Menu
Skills in this repository
SkillsMP has collected 4,114 skills from hiyenwong/ai_collection. Open a skill to review its source and details.
hiyenwong/ai_collectionShowing 40 of 4,114 collected skills.
Proactive memory agent that runs alongside an unmodified action agent to prevent behavioral state decay in long-horizon tasks. Updates structured memory bank from trajectory and selectively injects reminders. Plug-and-play with frontier agents. +8.3pp on…
Studies quantile-based distributional RL from statistical efficiency perspective. Non-asymptotic error bound O(√(m/n)) under W∞ metric. Achieves optimal √n convergence rate. Asymptotic distribution and semiparametric efficiency bound. Berry-Esseen theorem.…
Codebook-free binary spherical coding for extreme low-bit LLM weight compression. Maps weight chunks onto unit hypersphere and binarizes into sign streams. Residual BSQ stage for reconstruction error. Category-wise recovery distillation. Activation: binary…
Procrustes-conditioned Joint End-to-end Top-K SAE for extracting cross-seed universal features from independently trained BERT models. Combines Top-K sparsity, end-to-end optimization, and dead-feature revival. Pearson r ≥ 0.70 across seeds. Activation:…
Hierarchical structure-aware document analysis system using LLMs. Parses documents into hierarchical trees preserving layouts, builds structure-aware semantic indices for filtering and question answering. Handles academic papers, technical manuals, financial…
Training-free best-first draft tree for speculative decoding using Domino's conditional non-factorized correction. Achieves up to 6.6x speedup on Qwen3-4B and highest mean accept length (10.7 tokens/round). GPU-native CUDA-graph builder for efficient tree…
Shows that post-training quantization evaluation via accuracy/perplexity fails to capture behavioral changes. Introduces correctness agreement metric. Reveals non-linear breakpoints at low bit-widths. Query/key projections more sensitive than value/output.…
MAESTRO: Markov-chain Approximated Expert Sparsification via Transition-based Routing for MoE structured pruning. Models expert activation as Ergodic Markov chains for globally-aware importance. Outperforms baselines by 10.61% at 50% compression. Lower…
Balanced session-centric LLM scheduling for agent serving workloads. Routes first request in each session for load balance and follow-ups cache-aware. 10-16% TPS improvement. Leverages intra-session locality and 80%+ KV-reuse in agent traces. Activation: LLM…
Shows that Super Weight pruning degradation doesn't universally apply. Training Super Weights in isolation drops accuracy to random-guessing. Parameter importance ≠ trainability. Vanilla LoRA with 0.16% parameters succeeds. Activation: super weights, LLM…
Practical investigation of training-free relaxed speculative decoding for LLM inference acceleration. Unifies existing approaches within a shared framework, benchmarks on contemporary settings. Relaxed speculation trades lossless guarantees for speed-ups and…
gspDAG-FL: secure decentralized federated learning via gossip and virtual voting. Derives consensus from gossip history, uses Hashgraph-style virtual voting on compact DAG. Byzantine resilience with payload validation and semantic audit. Activation:…
FabriVLA: lightweight VLA model combining InternVL3.5 VLM backbone with flow-matching action head. Gated self-attention across action tokens, shallow VLM layer fusion. 90.0% success on Meta-World MT50. 1B-scale VLM without billion-parameter backbone.…
LingBot-VA 2.0: video-action foundation model built from the ground up for robot embodiment. Semantic visual-action tokenizer, causal pretraining from scratch, sparse MoE backbone, asynchronous inference for real-time closed-loop control. Few-shot…
Methodology for studying magnetic field effects on chimera states in Hindmarsh-Rose neuronal networks. Covers traveling chimera, multicluster chimera, and multicluster chimera breather transformations under spatial magnetic field applications.
Non-Hermitian Potential Well Formalism for Conscious-Preconscious-Subliminal Processing methodology. Models the Global Neuronal Workspace (GNW) as a complex-valued landscape where sensory encoding and conscious access are unified. Activation: non-Hermitian…
Dendritic In-Context Learning (DendriCL) — single-layer compartmental SNN that achieves ICL via dendritic subthreshold dynamics implementing online LMS. Use when: designing SNN architectures for in-context learning, exploring biologically-plausible ICL…
DRIADA: Open-source Python toolkit for cross-scale analysis of single-neuron selectivity and population dynamics. Unifies neural signals and behavior in shared data model for selectivity testing, dimensionality reduction, and network analysis. Activation:…
Non-Hermitian Potential Well Formalism for modeling conscious-preconscious-subliminal processing hierarchy. Uses nonlinear Schrödinger-type equations in imaginary time with non-Hermitian Hamiltonians to unify sensory encoding and conscious access.
arXiv paper search skill - search academic papers by keywords, authors, categories. Supports time filtering, category filtering, and paper detail retrieval. Activation: arxiv search, paper search, 论文搜索, search papers, arxiv 论文.
TRIBE v2 数据增强提升脑到图像解码性能方法论。使用大规模预训练编码模型生成合成fMRI数据，在小数据集上实现显著性能提升。
DendriCL methodology for dendritic in-context learning in single-layer spiking neural networks. Demonstrates that a single dendritic compartment with online-LMS dynamics implements complete in-context learning, eliminating the need for attention, depth, or…
Doob-Barrier-Conditioned Diffusion methodology that turns analog neuromorphic device noise into a continual-learning resource. Casts per-synapse consolidation as a Doob h-transform, creating a noise-amplified restoring force that consolidates memories —…
Non-Hermitian potential well formalism for the subliminal-preconscious-conscious processing hierarchy in the Global Neuronal Workspace. Uses nonlinear Schrödinger-type equation in imaginary time with non-Hermitian, non-normal Hamiltonian to model conscious…
Optimal control-based strategy for enhancing impulse estimation in Gaussian quantum systems via parametric modulation. Use when designing quantum sensing protocols, estimating transient disturbances in quantum systems, optimizing state preparation for impulse…
Framework for designing magnetic-field-free quantum reservoir computing systems using engineered organic materials. Bridges quantum computing and neuroscience through the 3-layer quantum brain hypothesis. Use when: designing organic qubit systems, quantum…
Penalty-free QAOA methodology for lattice protein folding using conflict graph independent set formulation. Use when: (1) quantum optimization for protein folding or molecular structure prediction, (2) QAOA without penalty terms, (3) conflict graph maximum…
Penalty-free quantum optimization methodology for lattice protein folding using QAOA and quantum annealing. Avoids constraint penalty terms that cause energy landscape distortion. Maps protein conformations to binary optimization without penalty parameters.…
Pre-asymptotic trainability analysis for photonic variational quantum circuits under postselection. Covers barren plateau dynamics in passive linear-optical circuits, Lie algebra dimension scaling, postselection-induced gradient concentration (allow-bunching,…
Q-ANCHOR architecture for Quantum Federated Learning (QFL) that addresses double-drift phenomenon (client drift from non-IID data + hardware bias from noisy quantum gradients). Uses ZNE-guided server anchoring and stateful client correction. Proves…
Unsupervised clustering via steady-state quantum transport in open quantum networks (GKSL master equation). Encodes data as input states and infers cluster assignments from terminal current observables - no full state tomography required. Use when: quantum…
Quantum Machine Learning with Equilibrium Propagation for medical image analysis. Energy-based training without backpropagation using Variational Quantum Circuits (VQCs) for resource-constrained quantum hardware. Use when: analyzing blood cells, leukemia…
Design framework-agnostic quantum machine learning (QML) systems using the Model-Agnostic Learning System (MALS) paradigm. Extracts QML models from any framework (PennyLane, Qiskit, TensorFlow Quantum, etc.) into portable representations with auto-validation…
Quantum Machine Learning model testing and robustness analysis methodology. Covers mutation testing for QNN circuits, accuracy/robustness evaluation of Variational Quantum Circuits (VQCs), and practical considerations for deploying QML models on NISQ-era…
Quantum-enhanced AI reliability patterns from cutting-edge research. Covers certified training of quantum neural networks, quantum interval bound propagation (QIBP), genetic algorithm-based HQNN optimization (GAT-QNN), distributed quantum reinforcement…
Quantum compiler qubit mapping and routing methodology for scalable quantum circuit compilation. Covers position graph abstraction, heuristic mapper optimization (SABRE), memoized congestion resolution, and architecture-aware compilation for heterogeneous…
Engineering patterns for reliable, efficient quantum control systems. Covers pulse-level gate optimization, real-time closed-loop QEC, dynamic decoder scheduling, physics-informed LLM control, and thermodynamic control optimization. Use when designing quantum…
Quantum Erasure Imaging (QEI) methodology — turns delayed-choice quantum erasure into practical dual-modality imaging protocol. Simultaneously reconstructs absorption T(x,y) and phase-sensitive quadrature from a single entangled photon run. Use when: quantum…
Evaluate quantum error-correcting codes under hardware-motivated and biased noise models. Benchmark fault-tolerant quantum computing primitives via noisy stabilizer simulation. Use when: (1) evaluating QEC code performance, (2) designing fault-tolerant…
Circuit-level backdoor detection methodology for Quantum Federated Learning (QFL) systems. Identifies malicious circuit patterns in variational quantum circuits during federated training. Use when: (1) securing QFL systems, (2) detecting quantum circuit…