hiyenwong/ai_collection - Page 40

unifying-von-neumann-hpc-neuromorphic-ebbrains

统一冯诺依曼HPC与神经形态计算的EBRAINS工作流框架 - 透明跨平台执行SNN，支持异构架构无缝切换。

Source text: Chinese

structured-sparse-autoencoders-cross-modal-concepts

Structured Sparse AutoEncoder (S²AE) that enforces concept consistency in vision-language models. Uses grouped image patches with attention similarity and spatial proximity for structured sparsity regularization. Achieves 6.06% improvement in semantic…

beyond-backpropagation-monte-carlo-train-deep-networks

Shows that simple Monte Carlo random mutation can train deep neural networks without gradients. No batch normalization or residual connections needed. Supports pure pruning training, discrete weights, and unconventional transfer functions. Demonstrated on 20+…

slorr-in-training-low-rank-regularization

Low-rank factorization for neural network compression that directly regularizes weight matrices using GPU-friendly approximations. Stateless, architecture-preserving, with less than 8% training overhead. Evaluated on ImageNet and LLM pretraining at 135M and…

druggen-2-disease-aware-language-model-drug-discovery

DrugGen-2: generative model that designs small molecules conditioned on disease ontology and target protein sequences. Fine-tuned GPT-2 with SFT + GRPO. Outperforms baselines on diabetic nephropathy targets with improved binding affinities. Use when working…

cognitive-structured-multimodal-agent-understanding-generation

Cognitive-structured Multimodal Agent (CMA) with Episodic Visual Memory for long-horizon multimodal dialogue. Perceptual Abstraction Engine, Cognitive Retrieval Engine, and Multimodal Executive Controller. 8B agent achieves 91.4% retrieval accuracy over…

wcog-vla-dual-level-world-cognitive-autonomous-driving

WCog-VLA: dual-level World-Cognitive VLA framework bridging semantic world forecasting with generative world evolution for proactive autonomous driving. Features Game-theoretic Chain-of-Thought reasoning and Aligned Decoupled Diffusion Transformer. SOTA PDMS…

ai-guided-stimuli-discovery-facial-emotion-autism

AI-guided framework using population-specific neural network models to discover and generate facial stimuli that maximize perceptual differences between autistic and neurotypical adults. Uses GAN to transform diagnostic images for behavioral assay…

contravariance-theory-strong-alignment-minimal-solutions

Theory formalizing contravariance in NeuroAI: weak alignment of network representations via affine mappings guarantees strong alignment of privileged axes, and alignment zippers up the network hierarchy. Shows convergent evolution between artificial and brain…

llm-judge-theory-agnostic-personality-recognition

JAM: theory-agnostic framework for personality recognition using LLM-as-a-Judge for adaptive metric alignment. Attention-Pooled Graph Prototypical Network with Cross-Theory Harmonization. LLM operates in before-the-loop and in-the-loop configurations for…

anytime-lidar-resolution-scaling-object-detection

Anytime computing method for LiDAR-based 3D object detection in cyber-physical systems. Multi-resolution inference with single DNN model, deadline-aware scheduler predicts execution time for all resolutions. Deployed in simulated autonomous driving with…

self-adaptive-anomaly-detection-rl-human-feedback-vehicles

Online anomaly detection framework for connected vehicles integrating factorized deep Q-network with self-attention, statistical drift detectors, and human-in-the-loop retraining. F1 score 0.69 vs 0.11 for single detectors, with sustained adaptation after…

ardy-autoregressive-diffusion-interactive-human-motion

Streaming generation framework for real-time 3D human motion controllable via online text prompts and kinematic constraints. Uses hybrid representation combining explicit root features with latent body embedding. Two-stage autoregressive transformer denoiser…

pose-to-biomechanics-bridging-3d-pose-biomechanical

BioModule: lightweight plug-in temporal transformer that attaches downstream of any 3D pose estimator to predict biomechanical attributes from standard 17-joint 3D skeletons. Estimator-agnostic, requires no modification of upstream pose model. Use when…

frozen-lgp-qaoa

FrozenLGP adaptive decomposition framework for divide-and-conquer QAOA. Transforms partitionability from an assumption into an enforceable property by freezing obstructing vertices and folding their energy into linear bias terms. Enables 100% decomposition…

bosonic-stellar-rank-qec

Stellar rank as resource measure for bosonic quantum error correction. Designs and benchmarks bosonic codes under finite non-Gaussian resources, revealing noise-adapted code structures and concrete resource thresholds.

hlbg-hyperbolic-learning-brain-graphs

Hyperbolic Learning on Brain Graphs (HLBG) methodology for brain network analysis using Lorentzian hyperbolic space to model hierarchical ROI-community-whole-brain relationships. Introduces Graph-aware Mamba (GaMamba) for capturing long-range dependencies…

krein-space-riemann-xi

Spectral interpretation of the Riemann xi-function via Krein space quantization in de Sitter QFT. Uses invariant two-point functions, Legendre functions, Lorentzian harmonic analysis, and Mehler-Fock transform to construct a retarded propagator with…

pathwise-metastability-galves-locherbach

Pathwise approach to metastability for Galves-Löcherbach (GL) stochastic spiking neural network models. Reviews metastability theory from chemistry to probability theory, provides general definition encompassing GL model variants, surveys established…

stark-units-sic-overlaps

Number-theoretic characterization of SIC-POVM overlap units via Stark units from ray class fields. Bridges algebraic number theory (Stark units, ray class fields, Shintani-Faddeev cocycle) with quantum information (SIC-POVM geometry, mutual scalar products).…

topological-grid-cell-decoding-codes

Topological decoding of grid cell activity via path lifting to covering spaces. Uses TDA to extract toroidal coordinates from grid cell populations and reconstructs spatial trajectories from a single module without training data or external position info.…

girsanov-quantum-control

Girsanov's theorem path-space regularization methodology for robust open quantum control. Bridges stochastic calculus (Girsanov theorem) with quantum optimal control, penalizing observable consequences of control on decoherence channels rather than control…

invariance-audits-quantum-kernels

Real-to-Hermitian taxonomy for auditing representation invariances in quantum machine learning. Covers Grassmann/flag projector kernels, quantum fidelity kernels, QVR anchor operators, and quotient-witness experiments for validating geometric lifts.…

quantum-grokking-analysis

Grokking and epoch-wise double descent analysis for quantum neural networks (QNNs). Covers delayed generalization transition, overparameterization effects, weight-norm regularization, and algorithmic stability in variational quantum circuits. Use when…

quantum-hoare-logic

Quantum Hoare Logic with Integer Hybrid Path-Sums (IHPS) for formal verification and resource estimation of hybrid quantum-classical programs with unbounded loops. Bridges formal methods, quantum programming, and static analysis.

quantum-reservoir-chaotic-forecasting

Quantum reservoir computing architecture for forecasting chaotic systems with a high-dimensional feature space diagnostic. Provides reproducible recipe for QRC circuit construction, data injection, readout training, and stability-number-based dimensionality…

quantum-minimax-estimation

Quantum minimax estimation methodology for high-order functionals — using quantum arguments to achieve optimal sample complexity for classical and quantum functionals (Rényi entropy, Tsallis entropy). Use when estimating high-order functionals of discrete…

solarchain-eval-a-physics-constrained-benchmark-for-trustworthy

SolarChain-Eval: A Physics-Constrained Benchmark for Trustworthy Economic Agents in Decentralized Energy Markets. As agentic AI systems are increasingly applied to cyber-physical environments, their evaluation requires assessment of both task performance and…

sampling-on-random-subspaces-under-limited-data-in

Sampling on Random Subspaces under Limited Data in the Context of Exploratory Landscape Analysis. Classical space-filling designs often fail to provide reliable statistical results for Exploratory Landscape Analysis (ELA) when only limited evaluation budgets…

delay-aware-active-triangulation-with-uncertainty-driven-multi

Delay-Aware Active Triangulation with Uncertainty-Driven Multi-Agent Reinforcement Learning for Counter-UAS. Multi-agent active visual triangulation enables precise 3D localization of aerial targets by coordinating mobile observers with controllable cameras.…

early-to-share-late-to-save-synchronisation-driven

Early to Share, Late to Save: Synchronisation-Driven Communication Gating in Bandwidth-Constrained Cooperative VLN. Most cooperative Vision-Language Navigation (VLN) methods assume unlimited communication, not considering real-world applications where…

from-triggers-to-emotions-a-cpm-grounded-appraisal

From Triggers to Emotions: A CPM-Grounded Appraisal Multi-Agent for Dynamic Emotional Evolution in Persona-Based Dialogue. Large Language Models (LLMs) have substantially advanced persona-based dialogue agents for emotion-sensitive role simulation in…

multi-agent-robotic-control-with-onboard-vision-language

Multi-Agent Robotic Control with Onboard Vision-Language Models. Vision Language Models (VLMs) and Vision Language Action (VLA) models have shown promise in robotic control. Yet, they face significant challenges regarding explainability, generalization, and…

webswarm-recursive-multi-agent-orchestration-for-deep-and

WebSwarm: Recursive Multi-Agent Orchestration for Deep-and-Wide Web Search. Large language model (LLM)-based web search agents are transforming information seeking from simple factoid question answering into complex, deep-and-wide search and research-oriented…

a-non-hermitian-potential-well-formalism-for-conscious

A Non-Hermitian Potential Well Formalism for Conscious--Preconscious--Subliminal Processing. We propose a phenomenological model of the Global Neuronal Workspace (GNW) in which early sensory processing generates an effective complex-valued landscape governing…

quantifying-entrainment-evidence-a-comparison-of-frequentist-and

Quantifying Entrainment Evidence: A Comparison of Frequentist and Bayesian Approaches for Information Processing Pathway Maps. Information Processing Pathway Maps (IPPMs) offer a scalable framework for formalizing the complex sequence of mathematical…

using-hierarchical-statistical-learning-models-to-model-individual

Using hierarchical statistical learning models to model individual statistical learning. Statistical learning is essential for individuals to discover structure in the sensory environment, especially during communication via speech or music. Individual…

ultrax-refining-pre-training-data-at-scale-with

UltraX: Refining Pre-Training Data at Scale with Adaptive Programmatic Editing. As available training data approaches its physical limit, gains from Scaling Laws have begun to diminish. Consequently, improving Large Language Models (LLMs) now depends less on…

workflow-as-knowledge-semantic-persistence-for-llm-mediated

Workflow as Knowledge: Semantic Persistence for LLM-Mediated Workflows. Large language model (LLM) applications increasingly use explicit workflows for tool use, retrieval, branching, checkpointing, and human approval. Existing workflow systems already…

learning-adaptive-solvers-for-distributed-factor-graph-optimization

Learning Adaptive Solvers for Distributed Factor Graph Optimization on Matrix Lie Groups. Modern robotic perception increasingly involves large-scale geometric optimization problems distributed across multiple robots or sessions. However, existing distributed…