AI Wiki — Knowledge Base

「AIは訓練するのではなく、自ら”育つ”ようにすべきだ」 Sakana AIのリサーチャー、@sebastianrisi がポッドキャスト @EyeOn_AI に出演。進化的手法でニューラルネットワークを構築するNeuroevolution手法の概要、継続学習や人工生命（ALif…

Sakana AI researcher Sebastian Risi appeared on the *Eye on AI* podcast to discuss neuroevolution — the approach of using evolutionary algorithms instead of gradient descent or reinforcement learn...

researchdavid-ha--x

10 Apr 2026

research

Adaptive Prompt Structure Factorization: A Framework for Self-Discovering and Optimizing Compositional Prompt Programs

Adaptive Symmetrization of the KL Divergence

AdaSpark: Adaptive Sparsity for Efficient Long-Video Understanding

Adversarial Evasion Attacks on Computer Vision using SHAP Values

AFL: A Single-Round Analytic Approach for Federated Learning with Pre-trained Models

AfriVoices-KE: A Multilingual Speech Dataset for Kenyan Languages

AI-Driven Marine Robotics: Emerging Trends in Underwater Perception and Ecosystem Monitoring

Algebraic Diversity: Group-Theoretic Spectral Estimation from Single Observations

An Empirical Analysis of Static Analysis Methods for Detection and Mitigation of Code Library Hallucinations

Approximate Replicability in Learning

AsyncTLS: Efficient Generative LLM Inference with Asynchronous Two-level Sparse Attention

AtomEval: Atomic Evaluation of Adversarial Claims in Fact Verification

AugLift: Depth-Aware Input Reparameterization Improves Domain Generalization in 2D-to-3D Pose Lifting

BadImplant: Injection-based Multi-Targeted Graph Backdoor Attack

Bag of Bags: Adaptive Visual Vocabularies for Genizah Join Image Retrieval

BDI-Kit Demo: A Toolkit for Programmable and Conversational Data Harmonization

Behavior-Aware Item Modeling via Dynamic Procedural Solution Representations for Knowledge Tracing

Between Century and Poet: Graph-Based Lexical Semantic Change in Persian Poetry

Beyond Pedestrians: Caption-Guided CLIP Framework for High-Difficulty Video-based Person Re-Identification

Beyond the Mean: Modelling Annotation Distributions in Continuous Affect Prediction

Bi-level Heterogeneous Learning for Time Series Foundation Models: A Federated Learning Approach

Bi-Level Optimization for Single Domain Generalization

Bi-Lipschitz Autoencoder With Injectivity Guarantee

Bird-Inspired Spatial Flapping Wing Mechanism via Coupled Linkages with Single Actuator

BLaDA: Bridging Language to Functional Dexterous Actions within 3DGS Fields

Bridging MRI and PET physiology: Untangling complementarity through orthogonal representations

Busemann energy-based attention for emotion analysis in Poincare discs

CAAP: Capture-Aware Adversarial Patch Attacks on Palmprint Recognition Models

ChangeBridge: Spatiotemporal Image Generation with Multimodal Controls for Remote Sensing

ChopGrad: Pixel-Wise Losses for Latent Video Diffusion via Truncated Backpropagation

Clickbait detection: quick inference with maximum impact

Code Sharing In Prediction Model Research: A Scoping Review

Compact Example-Based Explanations for Language Models

Complementary Filtering on SO(3) for Attitude Estimation with Scalar Measurements

Component-Adaptive and Lesion-Level Supervision for Improved Small Structure Segmentation in Brain MRI

Compressible Softmax-Attended Language under Incompressible Attention

Computational bottlenecks for denoising diffusions

ConceptTracer: Interactive Analysis of Concept Saliency and Selectivity in Neural Representations

Constellations

Contraction-Aligned Analysis of Soft Bellman Residual Minimization with Weighted Lp-Norm for Markov Decision Problem

Controller Design for Structured State-space Models via Contraction Theory

Cost-Efficient Multi-Scale Fovea for Semantic-Based Visual Search Attention

Countering the Over-Reliance Trap: Mitigating Object Hallucination for LVLMs via a Self-Validation Framework

Cram Less to Fit More: Training Data Pruning Improves Memorization of Facts

Cross-Tokenizer LLM Distillation through a Byte-Level Interface

CRPS-Optimal Binning for Univariate Conformal Regression

Current LLMs still cannot 'talk much' about grammar modules: Evidence from syntax