small language model

OSIRIS v4.2.1: a governed local console for a living language model

OSIRIS is a local-first console built around a small language model (osiris.nclm) that learns from every conversation and from the author's documents, and answers in its own voice only after passing a held-out test (30 held-out exchanges at ≤ 2.0 bits/byte and below a unigram baseline). Until then a local Ollama model...

Generative AI-Mediated Feedback and L2 Grammatical-Lexical Development: A Scoping Review of Longitudinal Research Gaps

The rapid adoption of generative artificial intelligence (AI) tools such as ChatGPT in second language (L2) writing instruction has produced a fast-growing body of research on AI-generated written corrective feedback. Two recent scoping reviews have mapped this literature broadly, focusing on feedback quality, learner...

Editorial: Artificial intelligence and anti-cancer drugs: advancing discovery, personalization, and precision oncology

It is fitting that the Topic opens with a broad review of the field itself. Nagrami et al. map AI's footprint across the entire anti-cancer drug pipeline -target identification, virtual screening, molecular docking, multi-omic integration, and the still-unresolved question of clinical validation. Their review is a usef...

Tron-1B: Fast, Calibrated Typed Decisions with a Set-Attention Option Head

Applications built around large language models make many small decisions: which queue a ticket belongs to, whether a message is a prompt-injection attempt, how urgent a request is, whether an agent's step succeeded. Routing each of these through a generative LLM is slow, costly and hard to calibrate. We present Tron-1...

A Driven, Finite‑Memory Dissipative System: The Empirical Boundary of Non‑Equilibrium Dynamics and Criticality in a Quantized Hybrid Language Model

Where the Answer Is Born: Causal Interventions on the Information Trajectory of a Quantized Hybrid Language Model

Few-Shot Versus Zero-Shot Detection of Novel Objects Using Pretrained Vision-Language Models: A Comparative Study on Kikoriki Characters

Pukara

A privacy-first, self-hosted gateway for local large language models (Ollama). Pukara runs a Dockerized Ollama server behind an application-layer encrypted proxy, and ships a desktop client that pseudonymises sensitive entities on-device (BERT + regex) before anything leaves the machine. The name Pukara (Quechua for "f...

Spot and Explain Public Trust: Large Multimodal Model-Based Forensic Reasoning for Synthetic Image Detection

Clinically Structured Retrieval: Can a Well-Designed RAG Layer Add Value Over a Base Language Model in Clinical Oncology?

Study 1 (currently under review) found that retrieval-augmented generation (RAG) and fine-tuning both *degraded* clinical answer quality relative to an un-augmented base model in oncology. That study bounded its own retrieval conclusion explicitly: its index used 500-character chunks with 100-character overlap, small r...

Filtering LLM-Synthesized Data for Cold-Start Recommendation: A Progressive Influence Function Approach

Leveraging Large Language Models (LLMs)-synthesized data to enable representation learning for cold-start entities has emerged as a promising solution for addressing cold-start recommendations. However, existing studies often over-rely on LLM-generated data while neglecting the detrimental noise it may introduce. This...

Master Blob v3.1: A Reconstructed Modular System for Synthetic Causal Learning and Experiential Memory

An internally-run experimental report on a reconstructed, modular runtime for synthetic causal learning and experiential memory. The system combines a law-blind causal learner, a reduced graph-based physical field, bounded episodic memory branches, graph-based fusion, and independent-judge interfaces. It is evaluated w...