small language model

"A Dataset for Comparative Performance Analysis of On-Device and Edge-Assisted Small Language Model Inference on Android Devices"

" This dataset presents experimental measurements for comparing on-device and edge-assisted text summarization using a small language model (SLM) on an Android device. The Android application is developed in Kotlin, with local inference performed using the MediaPipe LLM Inference runtime and edge inference supported by...

Can Zero-Shot Large Language Models Be Trusted for Credit-Risk Prediction? An Explainability Audit Against LightGBM on Loan Default Data

Zero-shot large language models (LLMs) offer a low-barrier route to credit-risk prediction, but regulated use requires evidence on discrimination, probability reliability, and auditable decision patterns. We compare four open-source LLMs (Gemma-2-9B, Llama-3.2-3B, Mistral-7B-v0.3, and Qwen2.5-7B) with LightGBM on 396,0...

Self-other value alignment shapes sustainable decisions

Sustainable behaviour is often viewed as requiring personal sacrifice to benefit others, yet how individuals integrate self- and other-regarding values in sustainable decisions remains poorly understood. Here we investigate a two-dimensional framework of self-other value alignment across four studies comprising seven i...

Sounding Like Which AI? Machine-Style Overlap of arXiv Abstracts with Language Models of Different Vintages

We ask not how much text was produced with LLM help but how much writing, by any author, falls on the LLM side of a stylistic boundary, and which LLMs define that side. We define Machine-Style Overlap (MSO): the share of documents that a calibrated classifier on style features that ignore content-word identity judges a...

FedCuraLM: Reliability-Calibrated and Resource-Aware Label Auditing with Small Language Models for Non-IID Federated AIoT

Federated Artificial Intelligence of Things (AIoT) activity-recognition systems learn from decentralized wearable and smart-home records with potentially incorrect labels, non-IID client distributions, and unequal edge-device resources. Our objective is to improve downstream learning from these labels while limiting ha...

"RFJ-Bench: A Multimodal Benchmark Dataset for Rheumatic Finger Joint X-Ray Imaging and Automated Report Generation"

"RFJ-Bench is a multimodal benchmark dataset for automated radiology report generation in rheumatology. It contains 2,239 finger joint posteroanterior (PA) X-ray examinations from 2,124 unique patients, retrospectively collected at Dazhou Central Hospital between October 2020 and March 2024 under an IRB-approved protoc...

IoT Devices as Layers: But What Is There to Autocomplete?

Connected consumer devices form an installed substrate of electricity, connectivity and heterogeneous computing resources. Their owners may control the physical equipment while suppliers retain authority to update its software and direct its behavior through device-initiated management sessions. This paper examines a c...

Beyond Generation: What a Decision-Native Model Does Well in a Long Task, and What It Needs from Its Harness

Many automated workflows end in a bounded decision among options that the software already knows, yet such decisions are usually obtained from language models by generating text. Decision-native models instead return one of the supplied options together with a probability distribution. We study one of them, Jev, on the...

Does More Deliberation Help? A Comparative Evaluation of Chain-of-Thought Prompting Strategies for Grade-School Math Reasoning in Open-Source Instruction-Tuned Language Models

Two Pre-Registered Nulls in Alternative Data: Insider Purchases and Legal-Risk Language in SEC Filings, 2014–2026

We report two pre-registered tests of popular alternative-data signals built from U.S. Securities and Exchange Commission (SEC) filings, each rebuilt end to end from public sources. Insider purchases: across 31,971 tradeable officer and director open-market purchases (Form 4, 2014 to mid-2026), the 63-day return in exc...

Model Collapse in Bangla and English: A Pre-Registered Study of Iterative Self-Training with Gemma-3-270m

Pre-registration of a computational experiment on model collapse in a low-resource language setting, frozen before any model is trained. We fine-tune the open-weight model google/gemma-3-270m (bfloat16) in Bangla and in English over six generations of self-training, where each generation is trained on text generated by...

Scientific Super Intelligence: From Science Tokens to ScienceRSI

Scientific reasoning moves beyond natural language. The search for a discovery runs across sequences, structures, fields, omics, images and measurements, while language models reason mainly in text. This Perspective sets out four proposals from the Scinetics team. First, we define science tokens: modality-appropriate u...