菜单
这个仓库中的 skills
SkillsMP 已收集 hiyenwong/ai_collection 中的 4,114 个 Skill。打开任一 Skill 可查看来源和详情。
hiyenwong/ai_collection已展示 40 / 4,114 个已收集 Skill。
TerraZero: Procedural Driving Simulation for Zero-Demonstration Self-Play at Scale - Training robust autonomous driving agents requires a simulator that is fast enough for reinforcement learning at scale, realistic enough to ground beh...
原文语言：英语
Unveiling Complex Collective Behaviors from Simple Rewards - Multi-agent Reinforcement Learning (MARL) holds great potential for robot swarms, but the black-box nature of neural policies complicates strategic an...
原文语言：英语
Dynamic neural manifolds methodology for flexible closed-loop control on neuromorphic hardware. Uses ring attractor networks with sensory-modulated control neurons (speed, shape, selection) to drive subspace rotations and fine-grained trajectory control in…
原文语言：英语
Human-AI Agent Interaction as a Neuroplastic Training Environment - Interaction with AI agents has become one of the most frequent activities of everyday digital life. Whether conversing with an assistant, working with...
原文语言：英语
Knowledge- and Gradient-Guided Reinforcement Learning for Parametrized Action Markov Decision Processes - In this paper, we study Reinforcement Learning in Parametrized Action Markov Decision Processes (PAMDP), where each decision consists of a symbolic ac...
原文语言：英语
LLMs Can See the Smoke but not the Fire: Evaluating Abductive Reasoning with Elenchos - Large language models (LLMs) excel at pattern recognition and text generation, but their capacity for abductive inference - inferring latent hypothese...
原文语言：英语
Do AI Agents Know When a Task Is Simple? Toward Complexity-Aware Reasoning and Execution - Large language model (LLM) agents increasingly automate multi-step engineering and informatics workflows, yet they rarely ask how much effort a task a...
原文语言：英语
MemOps: Benchmarking Lifecycle Memory Operations in Long-Horizon Conversations - Long-term memory has become a foundational capability for LLM-based agents that accompany users across extended, multi-session interactions. Existing ...
原文语言：英语
Tracing Agentic Failure from the Flow of Success - Failure attribution for LLM-based agentic systems, i.e., identifying which steps in a failure trajectory caused the task to fail, is critical for debu...
原文语言：英语
Who Grades the Grader? Co-Evolving Evaluation Metrics and Skills for Self-Improving LLM Agents - Self-evolving agent systems improve by creating, revising, and retiring their own skills, but every such loop rests on a hidden assumption: a reliable...
原文语言：英语
Win by Silence: Deletion Non-Monotonicity, Autonomous Exploitation, and Typed-State Gating in LLM Plan Evaluation - Plan evaluators can reward a strategic plan for becoming less explicit. This paper studies that failure in a staged expected-value scorer for…
原文语言：英语
A Learning-Rate-Gated Failure of GRPO in a Small Language and Vision-Language Model Web Agent: A Controlled Null and Its Mechanism - Reinforcement learning with verifiable rewards, and Group Relative Policy Optimization (GRPO) in particular, is now run…
原文语言：英语
Can Induced Emotion Bias LLM Behaviors in Sequential Decision Making? - As Large Language Models (LLMs) are increasingly deployed as autonomous agents in high-stakes domains, understanding contextual factors that may modul...
原文语言：英语
Constraint-Aware Aggregation for Federated Reinforcement Learning in Microgrid Energy Coordination - Federated Reinforcement Learning (FedRL) enables coordination of distributed energy resources without sharing raw local data, but standard aggregation...
原文语言：英语
PalmClaw: A Native On-Device Agent Framework for Mobile Phones - Large Language Model (LLM) agents have moved beyond generating responses to executing multi-step tasks by calling tools, observing the results, and it...
原文语言：英语
Research paper: StructAgent: Harness Long-horizon Digital Agents with Unified Causal Structure. Brief summary of key findings and contributions.
原文语言：英语
LLM interpretability methodology from Anthropic's "A global workspace in language models" (Jul 2026). Use when probing what a language model is "thinking but not saying" — its consciously-accessible / broadcast internal representations — via the Jacobian lens…
原文语言：英语
Alignment training methodology from Anthropic's "Teaching Claude why" research (May 2026). Use when fine-tuning or RLHF-aligning LLMs against adversarial/misalignment evals, or designing OOD-generalizing safety training data. Covers four lessons:…
原文语言：英语
Dendritic In-Context Learning (DendriCL): a single-layer compartmental spiking neural network whose apical/dendritic subthreshold dynamics implement online Widrow-Hoff LMS, giving in-context learning without attention, depth, or inference-time synaptic…
原文语言：英语
Dynamic neural manifolds methodology for flexible closed-loop control on neuromorphic hardware. From arXiv:2607.07373 (von Seeler et al., Jul 2026).
原文语言：英语
Methodology from Anthropic policy analysis examining possible trajectories for US-China AI competition by 2028 — focusing on compute advantage, export controls, distillation attacks, and two scenarios for democratic vs. authoritarian AI leadership.
原文语言：英语
分析AI能否具有意识的物理主义框架，区分模拟与实例化的本体论边界，提出制图者依赖的计算理论
原文语言：多语言混合
Expert guidance for web accessibility (WCAG 2.2). Treat a11y violations as compile errors. Use when implementing accessibility features, auditing a11y compliance, or working with screen readers. Triggers on: accessibility, wcag, a11y, aria, screen reader,…
原文语言：英语
Active Predictive Filtering paradigm for Spiking Transformers. Inspired by the brain's predictive coding mechanism, actively suppresses predictable signals and focuses on salient visual features. Activation: active predictive filtering, spiking transformer,…
原文语言：英语
Proposes that active sensing (energy expenditure for information) is not driven by sensory goals but is necessary for task-level control. Integrates empirical data and control theory to explain explore-exploit mode switching in biological sensorimotor…
原文语言：英语
Theoretical framework proposing that active sensing (movement for information) is not driven by sensory goals but is necessary for task-level control, with explore/exploit mode switching. Based on arXiv:2605.22988 (May 2026). Use when studying active sensing,…
原文语言：英语
Adaptive acquisition function selection for discrete black-box optimization (BOCS + GP-LCB hybrid). Based on arXiv:2605.10856 (Shikanai & Ohzeki, 2026). Use when optimizing QUBO, HUBO, or any discrete-variable black-box problem where BOCS stagnates. Combines…
原文语言：英语
Theory of phase-locked activity in delayed spiking networks using the Haken Lighthouse model — an analytically tractable event-based framework bridging integrate-and-fire networks and coupled phase oscillators. Derives self-consistency conditions for…
原文语言：英语
Two-stage adaptive confidence-gated neural decoding framework for quantum error correction — lightweight neural fast-path with high-confidence fallback to classical refinement. Use when designing real-time QEC decoding, neural-classical hybrid inference,…
原文语言：英语
Adaptive distributionally robust optimal control for handling Knightian uncertainty in stochastic systems. Addresses epistemic uncertainty through adaptive DROC methods. Use when: (1) Designing robust control systems with distribution uncertainty, (2)…
原文语言：英语
Adaptive flow routing methodology for uncovering latent communication patterns in brain networks. Models information flow through structural connectivity to reveal hidden communication pathways. Activation: brain network communication, flow routing, latent…
原文语言：英语
Adaptive-Frequency Resonate-and-Fire (ARF) neurons for spectral estimation of streaming signals. Neuromorphic-inspired method that dynamically adjusts internal frequency to match dominant frequency components, enabling real-time range/velocity estimation in…
原文语言：英语
MorphSNN: Adaptive Graph Diffusion and Structural Plasticity for Spiking Neural - Bio-inspired undirected diffusion for signal propagation in . Activation triggers: adaptive, graph, diffusion, neuroscience, SNN.
原文语言：英语
Adaptive Hybrid Quantum-Classical Feature Fusion methodology for medical image classification. Addresses optimization asymmetries between quantum and classical paradigms using Temperature-Scaled Hybrid Fusion (TSHF), Dynamic Hybrid Fusion (DHF), and Static…
原文语言：英语
Adaptive quantum-classical feature fusion methodologies for medical AI diagnosis. Covers Temperature-Scaled Hybrid Fusion (TSHF), tensor-network compression with quantum refinement, and multi-head quantum-aware encoding. Use when building hybrid…
原文语言：英语
Adaptive Spiking Neuron (ASN) methodology for vision and language modeling. Implements trainable membrane potential dynamics with adaptive firing mechanisms for efficient Spiking Neural Networks (SNNs). Activation: adaptive spiking neuron, ASN, spiking neural…
原文语言：英语
Adaptive Spiking Neuron (ASN) methodology for vision and language modeling. Trainable membrane potential dynamics, integer training + spike inference, NASN variant with normalization. Activation: adaptive spiking neuron, asn, nasn, trainable spiking neuron,…
原文语言：英语
Adaptive Spiking Neuron (ASN) methodology for vision and language modeling - a general-purpose spiking neuron family evaluated on 19 datasets across 5 distinct tasks. Use when: (1) Implementing energy-efficient vision models with SNNs, (2) Building language…
原文语言：英语
Energy-efficient Spiking Transformer using attention-driven sparse spike propagation. Reduces FLOPs 87.7-97.5% vs dense Transformers by replacing softmax with temporal spike coding. Activation: spiking transformer, energy-efficient vision, spike-based…
原文语言：英语
Theoretical framework comparing in-context learning (fixed queries) vs agentic learning (adaptive queries) under neural network realizability constraints. Use when: analyzing when adaptive querying helps or hurts, comparing ICL with agentic RL, understanding…
原文语言：英语