small language model

Analysis Code for: Dissociation Between Clinical Decision Accuracy and Citation Reliability Across Large Language Models in Small Renal Mass Management

Analysis scripts accompanying the study “Dissociation Between Clinical Decision Accuracy and Citation Reliability Across Large Language Models in Small Renal Mass Management.” The scripts reproduce the statistical analyses reported in the manuscript, including inter-rater reliability (Cohen’s kappa), model comparisons...

The Eigenweight Engine

We demonstrate that the weight matrices of multiple large language models sharing the same architecture but trained for different purposes (code generation vs conversational instruction-following) occupy a shared low-rank manifold. By computing a joint eigenspace (via SVD) across model weights, we factor N separate mod...

The Eigenweight Engine

We demonstrate that the weight matrices of multiple large language models sharing the same architecture but trained for different purposes (code generation vs conversational instruction-following) occupy a shared low-rank manifold. By computing a joint eigenspace (via SVD) across model weights, we factor N separate mod...

The Humidity Goes from 53.80 Per Cent to 100 Without a Drop of Water Being Added ── lift the parcel 1250 m and temperature and dew point meet at exactly 12.75 degrees ── cloud bases are flat because only height decides condensation ── the strength of the updraught does not appear in the formula ── [Paper 399]

Humidity rising sounds like water being added. This paper shows that the humidity can be taken to 100 per cent with the amount of water untouched──it is enough to lift the air. No new theorem or law is claimed. Scope of this paper (scope note): No new theorem or law is claimed──the lifting condensation level, the dry a...

Put a Number on a Pendulum's "Small Angle" and Ten Degrees Costs 164.78 Seconds a Day ── two degrees costs 6.58 s/day and thirty degrees 1504.12 s/day (25 minutes) ── yet at 170 degrees the period is only 2.4394 times, and it reaches three times at 175.8765 degrees ── the divergence bites only in the last fraction of a degree ── [Paper 396]

We are taught that a pendulum is isochronous if the amplitude is small. This paper asks how small is small──even ten degrees costs 164.78 seconds a day, which is useless for a clock. No new theorem or law is claimed. Scope of this paper (scope note): No new theorem or law is claimed──the exact period of a simple pendul...

What Actually Fits: Context Selection for Local-First Retrieval on Memory-Constrained Hardware

Running a language model on your own machine keeps your files private, but consumer hardware imposes a context budget that server deployments never feel. We report measurements from a deployed local-first system that indexes a user's files and answers questions over them. Early versions injected whole-project content i...

A Diagnostic Evaluation of Small Language Models for Offline Writing-Feedback Tutoring in Brazilian Portuguese: A Study on Structural and Pedagogical Adherence under Public-School Infrastructure Constraints

Diagnostic benchmark of eight open Small Language Models (≤3.8B parameters) for writing-feedback tutoring in Brazilian Portuguese, evaluated strictly offline on CPU against a metalinguistic rubric derived from Koch's Textual Linguistics. Includes a four-isolation falsification protocol on the Llama 3.2 family, supporte...

Observed Recoverable Behavioral Failures in LLM Workflows

Standard large-language-model evaluations primarily score terminal outputs: whether a final answer is correct, safe, preferred, or useful. This paper studies a recurring class of workflow failures referred to here as recoverable behavioral failures (RBFs). In the operational usage adopted in this paper, an RBF occurs w...

Employing General-Purpose and Biomedical Large Language Models with Advanced Prompt Engineering for Pharmacoepidemiologic Study Design

Abstract Background The potential of large language models (LLMs) to automate and support pharmacoepidemiologic study design is an emerging area of interest, yet their reliability remains insufficiently characterized. General-purpose LLMs often display inaccuracies, while the comparative performance of specialized biom...

Only 2 of 154 Most spoken Languages from 23 Language Families Employ a Labial for "Tongue" in 1st Consonant Position Demonstrating Embodied Iconicity for Tongue Vs Mouth in "inverse" Control

Class-Level Embodied Iconicity in Words for TONGUE and MOUTH: Evidence From 154 Major Languages, With Genealogical Controls and a Balanced-Sample Replication Author: M. Phillips Affiliation: Independent Contact: mg@mgcode.mozmail.com Version: v6 (September 2026). Supersedes v4 and 5 (Zenodo record 18391312). Concept DO...

A Preliminary Study on Spanish–Zapoteco Machine Translation Using a Multi-stage Fine-Tuning M2M100 Model in a Low-Resource Setting

spec-resistance corpus: 30-model, 627,491-trial dataset for "Large language models that near-perfectly evaluate products systematically recommend a worse familiar brand"

Primary computational corpus for the manuscript "Large language models that near-perfectly evaluate products systematically recommend a worse familiar brand" (Affonso, Spears School of Business, Oklahoma State University). Contains 627,491 controlled product-recommendation trials across 30 frontier language models from...