Skip to content

Category

small language model

2,925 papers

#large language models Open access Sep 2026

Sufficient for Whose Answer? Three Predicates Under One Word in Selective Retrieval-Augmented Generation, and Why the Only One Tied to Correctness Cannot Be Computed Where the Gate Runs

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. A retrieval-augmented system that answers only when its retrieved context suffices needs a predicate that says when it does. Three incompatible predicates are in circulation under the word "sufficient", and results obtained under different ones are routinely compar...

Pranay M. Mahendrakar · 0 citations
#large language models Open access Sep 2026

How Much of a Leaderboard Ranking Survives Its Own Sampling Error? A Paired-Resolution Audit of Five AI Benchmarks

A benchmark leaderboard is read as a stopwatch: the rank beside a model is treated as a fact. Yet every entrant is scored on the same items and publishes which it solved, so a leaderboard is a paired experiment, and its ordering can be tested rather than trusted. We test each adjacent-rank pair for statistical separabi...

Ilpo Väätäinen · 0 citations
#edge computing Open access Sep 2026

fracpll: charge-pump fractional-N PLL design from measured pieces

First release: the charge-pump fractional-N loop, noise, delta-sigma, tuning-curve, lock and measurement-planning machinery distilled from a monolithic GaN-on-SOI HEMT-varactor PLL co-design study, generalized to any technology that fits a charge-pump loop -- CMOS, SiGe, GaN or board-level. The physics-informed-ML and...

Tanvir M. Mahim · 0 citations
#small language model Open access Sep 2026

Critical Mass Is Set by Skin Balancing Content, Yet a Cube's Extra Volume Does Not Match Its Surface Coefficient ── in a bare one-group diffusion sphere, neutrons leaking through the skin over those gained in the content equal the square of the critical radius over the radius, and a cube's critical-volume factor 1.240490 is smaller than its equal-volume surface coefficient 1.240701 ── the separator is whether the weight of a shape is set by surface area or by the first eigenvalue ── [Paper 1007]

Whether a fission chain reaction keeps going is decided by the balance between neutrons gained in the content and neutrons leaking through the skin. This paper counts that balance in a bare one-group diffusion core (no reflector, no extrapolation distance, lengths measured in the diffusion length L, and the material re...

Yuuki Yamagishi · 0 citations
#small language model Open access Sep 2026

Continued-Fraction Statistics Cannot Tell Known Transcendentals from Unknown Ones or from a Cubic Irrational ── the geometric mean of partial quotients, the growth of denominators and the share of ones all fall within the scatter of random numbers for eight named numbers, and only periodic quadratic numbers and the patterned e stand out ── the separator is whether one speaks of almost all numbers in measure or of this number by name ── [Paper 1021]

Whether π + e, πe or γ is transcendental is not known. On the other hand, it is known that for almost all real numbers the geometric mean of the partial quotients of the continued fraction approaches Khinchin’s constant (Khinchin). This paper expands to 3000 terms the continued fractions of numbers known to be transcen...

Yuuki Yamagishi · 0 citations
#small language model Open access Sep 2026

NHOS Children's Symptoms Checker™ A Client-Side Naive Bayes Pediatric Triage Aid with a 50-Level Triage Severity Score, 3,392-Condition Library, 17-System Coverage, Integrated Interactive Visualization, and a Nine-Module Complete Clinical Upgrade Pack Technical Research White Paper — Architecture, Methodology, Validation Framework, Reproducibility Specification, and Complete Clinical Upgrade Manifest

The NHOS Children's Symptoms Checker™ (NHOS-CSC) is a fully client-side pediatric symptom-triage artifact released as informational research software. It is designed to answer a narrow methodological research question: can a compact, fully inspectable, browser-local pediatric symptom-triage tool expose its assumptions...

Inc. NHOS Labs, Adedapo Ogundiran · 0 citations
#small language model Open access Sep 2026

A laboratory method for AI-assisted science: formal proof, exact arithmetic, controlled experiment, and a claim ledger, tested for a month on the Navier-Stokes problem

Large language models produce mathematics and physics that read well and are sometimes wrong, and the discipline that has grown up around them is verification: proof assistants for proofs, and now public registries that check whether a repository proves what it claims. That discipline is necessary and it stops at the p...

Socrate AI Lab · 0 citations
#small language model Open access Sep 2026

Mersenne Primes Are Expected Without End and Fermat Primes to Stop, the Growth of the Exponent Splitting the Sum ── correcting for not being divisible by small primes, the expected number of Mersenne primes grows by the same amount each time the exponent limit is multiplied by 10, while for Fermat numbers it is used up within the first ten or so ── the separator is whether the exponent grows one step at a time or by doubling ── [Paper 1018]

Whether infinitely many numbers 2 to the power p minus 1 (Mersenne numbers) are prime, and whether any number 2 to the power 2 to the power n plus 1 (a Fermat number) beyond F4 is prime, are both open. That there are infinitely many Mersenne primes is the same question as whether there are infinitely many even perfect...

Yuuki Yamagishi · 0 citations
#small language model Open access Sep 2026

Sovereign AI and Epistemic Dependence: Infrastructure, Language, and Who Answers for a Region's Knowledge Layer

Sovereign AI is measured in megawatts, chips, and capital. A region announces a gigawatt campus, secures an export licence for tens of thousands of accelerators, and is described as having achieved compute sovereignty. This article argues that the accounting is measuring the visible layers rather than the determinative...

Syed Shahzad · 0 citations
#small language model Open access Sep 2026

CloudSentinel: An AI-Assisted Cloud Misconfiguration Detection and Remediation Framework for Resource-Constrained Engineering Teams

Cloud misconfiguration remains a persistent security challenge for small software development teams operating without dedicated security expertise. This paper presents CloudSentinel, an AI-assisted cloud misconfiguration detection and remediation framework that improves the actionability of security findings while pres...

Binamra Bikram Pandey · 0 citations
#small language model Open access Sep 2026

Primitive Roots Break the Independent Estimate Only When the Square-Free Base Leaves 1 on Division by 4 ── for a = 5, being a quadratic non-residue and the remainder of p on division by 5 are tied together by reciprocity, and the share of primes with 5 as a primitive root becomes 1 + 1/19 times the independent product ── the separator is whether a leaves 1 or 3 on division by 4 ── [Paper 1020]

An integer a is a primitive root of a prime p when its powers produce every nonzero remainder on division by p. Whether infinitely many primes have a as a primitive root (Artin’s conjecture) is open for every a that is neither a square nor −1. This paper counts the share of primes for which a is a primitive root and li...

Yuuki Yamagishi · 0 citations

From tech blogs

See all →
Microsoft Research Blog Oct 12, 2026

Simulating student learning for better AI tutoring

StudentSim is a research system for evaluating and improving AI tutoring tools before they are tested with learners. It models how individuals may engage with tutors, giving feedback to improve tutor performance across subjects like chess, writing, and mathematics. The post Simulating student learning for better AI tutoring appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.