small language model

How Many Frequencies Can Be Nulled Is Fixed by the Number of Coefficients ── five coefficients null two frequencies, a third raises the count to 7, and truncating back to five leaves none of the three at zero ── [Paper 520]

How many demands of the form “null this frequency” can be met? What this paper shows is that the ceiling is set by the number of coefficients, and that falling short does not cost one demand but all of them. No new theorem or law is claimed. Scope of this paper (scope note): No new theorem or law is claimed──the fundam...

Historical Detective Fiction: An Annotated Bibliographic Dataset of 301 Novels and Series (first publications 1594–2023)

About This is a single-author, source-verified bibliographic dataset of 301 works of historical detective fiction – single novels and whole series, first published between 1594 and 2023, in 36 original languages, in 46 countries. Each of the 301 entries is a bibliographic record carrying an annotation, describing why t...

precis-mcp: a Model Context Protocol server for papers, documents, and agent state

precis-mcp is a Model Context Protocol (MCP) server that gives language-model agents a small, uniform API for reading, writing, and searching across scientific papers, documents, personal state, code, and cached tool calls. Small-model-friendly (7B-class agents are the design target), it exposes a handful of typed verb...

Window Theory • AI — Appendix A: Metaphor Material and Case Verification

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...

Window Theory • AI — Paper 0: Prologue

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...

The Energy Fluctuation of Light Is a Particle Term Plus a Wave Term, and the Two Are Equal Exactly When the Occupation Number Is 1 ── in the energy fluctuation of a single mode, the particle term divided by the wave term equals the reciprocal of the occupation number and becomes 1 at hν = kT ln 2; at the Planck peak the particle term is 15.8 times the wave term, and for visible sunlight 91.5 times ── the separator is whether the occupation number is above or below 1 ── [Paper 1022]

In 1909 Einstein showed that the energy fluctuation of a single mode of black-body radiation is the sum of a term like that of light quanta and a term like that of classical waves. This paper checks by three routes, the formula, the distribution of the occupation number and thermodynamics, that the ratio of the two ter...

The Energy Fluctuation of Light Is a Particle Term Plus a Wave Term, and the Two Are Equal Exactly When the Occupation Number Is 1 ── in the energy fluctuation of a single mode, the particle term divided by the wave term equals the reciprocal of the occupation number and becomes 1 at hν = kT ln 2; at the Planck peak the particle term is 15.8 times the wave term, and for visible sunlight 91.5 times ── the separator is whether the occupation number is above or below 1 ── [Paper 1022]

In 1909 Einstein showed that the energy fluctuation of a single mode of black-body radiation is the sum of a term like that of light quanta and a term like that of classical waves. This paper checks by three routes, the formula, the distribution of the occupation number and thermodynamics, that the ratio of the two ter...

The Extrapolation Crosses Zero Because the Heat-Capacity Difference Was Held Constant ── With a melting entropy of 10 and a heat-capacity difference of 30, the extrapolated entropy difference crosses zero at T=214.959393, or 0.716531 of the melting point ── let it vanish as the third law requires and it need not cross at all ── [Paper 503]

The Kauzmann paradox is spoken of as a property of supercooled liquids. This paper shows that crossing zero follows from holding the heat-capacity difference constant──held constant, however small, it always crosses. No new theorem or law is claimed. Scope of this paper (scope note): No new theorem or law is claimed──t...

The Spatial Curvature of a Rotating Disk Is −3Ω²/c², and Its Argument Differs from the −1/c² of Velocity Space ── the Gaussian curvature of the spatial metric of a rotating frame is −3Ω²/c² at the centre and vanishes when Ω is set to 0; for a disk whose rim moves at half the speed of light the circumference over the diameter is πγ = 3.627599 and the area is 1.071797 times the Euclidean one, growing less than the circumference; the curvature −1/c² of velocity space contains no Ω and has different units ── the separator is whether what bends is rotation or the composition of velocities ── [Paper 1027]

The circumference of a rotating disk is longer than π times its diameter (Ehrenfest 1909). Einstein used this disk as a clue that geometry under gravity would not be Euclidean. This paper obtains the Gaussian curvature of the spatial metric of a rotating frame by two routes, symbolic computation and numerical different...

The Extrapolation Crosses Zero Because the Heat-Capacity Difference Was Held Constant ── With a melting entropy of 10 and a heat-capacity difference of 30, the extrapolated entropy difference crosses zero at T=214.959393, or 0.716531 of the melting point ── let it vanish as the third law requires and it need not cross at all ── [Paper 503]

The Kauzmann paradox is spoken of as a property of supercooled liquids. This paper shows that crossing zero follows from holding the heat-capacity difference constant──held constant, however small, it always crosses. No new theorem or law is claimed. Scope of this paper (scope note): No new theorem or law is claimed──t...

Small Models, Big Claims: Choosing and Quantizing Local LLMs on an Integrated GPU

Running large language models locally on integrated graphics is now practical, but the community's model-selection advice is dominated by vibes: "just grab a 2B at Q4." This report presents a controlled, single-machine measurement study of twelve model/quantization configurations across three model families (Qwen2.5-1....

Light-MFND: Efficient Multimodal Fake News Detection via Fine-Tuned Small Vision-Language Model