Skip to content
Log in
Sponsored AI Networking Cookbook: Practical recipes for AI-assisted network automation and development AI Networking Cookbook: Practical recipes for AI-assisted network automation and development

Category

small language model

2,883 papers

#small language model Book Open access Aug 2026

Invariant Pretraining for Robust Code Representations

InvPT applies semantic-preserving transformations to the corpus and combines masked language modeling with multi-positive supervised contrastive learning that treats all augmentations of the same source function as positives, mixing self-contrast pairs with invariant-contrast pairs for positives of varying difficulty.

Yifeng He, Yun-Di Xu, Christopher Castro Gaw Gonzalo et al. · 0 citations
#small language model Preprint Sep 2026

Path2Spec: Path-Aware Specification Generation via Large Language Models

This work introduces Path2Spec, a divide-and-conquer framework that leverages LLMs to extract all execution paths from an input program, generates path-specific specifications for each, and merges them into a comprehensive overall specification.

Dan Huang, Zhensu Sun, Hui-Hui Huang et al. · 0 citations
#small language model Preprint Sep 2026

Position Aware Layer Queries for Test Time Training in Vision Language Models

Test-Time Training (TTT) adapts models to incoming test samples (e.g. out-of-distribution, (OOD)) when conventional fine-tuning is infeasible. Existing TTT methods for Vision-Language Models (VLMs) create supervision from several augmented views, each requiring forward (and often backward) passes through the entire VLM...

Rajat Modi, P. Pathak, Xin Liang et al. · 0 citations
#small language model Preprint Sep 2026

SCOPD: Sparse-Context On-Policy Self-Distillation for Efficient Vision-Language Models

Reasoning vision-language models (VLMs) process images and videos as long sequences of visual tokens, making inference expensive. Training-free token pruning reduces this cost, but aggressive compression can sharply degrade performance, often attributed to irreversible loss of task-relevant visual information. We show...

Ahmadreza Jeddi, En-Ming Zhang, Jasper Gerigk et al. · 0 citations
#small language model Preprint Sep 2026

Towards Interpretable Framework for Neural Audio Codecs via Sparse Autoencoders: Exploration toward Age, Gender, and Accent Steering

Neural audio codecs (NACs) are widely used in speech generation and audio-language modeling, yet how they encode speaker-trait information remains poorly understood. Prior work applied sparse autoencoders (SAEs) to investigate accent information in NACs through task-level analysis. Here, we extend this analysis to the...

Shi Wang, Tian-Tian Feng, Aditya Kommineni et al. · 0 citations
#small language model Review Open access Sep 2026

Sistem Informasi E_Commerce UMKM Bolu Pisang Ine Pare Hanbai Kanbai Berbasis Web

An e-commerce–based sales information system that helps improve the efficiency of transaction management, enhances the quality of customer service, and simplifies the process of ordering and making online payments is designed and developed.

Clarita Ilona Seran, K. J. Tute, B. Bhae · 0 citations
#small language model Open access Sep 2026

Stretching the Stopping Distance 300 Times Cuts Peak Acceleration to 1/300 but HIC to 1/20784.609691 ── acceleration is inversely proportional to distance, but HIC15, driven by acceleration to the 2.5 power and a 15 ms window, falls to 1/20784.609691 ── the separator is whether danger is measured by peak acceleration or by a time-weighted criterion, and the shape of the force ── [Paper 790]

The shock of a collision gets smaller the longer the distance over which one stops. That is why cars are built to crumple on purpose. Varying the stopping distance and the shape of the force, how much the peak acceleration and HIC15, a criterion used for head injury, fall is counted. No new theorem or law is claimed. S...

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

What Sets the Life Is the Initial Flaw, Not the Final Size ── With an initial flaw of 0.5 mm and a final size of 25 mm, doubling the final size extends the life only to 1.048244, while halving the initial flaw extends it to 1.482441 ── and an infinite final size buys only 1.164716, approached slowly at that ── [Paper 565]

A tougher material is expected to last longer once a crack is present. What this paper shows is that most of the life is spent while the crack is small, so that enlarging the final size extends the life almost not at all. No new theorem or law is claimed. Scope of this paper (scope note): No new theorem or law is claim...

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

An Aperture and Its Complement Make the Same Pattern Only Away From the Forward Direction ── the intensities of a slit and of its complement converge away from the forward direction, while in the forward direction alone they do not agree, the ratio being 18.9055 ── [Paper 546]

Babinet's principle says that an aperture and its complement produce the same diffraction pattern. What this paper shows is that this holds only away from the forward direction, and that why it holds and why it fails follow from one and the same identity. No new theorem or law is claimed. Scope of this paper (scope not...

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

One and the Same "Proportional To" Carries Four Different Kinds of Constant ── a sphere's S/V^2/3 is fixed by geometry, a pendulum's Tsqrt(g/L) moves with amplitude, a power law fitted to the primes moves with the window, and the Catalan coefficient reaches only 0.99988751 of its limit even at n=10000 ── [Paper 530]

When one writes "A is proportional to B", the standing of the constant in front is rarely asked about. What this paper shows is that such constants come in four kinds, some of which move when something is changed and some of which do not. No new theorem or law is claimed. Scope of this paper (scope note): No new theore...

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

Condensation Happens Without Any Attraction: What Decides Is the Wavelength Overtaking the Spacing ── the condensation temperature of an ideal Bose gas contains only constants, the mass and the number density, with no interparticle potential at all ── the separator is whether switching off the interaction makes it disappear ── [Paper 663]

It is natural to think that a phase changes because particles attract one another. But the condensation of bosons happens without any attraction at all. The condensation temperature of an ideal Bose gas is computed in three cases, and what fixes that temperature is examined. No new theorem or law is claimed. Scope of t...

Yuuki Yamagishi · 0 citations

From tech blogs

See all →

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