AI Wiki — Knowledge Base

R9V Update: Created and adopted KVA projections based on Deepseek V4.1 Flash + HySparse2/MiMo-V3 for Qwen3.8 Flash Next. This is a game changer for models that don't natively implement it. 1.45-1.85x speedup in prefill to 3k+ at a small deficit to perplexity. [2x R9700, 128GB DDR5]

Here's my *first* implementation of KVA projectors on QFN (just the uncensored model for now) the highlights are basically as follows for using the projectors at each different layer: Starting at laye

model-releasesr-localllama

24 Sept 2026

Search: “engineering”

PSA for M5Ultra owners running LLMs: set your prefill step to 8192

Reliable, fast inference is essential. As Ted shared at Apsara Conference 2026, Together makes it easy for teams to get access to frontier o…

Rufus-Air: An Open LLM Post-Training Recipe

SplatLabel: Pseudo-Labelling through 4D Gaussian Splatting

Stable and Faithful Explanations for Knowledge Tracing

The Entropy Triangle Method (ETM): A novel framework for the prevention of cardiac arrhythmia with a review of more than 10,000 patients

We announced our RSI Lab earlier this year: https://sakana.ai/rsi-lab/ Over the last two years, we have systematically shipped the foundatio…

When Agents Act Unwatched: The Reduced-Supervision Paradox in Agentic AI

Agent-Editing World Model: Rethinking World Modeling for LLM Agents

Bounded Loops: Pre-Run Spend Bounds, Proved Termination, and Verified Completion for Agent Harnesses

Controlling Collectives of AI Agents in Reasoning Space with Spatial Transformers

COPE: Continual Personalization of LLMs under Sparse User Feedback via User Embeddings and Self-Evaluation

Databricks acquires spreadsheet startup Row Zero to enhance its AI capabilities

Enhancing Multiclass Malware Classification in Resource-Constrained Environments

Evaluation of pre-trained models for pedagogical assessment of novel AI-assisted educational questions

FDE-Bench: Evaluating LLM Agents for Deployment Environment Configuration

Harness as a Language: A Minimalist Agent Framework With Maximal Expressivity

Intelligence Across Embodiments

JEV almost dead: CLM vs JEV

Learning to Detect Symbolic Failure: Machine Learning and the Limits of Black-Scholes

Live Assistant: Learning Whether, When, and Whom to Assist in Real-World Live Social Streams

Power your agents: Gemini 3.8 Live with Live Avatar is now generally available

R9V Update: Created and adopted KVA projections based on Deepseek V4.1 Flash + HySparse2/MiMo-V3 for Qwen3.8 Flash Next. This is a game changer for models that don't natively implement it. 1.45-1.85x speedup in prefill to 3k+ at a small deficit to perplexity. [2x R9700, 128GB DDR5]

Safe Multi-Robot Coordination via VLM-LLM Reasoning and Reachability Analysis

The Recall Ceiling of LLM Recommendation Reranking

A Compact Stance-Indexed Anterior-Posterior COP Representation for Parkinson's Disease Classification from Plantar VGRF

A digital-twin framework for forecasting treatment-day imaging with contour uncertainty in adaptive proton radiotherapy

A Hybrid AI Framework for Academic Advising: Integrating Ensemble-Based Grade Prediction and a Rule-Based Expert System

An Implant-to-Wearable IR-UWB Transmitter-Receiver Architecture and Layered Protocol for High-Density Brain-Computer Interfaces

Bridging the Data Gap: Digital Twin as a New Paradigm for AI-based Radio Sensing

CoEvo: Oracle-Grounded Self-Evolution of a Single Model for Multi-Step Causal Reasoning

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…

FleXray: Universal Clinical X-ray Segmentation

GameDirector: Decoupling Gameplay Logic from Rendering for Player-Configurable Game World Models

Generalizing Manipulation Skills with a Local Coding Agent

GeoPair: Geometry-Preserving Cross-Layer Factorization for Training-Free Transformer Compression

Graphic producing new model Ming works fine in Comfy

GUI harness [video]

Heating in human-HVAC interaction for smart homes: An interdisciplinary overview

ICDAR2026 Competition on Multimodal Reasoning over Documents in Multiple Domains

KwaiMind Technical Report

LEAP-NBV: Lightweight Edge Active-Perception for Foundation-Model Next-Best-View Planning

Learning Diagnostic Reasoning for Decision Support in Toxicology

Manipulation of Deformable Linear Objects Using Model Predictive Path Integral Control with Bidirectional Long Short-Term Memory Learning

NPU Accelerator: Quantized Real-Time Vehicle Detection on PYNQ-Z1 Using FINN

OMatG-flash: An All-Atom Flow Map with Reinforce Adjoint Matching for Scalable Materials Discovery

SPECTRA: Adaptive Execution of Speculative Decoding on a Runtime-Reconfigurable Tiled Architecture