large language models

Scalable Training of Spatially Grounded 2D Vision–Language Models for Radiology

The Alignment Paradox of Medical Large Language Models in Infertility Care: Decoupling Algorithmic Improvement From Clinical Decision-Making Quality

Outcome-based metrics alone are insufficient proxies for clinical utility in ART decision support and algorithmic improvement and clinical alignment suggest a possible decoupling; a phenomenon the authors term the alignment paradox.

The Immutable Tensor Architecture: A Pure Dataflow Approach for Secure, Energy-Efficient AI Inference

The deployment of Large Language Models (LLMs) on consumer edge devices is throttled by the “Memory Wall”—the prohibitive bandwidth and energy cost of fetching gigabytes of model weights from DRAM for every token generated. Current architectures (GPUs, NPUs) treat model weights as mutable software data, incurring massi...

How Instruction-Tuning Imparts Length Control: A Cross-Lingual Mechanistic Analysis

Longitudinal and Transverse Belong to the Field; x and y Belong to the Coordinates ── rotating the coordinates leaves the longitudinal share at 0.647222222222 without moving a digit, while the x share moves with every rotation ── both are called "splitting the field in two", and only one of them belongs to the field ── [Paper 517]

There are many ways to split a vector field in two. What this paper shows is that which of them belongs to the field and which to the coordinates can be measured by rotating the coordinates. No new theorem or law is claimed. Scope of this paper (scope note): No new theorem or law is claimed──the Helmholtz decomposition...

Spectral SU(3) Gauge Invariant on Four-Dimensional Euclidean Lattices: Wilson Loop Expectation Values, Area Law Confinement, and the Non-Perturbative Mass Gap

Complete lattice computation of Wilson loop expectation values under SU(3) gauge theory on a four-dimensional Euclidean lattice. Establishes color confinement via the area law and extracts the non-perturbative Yang-Mills mass gap of 77.71 MeV from the two-point glueball correlator spectral decomposition. ⚖️ PROPRIETARY...

Goal-Questioning Beyond Explicit Conflict: Latent Conflict Discovery and Re-Dimensioning in CHORDI Meta-Reasoning

As large language models (LLMs) are increasingly used for autonomous decision-making and action support, it becomes important not only for them to achieve a given objective efficiently, but also to question whether that objective should be optimized as stated. This study positions Goal-Questioning as a Meta-Reasoning c...

Genesis Cube or Squid Cube? A Dual-Use Framework for Artificial Collective Organisms

Large-language-model agents can be combined into persistent systems with differentiated roles, memory, coordination, tools and operational continuity. Yet aggregation, scale and commercial productivity alone do not create an organism. This paper introduces the concept of the artificial collective organism: a bounded di...

TCGA_brest_extraction

# Structured extraction of breast cancer pathology reports by five large language models (TCGA cohort) ## Description This dataset accompanies the manuscript *"Secure large language models for structured breast cancerpathology extraction: a consensus- and embedding-based evaluation"* (Astruc et al., manuscript inprepar...

The prose penalty in AI-assisted authorship

Generative AI based on large language models (LLMs) is increasingly used in academic, journalistic, and other professional and online writing, yet its implications for human authorship remain contested. We asked whether assignment of authorship credit depends on which task within the larger writing process a writer del...

Light Is Gathered by a Gravitational Lens, Yet the Curvature of the Geometry It Travels In Is Negative ── The deflection is the curvature integrated over the region outside the ray, by the Gauss-Bonnet theorem: tracing the ray gives 4.011823809924e−3 and the outside integral 4.000000000000e−3. Integrating inside diverges ── the separator is whether the region lies inside or outside the ray ── [Paper 1044]

Gravity bends light and gathers it as a lens does. Yet measure the curvature of the very geometry that light travels in, and it is negative. This paper counts that discrepancy as a difference in the region of integration. No new theorem and no new law is claimed. Scope of this paper (scope note): No new theorem and no...

Slicing a Light Sphere by Another Frame's Simultaneity Gives an Ellipse of Eccentricity β with the Emission Point at a Focus ── The radius in each direction is exactly the Doppler factor for that direction, the two vertices are k and 1/k, the semi-major axis is γ, the semi-minor axis is cT, and the centre is where the moving observer stands ── the separator is whose simultaneity does the cutting ── [Paper 1036]

Light emitted at the origin is said to spread as a sphere in every frame. This paper counts what comes out when that sphere is cut by the simultaneity of another frame. No new theorem and no new law is claimed. Scope of this paper (scope note): No new theorem and no new law is claimed──the Lorentz transformation, the i...