small language model

GEOSTEER: Geodesic Optimization for Activation Steering in Large Language Models

GeoSteer is proposed, an optimization-based method for norm-preserving activation steering that consistently improves over state-of-the-art activation steering baselines and replaces predefined one-step edits with adaptive, geometry-aware optimization.

Chain-of-Thought supervision enables explainable and efficient medical summarization with Small Language Models

Major Change in Human Life: The Decline of the Traditional PC and the Rise of the Supercomputer PC

The "death of the normal PC" should not be seen as the extinction of computers used for routine work, rather, it symbolizes the demise of the conventional computing paradigm, where the PC is essentially a passive instrument for executing programs.

Extracting AI agent-accessible data from biodiversity literature with corpus

[1.4.0] - 2026-09-09 Theme — v1.4 silent wrongs and the hazards behind them v1.3 made the pipeline show where evidence came from. This cycle went after what survives behind that: results that are wrong without saying so, and build hazards that cost hours per run without ever failing. The organizing complaint is that a...

One and the Same log_2 Counts a Ratio in One Law and Items in the Other ── Fitts's Law Stays at 4.3219 When D and W Are Scaled Together ── Hick's Law Turns 32 Times the Choices into Only 2.7333 Times the Time ── [Paper 345]

Fitts’s law and Hick’s law both say that time goes as log_2. This paper asks whether they are saying the same thing──the answer is that what is inside the log_2 differs: a dimensionless ratio in one, a count in the other. No new mathematical theorem and no new law is claimed. Scope of this paper (scope note): No new ma...

The Size of the Hole Is Not in the Stress Concentration ── Two Elliptical Holes Differing 10000-Fold in Area Carry the Same K_t=21.0 ── What Enters Is the Sharpness a/b Alone: 3.0 for a Circular Hole and 201.0 at a/b=100 ── [Paper 343]

Cut a hole in a plate and the stress rises at its edge. This paper asks what decides how far it rises──the answer is not the size of the hole but only its sharpness. No new mathematical theorem and no new law is claimed. Scope of this paper (scope note): No new mathematical theorem and no new law is claimed──Inglis’s s...

Same Atomic Number, Same Charge, Same Electrons: Change Only the Mass and the Rate Moves Sevenfold ── The C--H to C--D frequency ratio is 0.734230, which is not 1/sqrt(2)=0.707107 ── A zero-point difference of 4.7690 kJ/mol makes k_H/k_D=6.8468 at room temperature ── [Paper 341]

Replace hydrogen with deuterium and a reaction rate changes. This paper asks why changing only the mass moves the rate──the answer is that the energy of standing still is set by the mass. No new mathematical theorem and no new law is claimed. Scope of this paper (scope note): No new mathematical theorem and no new law...

Selberg–Residue Analysis of the Equally Spaced Richardson–Gaudin Pairing Model

Authorship and verification statement. The scientific problem, the original methodological direction, and the research objectives were formulated by W. V. Pogosov. The analytical derivations and numerical computations in this project were carried out by an advanced large language model (LLM), through an iterative inter...

The Forced Alternative Is a Filter: What a Constructed Navier-Stokes Blowup Proves, What It Does Not, and One Order-Type Demand on Its Correction Tower

A direct reading of the OpenAI forced Navier-Stokes construction of 8 September 2026. On 8 September 2026 a construction was released, under the author line OPENAI, claiming a smooth compactly supported force under which a solution of the three-dimensional incompressible Navier--Stokes equations develops unbounded velo...

An Empirical Benchmarking of Traditional Machine Learning and DistilBERT-Based Zero-Shot Hierarchical Sentiment Analysis on Large-Scale Twitter Data

A zero-shot classification pipeline that uses Natural Language Inference for classification of 1,000 tweets into a two-layer taxonomy of 30 parent topics and 330 subtopics has been created without any human-labeled samples.

Augmented Screens, Not Augmented Reality: Why Most Smart Glasses Displays Fail the Spatial Computing Test

This paper applies the categorical test for genuine spatial computing to consumer smart glasses and argues that most current display-equipped models are augmented screens, showing text and icons fixed to the wearer's head rather than content positioned in the three-dimensional space around them.The paper separates two...

Editorial: Machine learning algorithms and software tools for early detection and prognosis of schizophrenia

Schizophrenia and related psychotic disorders exhibit complex neurobiological mechanisms with variable disease characteristics. Early diagnosis currently relies on clinical assessments alone, since validated biomarkers and generalized prediction models are limited. Recent advances in machine learning (ML) have created...