We describe TinySQL-125M, a decoder-only language model for text-to-SQL in which every component of the model path is written from scratch: a byte-level BPE tokenizer, the transformer itself (grouped-query attention, rotary embeddings, SwiGLU, RMSNorm), the optimizer, the training loop, the data pipeline, a grammar-con...
Aditya Narayan Jha· Zenodo (CERN European Organi...· 0 citations
Artificial intelligence (AI) is increasingly used in healthcare, where evaluations usually ask whether a system is accurate on ordinary clinical data. This review asks instead what can happen when someone deliberately changes the data, instructions, or training to make a system fail, behave in hidden ways, or reveal co...
Patryk Sielaff, Artur Brudek, Adrian Nowakowski et al.· International Journal of Inn...· 0 citations
The evaluation of language-model agents has developed mainly around programming tasks, where correctness can be checked mechanically. Analyses in the humanities, social sciences and economics have no such criterion: correct values can support conclusions the data do not warrant. This article evaluates the harness — the...
Vagner de Carvalho Bessa· Zenodo (CERN European Organi...· 0 citations
GenoSQL-AI is an open-source framework that integrates PostgreSQL hybrid data models with AI-powered natural language querying to enable scalable and cost-effective genomic data analysis.
Joseph Edet, Uchenna Splendor Ejike· Zenodo (CERN European Organi...· 0 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
Weight-only post-training quantization (PTQ) pipelines such as AWQ ultimately commit to a deterministic rounding decision for every scalar weight, and recent discrete coordinate-descent quantizers refine those decisions one weight at a time. We show that both are blocked by the same obstacle: correlated activation chan...
Anh Nguyen Duc, Linh Ngo Van, Sang Dinh Viet et al.· 0 citations
This dataset accompanies the paper "Analysis of Epistemic Isolation Components in Kubernetes Manifest Generation by Small Language Models" by Andrey V. Kozachok, Anatoliy A. Bakaev, Aleksandr V. Kozachok, Shamil G. Magomedov and Artem N. Noev. It contains the data split, code and evaluation results of an ablation study...
Andrey V. Kozachok, Anatoly A. Bakaev, Alexander Vasilievich Kozachok et al.· Zenodo (CERN European Organi...· 0 citations
Enterprise AI procurement often combines a choice of application bundle with a choice of language model. This exploratory study examined three model selections available through enterprise interfaces: Opus 5.5 and GPT-6.1 Sol in the installed Microsoft 365 Copilot desktop application, and Gemini 4 Argon in the Gemini E...
Some sign language avatars play a recorded sign for each word in their dictionary and fingerspell the other meaningful words, such as nouns, verbs and names, spelling them letter by letter with the hand alphabet. Input in another language must be translated before it is signed, and the translation can be fitted to the...
Nikolaos D. Tantaroudas, Andrew J. McCracken, Ilias Karachalios et al.· Preprints.org· 0 citations
We analysed how key aspects of refugees’ reception and civic integration experiences relate to their probability of exiting social assistance for employment in the Netherlands. We focused on the duration of stay in reception centres, relocations between centres, completion of the civic integration program, and the lang...
Барсегян Вардан Меружанович, Samir Mustafa Negash· 0 citations
External context can specify how a language model should respond, but a response that satisfies operational criteria need not accurately describe the process that produced it. We examine two small evaluations of one operational deployment. In EXP-001, scaffolded outputs scored lower on the accuracy of process self-repo...
Ondřej Šrámek· Zenodo (CERN European Organi...· 0 citations
Background Domain-specific medical question-answering (QA) models face bottlenecks due to scarce labeled data, specialized terminology, and constrained computational hardware. Objectives The study aims to develop an efficient pharmaceutical question-answering (QA) system that can be trained and deployed on consumer-gra...
Shenbagarajan Anantharajan, Saranraj Loganathan, Rajkumaran Soundarraj et al.· Frontiers in Artificial Inte...· 0 citations
Search-optimisation advice for AI answers lists dozens of ranking factors, most of them taken from vendor studies that differ in data, platform and method. This preprint compares 38 of them with one snapshot of Russian-language results. The main collection covers 2400 queries in 16 niches on Google AI Overviews (88230...
Stanislav Kirichenko· Zenodo (CERN European Organi...· 0 citations
Assistant Professor Benjamin Lindquist studies how engineers deliberately injected randomness into early AI, giving rise to deep-learning models that sometimes hallucinate.
Flint-interactive gives AI agents a language for creating interactive visualizations. Click a chart, filter data, highlight related content, and explore beyond the view the author originally chose. The post Flint-interactive: A language for creating interactive data visualizations appeared first on Microsoft Research.
