AI Wiki — Knowledge Base

Massive Update to my Krea 2 Multi-Lora Bounding Box workflow, now bounding boxes control placement with better accuracy. Also introduced Edit features like Scene and Outfit transfer, put multiple character loras in a scene or outfit of your choosing! Token drift also fixed by facial detailer stage

Krea 2 has been my favorite base model for character work, but the moment you put two character LoRAs in the same generation they smear into one blended face. Attention bias, prompt engineering, and C

safetyr-stablediffusion

31 Jul 2026

Search: “routing”

Compliance2LoRA: On-Demand Safety Alignment on Arbitrary Policy Subsets via Hypernetwork-Generated LoRA Adapters

Contrastive Reinforced Policy Optimization via Privileged Self-Distillation

Distributions In, Distributions Out: The Case for Soft-Label Training

ENCORE: Event-Assisted Complementary Motion Refinement for Learned Video Compression

eta-OPSD: Deriving with Policy Optimization, Training with Self-Distillation

HALO: Heterogeneous Admission through Localized Obligations for Safe Agentic Execution

Massive Update to my Krea 2 Multi-Lora Bounding Box workflow, now bounding boxes control placement with better accuracy. Also introduced Edit features like Scene and Outfit transfer, put multiple character loras in a scene or outfit of your choosing! Token drift also fixed by facial detailer stage

MUGEN: A Unified Framework for Efficient Motion Understanding and Generation

Nanoparticle Networks for Neuromorphic Computing

PUDA: An AI-Native Hardware Harness for Self-Driving Laboratories

RiO-DETR: DETR for Real-time Oriented Object Detection

RoboBRIDGE: A Modular Framework for Bridging Policies to Robust Real-World Robotic Agents

Shared Symbolic Backbones for Physically Consistent Multi-Output Symbolic Regression

Stateless MCP has recaptured my interest (and inspired mcp-explorer and datasette-mcp)

We created a document OCR router that can estimate the complexity of every single page and parse it with the relevant mode 💫 * Some pages a…

What’s new in AI infrastructure and orchestration this month

A Compositional Theory of Causally Masked Transformers

Aligning LLM-Simulated and Human Examinees for Psychometric Calibration: A Cognitive Diagnostic Profiling Approach

Anatomy Contextualized Adaption of CT Foundation Models

ASCD: Attention-Steerable Contrastive Decoding for Reducing Hallucination in MLLM

CineWeaver: Training-Free Reference-Controllable Multi-Shot Long Video Generation for Cinematic Storytelling

DataBahn raises $40M as AI agents queue up for enterprise telemetry

Decoupled Visual Processing: Efficient Multimodal Adaptation via Modality-Specific Transformer Substitution

DP-MGTD: Privacy-Preserving Machine-Generated Text Detection via Adaptive Differentially Private Entity Sanitization

DuplexGen: Adaptive Synthesis of Human-AI Turn-Taking Dialogues

From Uncertainty to Determinism: Coarse-to-Fine Visual Floorplan Localization without Ray Matching

HeteroPROPMT: A Real-time and Privacy-Preserving Heterogeneous Collaborative Perception Framework

Latent-IM: Latent Interaction Management for Speech LLMs

LISA-3D: Lifting Language-Image Segmentation to 3D via Multi-View Consistency

SPROUT: A Scalable Diffusion Foundation Model for Agricultural Vision

Structurally Separated Uncertainty in Supervised Latent Variable Models

SymmGrid: Super-Scaling On-Robot Learning with Parallelized Symmetries and Egocentric-Exocentric Visual Perception

Temporally Centered SIGReg Improves Multi-Task LeWorldModel Learning: From Analysis to Method

When Does Span-Guided Detoxification Help? Human Preferences and Evaluator Diagnostics in a Controlled Comparison

A Study of Crosslinguistic Influence in Language Models

Automate your agent development lifecycle using any coding agent

CogEEGAgent: Toward Autonomous Cognitive EEG Analysis with Grounded Execution and Selection-Aware Verification

Configuring Dedicated Model Inference

Context Assembly as the Controlled Variable: A Control-Theoretic View of Harness Policies for Frozen LLM Agents

ContractHIL-HLS: Contract-Aligned Multi-Agent Workflow with Hardware-in-the-Loop Feedback for HLS Design

FinAbstain: Uncertainty-Calibrated Multimodal RAG for Selective Financial Forecasting

Leak-Free Cross-Validated Stacking with Per-Architecture Calibration for Sand-Boil Segmentation in Earthen Levees

Less Data, Better Alignment: Data-Centric Multi-Evaluator Agreement for Preference Optimization

Memory for Large Language Models

MusiChat: Vibe Composing for Music Creation

Optimal Ansatz-free Hamiltonian Learning In Situ

Physics-Informed CNN-LSTM for Street-Scale Urban Flood Prediction: Reconciling Aggregate Accuracy and Street-Level Plausibility

Reasoning with Memory: A Temporal Granularity-Adaptive Framework for Training-Free Long Video Understanding