asked to complete a questionnaire consisting of 14 questions specifically concerning the environment of the ICU they had stayed in. Furthermore, they were asked whether they remembered any discomfort during their stay; if they did then they were asked to specify which sources of discomfort they could recall. A reference group of surgical ward patients, matched by sex and age to the ICU group, was studied to validate the questionnaire.RESULTS: A total of 125 patients discharged from the ICU were included in this study. Data for 123 ICU patients and 48 surgical ward patients were analyzed. The prevalence of recollection of any type of discomfort in the ICU patients was 54% (n = 66). These 66 patients were asked to identify the sources of discomfort, and presence of an endotracheal tube, hallucinations and medical activities were identified as such sources. The median (min-max) score for factual recollection in the ICU patients was 15 (0-28). The median (min-max) score for factual recollection in the reference group was 25 (19-28). Analysis revealed that discomfort was positively related to factual recollection (odds ratio 1.1; P < 0.001), especially discomfort caused by the presence of an endotracheal tube, medical activities and noise. Hallucinations were reported more often with increasing age. Pain as a source of discomfort was predominantly reported by younger patients.CONCLUSION: Among postdischarge ICU patients, 54% recalled discomfort. However, memory was often impaired: the median factual recollection score of ICU patients was significantly lower than that of matched control patients. The presence of an endotracheal tube, hallucinations and medical activities were most frequently reported as sources of discomfort. Patients with a higher factual recollection score were at greater risk for remembering the stressful presence of an endotracheal tube, medical activities and noise. Younger patients were more likely to report pain as a source of discomfort.