small language model

Idiomaticity Detection with Span Contrastive Loss

Counting intersections of rational cyclic grids: an application of Smith normal form

Choose two points from equally spaced grids on a circle. How many choices put both their sum and their difference on specified target grids? A small example leads from fractions to integer congruences. We then use classical Smith normal form to obtain the exact count, carefully allowing the two input grids to have diff...

Dosezy

v2.5.9: Custom Multi-Week Recurrence, Android 15 Edge-to-Edge, and Cross-Platform Schema Parity Summary of Changes Release v2.5.9 introduces custom multi-week recurrence scheduling (e.g. taking medications every 2 or 3 weeks on specific days), enforces strict portrait orientation across all activities, normalizes devic...

Agreement Between Large Language Models and Humans in Research Proposal Review - Data and Scripts

Agreement Between Large Language Models and Humans in Research Proposal Review — Data and Code This repository contains the data and code required to reproduce the analyses, statistical results, and figures presented in the associated manuscript. Files are organized by function and described below. All research proposa...

Beyond similarity retrieval: governance fidelity, consistency, and minimality in long-term conversational memory

Conversational language model systems persist user constraints, preferences, identity, commitments, and assigned roles across turns. The dominant memory architecture stores this alongside ordinary episodic content in a vector index and retrieves the top-k items by similarity to the active query. We argue that this desi...

Artificial intelligence readiness among audiovisual media professionals in Saudi Arabia and its barriers and facilitating conditions

Artificial intelligence is reshaping audio-visual content production, yet adoption among media professionals in Arab contexts remains under-researched, and existing studies rarely integrate individual, structural, and institutional explanations within a single analytical framework. This study examines the barriers to a...

Artificial intelligence in male infertility from diagnosis to treatment

Artificial intelligence has the potential to transform clinical decision-making for patients presenting with infertility by enabling objective, high-throughput, and personalised assessment of semen samples, but these tools should be used to augment rather than replace clinical expertise until models have been prospecti...

Mapping generative AI governance readiness in Kurdistan Region universities

It is argued that universities should move beyond plagiarism control toward transparent, multilingual, locally grounded AI governance that supports academic integrity, responsible learning, assessment fairness, data protection, and student and staff AI literacy.

Works for 10/6/2026 - Sergei Pavlovich Korolev

Listen and ask questions on Gemini Notebook: https://notebook.google.com/notebook/31300077-8066-4826-bca1-faf9fe8dbd2e?authuser=1 The Little Twist in the Middle: History, Accessibility, Logos, and the Problem of Making the Other Reachable develops a unified account of how accumulated history changes what becomes access...

Using Artificial Intelligence to estimate how common alcohol references and alcohol-related brand references are among influencers on Instagram

Background: Young people spend hours on social media every day, where following influencers is the norm. Influencers post frequently about alcohol and alcohol brands and there has been a rise in influencer marketing. However, prevalence estimates rely on manual analyses of only a small number of posts. We used a large...

COMPASS: Finding Where Reasoning Lives in Language Models

Comparing three model families and multiple math benchmarks, COMPASS outperforms the activation-steering baselines the authors compare against, improves GSM8K accuracy by 16 percentage points on average, and approaches CoT accuracy with 20-70\% fewer generated tokens.

Thin Evidence, Thick Priors: How Language Models Substitute Identity for Missing Financial Facts

Advisory systems built on large language models should be audited at the disclosure levels users actually reach, and judged across the whole identity space rather than one attribute at a time.