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: “models”

A Sparse Glimpse of the Whole: Train-Free Self-Speculative Decoding

AHA-Memes: A Fine-Grained Multimodal Benchmark for Understanding Hate in Arabic Memes

APO: Unsupervised Atomic Policy Optimization for 3D Structure Prediction of Atomic Systems

b10208

Beyond a Single Judge: Simulating Social Persona Panels for Generative UI Evaluation

Beyond Similarity: Grounded Agentic Extraction and Expert-Adjudicated Evaluation of Intertextuality in Classical Chinese Histories

Bunraku: Turning a Single Illustration into an Editable Live2D Character

ClawTrack: Towards Trace-Level Evaluation and Improvement of Real-World Autonomous Agents

CLIP-Guided Backdoor Defense through Entropy-Based Poisoned Dataset Separation

Compact Task-Aligned Imitation Learning for Laboratory Automation

Could we all crowdsource a dataset/model/finetune?

Cross-Embodiment Transfer via Behavior-Aligned Representations

deepseek-ai/DeepSeek-V4-Flash-0731

DinoLizer: Separating VAE and Diffusion Artifacts in Generative Inpainting Localization

Error Analysis of Neural-Network-Based Engression

Expected Survival-Time Bounds for Robust Optimization Over Time under Isotropic Gaussian Dynamics

Huawei opensouced openPangu-2.0-Pro, 505B-A18B

K-EXAONE 2.0 released

Learning Color Grading, No Photo Sharing: Federated Aesthetic Preference Learning for Personalized Image Enhancement

LLMs struggle to simulate human belief updates in controlled environments

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

MatCreatioNN: Machine learning-guided computational discovery of photocatalysts for environmental applications

MemSecBench: Tracking Agent Memory Poisoning from Persistence to Consequence and Repair

MiniMax H3 discussion

ObjectStream: Latent Objects as Memory Anchors for Streaming Video Understanding

Oxide and Friends: The Open Weight Revolution with Simon Willison

PlantBGC: Transformer for Plant BGC Discovery via Label-Free Domain Adaptation and Weak Supervision

PoseMaster: A Unified 3D Native Framework for Stylized Pose Generation

Psych-ECA: A Reproducible Semi-Synthetic Benchmark for Synthetic Control Arms in Longitudinal Psychiatry

Reasoning Consensus: Structural Ensembling of LLM Reasoning via Weighted DAG Aggregation

RefineSVG: Visual Feedback-Driven Reinforcement Learning for Image-to-SVG Generation

ROAD: Reciprocal-Objective Alignment of Discriminative Semantics for 3D Shape Generation

Safety-Gated Agentic Supervisory Control on a Coupled Distillation Benchmark: Regime Map, Auditable Gate, and Co-Design Findings

SE(3)-MeanFlow: Few-Step Protein Backbone Generation on Lie Groups

ServerlessT2I: Efficient Text-to-Image Workflow Serving on a Serverless Platform

Simplifying Neural Networks During Training

SkillSight: Calibrating Generic Content Bias for Skill Retrieval

TAPO: Transition-Aware Policy Optimization for LLM Agents

The Easy Trap: Why LLMs Underestimate Misconception-Driven Difficulty

Towards Generalized Synapse Detection Across Invertebrate Species

What Does It Take to Detect an AI Agent? Minimal Feature Sets for Behavioral Detection under Browser Automation

4090 + 5060 Ti + 64GB RAM: 206 t/s on a 35B-A3B, and a 122B at 37 t/s

AgentSnare: Learning to Delay, Divert, and Defuse Autonomous Penetration Agents

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

Dynamic Parameterization Is Not Dynamic Inference

exttt{AMEND++}: Benchmarking Eligibility Criteria Amendments in Clinical Trials

Gated Adaptation for Continual Learning in Human Activity Recognition

How close are we to local llama robotics for consumer price point?