Skip to content
Log in
Sponsored Design Multi-Agent AI Systems Using MCP and A2A: Engineer your own Pyt Design Multi-Agent AI Systems Using MCP and A2A: Engineer your own Pyt

Category

small language model

2,906 papers

#edge computing Open access Sep 2026

Adding One Chord Destroys Only the Property That Has a Local Certificate ── Add one chord to a hexagon and 2 vertices acquire odd degree, so the Euler circuit vanishes ── yet the number of Hamiltonian cycles stays at 2 ── [Paper 508]

Tracing every edge and visiting every vertex are taken to be two forms of one problem. This paper shows that adding one and the same edge destroys only one of them──one has a certificate at each vertex and the other has none. No new theorem or law is claimed. Scope of this paper (scope note): No new theorem or law is c...

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

Voting, Validation, and Vulnerability in Local Language Model Penetration Testing An Empirical Study of Module Selection Under Adversarial Evidence

Small local language models are attractive for privacy-preserving penetration-testing assistants, but their module choices must remain reliable when target-controlled banners contain misleading or instruction-like text. We evaluate module selection in three experiments on an Apple M2 system with 8 GB of memory. A verif...

Piyush Shukla · 0 citations
#small language model Open access Sep 2026

allen-proxmire/Primes: Primes Repo V2.2

v1.2 — September 2026 The collection audited itself, and gained a way to be read. Seventeen commits. Two claims retracted, two verified, one small new proposition, and — probably the most useful part — a layer of plain language so the work can be understood without the notation. New: you can now read this without a mat...

allen-proxmire · 0 citations
#small language model Open access Sep 2026

CAMS Protocol R (CAMS-R-2026-001): preregistration, single-run results, and rubric erratum final

Protocol F (CAMS-F-2026-001): Design Rationale Companion to preregistration_F_draft.json — 2026-09-20 Status: DRAFT. The forecast file does not exist yet; no outcome-year scoring has begun. Protocols R and R2 answered does the relational pattern generalise? — twice, under lock and key. Protocol F asks the harder questi...

Kari McKern · 0 citations
#small language model Open access Sep 2026

Window Theory • AI — Paper I: Closure Dynamics

Window Theory - AI (wtAI): A Falsifiable Framework for Flexibility-Fragility Coupling in Large Language Models This is a six-part research series examining a structural pattern in decoder-based next-token-prediction systems deployed as dialogue and instruction agents. It is written for readers with a background in mach...

Wai-Hung (Pan) Tam · 0 citations
#small language model Dataset Open access Sep 2026

quant_eval v7.21: Per-Case Evaluation Results and Run Provenance for Mistral-Nemo-Instruct-2407

Complete per-case evaluation output and run provenance for the two quant_eval version 7.21 runs that underlie the whitepaper "quant_eval: A Behavioral Evaluation Harness for Full-Weight and Quantized Large Language Models." Every pass rate, bucket score and latency figure reported in that paper can be recomputed from t...

Patrick Hill · 0 citations
#small language model Open access Sep 2026

Voting, Validation, and Vulnerability in Local Language Model Penetration Testing An Empirical Study of Module Selection Under Adversarial Evidence

Small local language models are attractive for privacy-preserving penetration-testing assistants, but their module choices must remain reliable when target-controlled banners contain misleading or instruction-like text. We evaluate module selection in three experiments on an Apple M2 system with 8 GB of memory. A verif...

Piyush Shukla · 0 citations
#small language model Open access Sep 2026

Split the Surplus Three Ways and a Curved Problem Becomes a Flat One ── subtract a third of it from each angle and solve as a plane triangle, and the residual grows as the square of the excess, staying at 4.106x10^-4 of it even at a million km^2 ── [Paper 512]

A triangle on a curved surface is taken to require curved tools. This paper shows that splitting the surplus three ways lets flat tools suffice──the residual grows as the square and vanishes within practical range. No new theorem or law is claimed. Scope of this paper (scope note): No new theorem or law is claimed──the...

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

A Square Net and a Triangular Net Describe the World With the Same Exponent, and Only the Coefficient Differs ── counting lattice points inside a circle with square, triangular and rectangular lattices of the same density, the error grows as a power of the radius whose exponents 0.525037, 0.521921 and 0.542771 differ by only 0.02 while the coefficients 1.392341, 1.657107 and 1.119127 differ by 0.30; Tractatus 6.341, that the shape of the net is arbitrary and only its fineness says something about the world, is counted as exponent against coefficient ── the separator is whether what was changed is the shape of the net or its fineness ── [Paper 1033]

In Tractatus 6.341 Wittgenstein likened mechanics to a net laid over the world: the shape of the mesh (square or triangular) is arbitrary, and all that can be said of the world is that it can be described by a net of this fineness. This paper counts the simile with lattice points. Three lattices of the same density are...

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

The Time Difference for Light Going Once Round a Rotating Loop Is Set by the Enclosed Area Alone, Not by the Perimeter ── A square, an equilateral triangle, a rectangle with 2.125 times the perimeter and a regular 64-gon of the same area all give 4AΩ/c², and a figure of eight gives zero because its signed area is zero ── the separator is first order or second ── [Paper 1037]

Send light both ways round a rotating loop and the two beams return at different times. This paper solves, edge by edge, the motion of mirrors that move while the light is on its way, and counts what fixes that difference. No new theorem and no new law is claimed. Scope of this paper (scope note): No new theorem and no...

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

CAMS Protocol R (CAMS-R-2026-001): preregistration, single-run results, and rubric erratum final

Protocol F (CAMS-F-2026-001): Design Rationale Companion to preregistration_F_draft.json — 2026-09-20 Status: DRAFT. The forecast file does not exist yet; no outcome-year scoring has begun. Protocols R and R2 answered does the relational pattern generalise? — twice, under lock and key. Protocol F asks the harder questi...

Kari McKern · 0 citations
#small language model Open access Sep 2026

Failing to Have a Transition for Want of Size Is Not the Same as Failing for Want of Dimension ── The free energy of the one-dimensional Ising model shows no jump in its second derivative at finite size (a largest step of 0.282856 at n=64) ── taking the limit is necessary for a transition, not sufficient ── [Paper 502]

The absence of a phase transition is put down to the system being finite. This paper shows that a system can fail to have one even in the thermodynamic limit──the one-dimensional Ising model, whose largest step stays at 0.510568. No new theorem or law is claimed. Scope of this paper (scope note): No new theorem or law...

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.