AI Wiki — Knowledge Base

I turned Qwen3.8-27B Q2_64 + llama.cpp into a fully TypeSafe AI-compatible Jev-like system. OpenAI API still intact! World’s first Vision-enabled Jev-like model! <10 GB VRAM, 170 ms on an RTX 3090 and ~140 tok/s in chat. 76% vs. 88% Jev-1.13 Acc. on a diverse 22,000-request typed-decision benchmark

Yesterday I released Bonsai-Llama-Jev - a Jev-like Typed Decision inference system for locally operated, sovereign AI. In my new typed-decision-bench, covering more than 22,000 decisions across 175 us

model-releasesr-localllama

23 Sept 2026

model releases

EDGEGEN: Improving Tool-Calling Agents Beyond Happy Paths with Synthetic Edge Case Generation

Efficient Cost-Aware LLM Evaluation via Bayesian Bandit Gittins Indices

Efficient Iterative Retrieval with Heterogeneous Batching

EMGBlend: Heterogeneity-Aware Self-Supervised Pretraining for Gesture and Force Decoding

Engram gone wild! 2b model update...

ENPIRE: Agentic Robot Policy Self-Improvement in the Real World

Enriching Speech Emotion Representations with Conversational Context

EquivSVA: A Formally Verified Dataset of Behavioral Assertions Across Equivalent RTL Implementations

Error Bounds for Statistical Estimators in BTL Model with Parametric Multivariate Utility Functions

Et Tu, Brute? Economic Misalignment in Personal AI Agents

Evaluating Accuracy and Probabilistic Reliability of Zero-Shot Time Series Foundation Models

Evaluating the Effectiveness of SechKAN on 1D Data

Evidence-gated multimodal parsing and vectorization of architectural floor plans

Exciting work from NVIDIA. (bookmark it) Interesting to see this approach to turn public Agent Skills into RL environments. Lots of exciteme…

eXplaining to Learn (eX2L): Regularization Using Contrastive Visual Explanation Pairs for Distribution Shifts

Exposing Blind Spots in Deep Imbalanced Regression Evaluation

FAST-ML: A Hybrid Physics-Machine Learning Framework for Tropical Cyclone Intensity Forecasting

Federating Quantum and Classical Computing: A Privacy-Preserving Hybrid Approach

FineWeb-CLaR: Culture, Language, and Region Annotations for Benchmark-Aligned Corpus Auditing

FinFIRST: Benchmarking Search Agents for Financial Information Retrieval, Sourcing and Traceability

FinInteract: Benchmarking Clarification and Intent Integration in Ambiguous Financial Question Answering

FIRE: Failure-Informed Runtime Engineering for Reliable Language-Model Agents

FireWorldBench: Benchmarking Complex Physical World Intelligence through Coupled-Field Fire Dynamics

For the last several years, I have said that OpenAI models are better at ideation, creation, and exploration. If I had to brainstorm somethi…

From Capability to Assurance in Autonomous Penetration-Testing Harnesses: A Framework and Reference Implementation

From Code to Requirements: Agentic Reverse Engineering of Business Rules at Enterprise Scale

From Experts to Sub-experts: Fine-grained Parameter-Efficient Fine-Tuning for MoE LLMs

From IceCube to IT-Sphere: A Hybrid Quantum-Classical GNN for Banking IT Root Cause Analysis

From Inference Engine to Inference Control Plane: Connecting vLLM, llm-d, and the Evolution of Efficient Distributed LLM Serving

From Instrument-Mounted Demonstrations to In-Vivo Execution: Learning Bimanual Laparoscopic Appendectomy Without Robot-Collected Demonstrations

From Utterances to Networks: Modelling Slang Adoption and Diffusion Across Subreddits

FrontierMath Erdos

FuncCode: Compressing Kolmogorov--Arnold Networks in Function Space with Hardware-Aware Quantization

FusionMMT: A Unified Multimodal and Multitask Learning Framework for Nuclear Fusion

Gemini 3.8 TTS Playground

GGUFs in transformers natively!

glance-vlm speedlab is now open source! read: https://glance.yohei.me/speed/ try: https://github.com/yoheinakajima/glance-speedlab turn any …

Glucose-ML: A collection of longitudinal diabetes datasets for development of robust AI solutions

Google DeepMind SVP Koray Kavukcuoglu says Gemini 4 is in the early phase of post-training and hopes it will be released 'much earlier' than the end of the year (Laura Bratton/The Information)

Graph Memory for LLM Agents: At What Cost? A Comparative Evaluation of Query, Ingest, and Update Performance Across Graph Database Engines

Graphic producing new model Ming works fine in Comfy

GTR: Gated Token Recurrence for Efficient Dense Prediction

GUI harness [video]

gwBenchmarks: Stress-Testing LLM Agents on High-Precision Gravitational Wave Astronomy

Hierarchical Sparse Bayesian Multitask Learning for Disease Prediction in Pooled Microbiome Studies

Hill Sampling for Test-Time Scaling: A Simple and Better Alternative to Repeated Sampling, Evolution, and Training

If you've noticed how fast https://claude.ai and the Desktop app have become in the last few weeks, here's how we did it. Lots of juicy lear…