hiyenwong/ai_collection - Page 6

grounded-world-models-in-biological-organisms-and-future-embodied-ai

Skill for extracting and applying the grounded world modeling framework from biological organisms to inform future embodied AI systems

Langue du texte source : anglais

optimal-photostimulation-selection-for-connectomics

Optimal photostimulation selection framework (OPhELIA) for efficient causal connectomics mapping using Bayesian experimental design. Enables reconstruction of exhaustive functional connectomes with minimal trials by combining Bayesian inference, active…

Langue du texte source : anglais

positive-experience-principle-pep-methodology

Methodology for forecasting conscious choices using the Positive Experience Principle (PEP) derived from the Universal Consciousness Code theory.

Langue du texte source : anglais

project-deal

Project Deal

Langue du texte source : anglais

reservoir-computation-organization

Analyze how information processing capacity is organized across reservoir state-space modes. Use SVD mode-task projection, degree-wise representation energy, and noise-aware capacity to design or diagnose echo-state networks and physical reservoirs.

Langue du texte source : anglais

state-dependent-observation-noise-active-inference

State-Dependent Observation Noise methodology that reintroduces epistemic value in Linear-Gaussian Active Inference models. This skill provides the mathematical framework and implementation guidance for restoring curiosity-driven behavior in Gaussian agents…

Langue du texte source : anglais

stsbench-dataset

A skill for understanding and using the STSBench dataset for modeling neuronal activity in the dorsal stream of primate visual cortex. Based on arXiv:2607.15631.

Langue du texte source : anglais

subjective-functions-251215948

A skill for modeling subjective experiences as emergent properties of neural processes based on the mathematical framework of subjective functions. Use this skill when you need to bridge objective neural measurements with subjective reports, model endogenous…

Langue du texte source : anglais

thoughtseeds-dual-process-meditation

A computational phenomenology framework for modeling focused-attention meditation using dual-process active inference and hierarchical Markov-blanket architecture. Use when modeling meditation states, attentional dynamics, or cognitive phenomenology with…

Langue du texte source : anglais

transient-synapse-activity-regen

Skill for understanding and applying the Latent Excitable Recruitment (LER) framework from arXiv:2607.14000, which predicts activity regeneration in neuronal networks based on transient synaptic memory.

Langue du texte source : anglais

vnvspec-requirements-vv-specification

VNVSpec methodology for machine-readable verification and validation (V&V) specifications that bridge high-level systems-engineering requirements with low-level test results. Use when: (1) designing or auditing requirements-to-tests traceability for…

Langue du texte source : anglais

what-81000-people-want-from-ai

What 81,000 people want from AI

Langue du texte source : anglais

wind-solar-csfs-feature-selection

Skill for applying Cluster-based Sequential Feature Selection (CSFS) to improve feature selection in wind and solar power prediction tasks. Use when working with renewable energy prediction datasets that have many environmental variables and need efficient,…

Langue du texte source : anglais

rts-neural-physics-ode-learning

Hybrid neural-physics framework for learning unknown components of ODEs using Rauch-Tung-Striebel smoother and neural networks

Langue du texte source : anglais

entanglement-detection-unitary-ancilla-methodology

Methodology for measuring entanglement measures (concurrence and 3-tangle) using unitary transformations and ancilla measurements, as proposed in arXiv:2607.15201.

Langue du texte source : anglais

stochastic-quantum-spiking-neural-networks

Novel stochastic quantum spiking (SQS) neuron model with multi-qubit quantum circuits for internal quantum memory, enabling event-driven probabilistic spike generation and hardware-friendly local learning without backpropagation.

Langue du texte source : anglais

differentiable-clone-structured-causal-graphs

Skill for implementing the differentiable Clone-Structured Causal Graph (gradCSCG) algorithm for end-to-end cognitive map learning from raw image sequences, as described in arXiv:2607.12382.

Langue du texte source : anglais

modular-state-space-model-260714078

Model human perception, cognition, and decision dynamics as a modular perception-cognition-decision pipeline state-space model. Provides mathematical formulation, stability conditions, and application to rehabilitation control. Use when you need interpretable…

Langue du texte source : anglais

cognisnn-random-graph-architecture

CogniSNN: Enabling Neuron-Expandability, Pathway-Reusability, and Dynamic-Configurability with Random Graph Architectures in Spiking Neural Networks

Langue du texte source : anglais

neuroscience-spiking-afe-20260714

Skill summarizing the 32-channel event-based bio-signal analog front-end with adaptive delta and pulse frequency encoding for neuromorphic signal processing.

Langue du texte source : anglais

spikingmot-spike-driven-multi-object-tracker

SpikingMOT: A Spike-Driven Multi-Object Tracker that uses brain-inspired spiking neural networks for efficient trajectory prediction and target association. Achieves state-of-the-art performance while reducing parameters by 72% and energy by 86.7%. Use when…

Langue du texte source : anglais

formal-verification-probabilistic-snn-quotient

Formal verification toolchain for probabilistic spiking neural networks using weight-discretized quotient abstractions. CogSpike framework integrates SNN design, simulation, and PRISM-based verification. Key contributions: weight-discretized quotient model…

Langue du texte source : anglais

visual-semantic-decoding-ecog-video-stimuli

End-to-end deep learning framework for visual semantic decoding from ECoG, demonstrating promising performance without handcrafted features while maintaining interpretability.

Langue du texte source : anglais

causal-emergence-active-inference-perspective-latents

Analysis of causal emergence in active inference agents using Integrated Information Decomposition, showing how architectural separation of fast perception and slow global latents affects Φᵣ dynamics.

Langue du texte source : anglais

transition-related-potentials-narrative-comprehension-eeg

Transition-Related Potentials (TRPs) as markers of narrative comprehension in continuous EEG using deep neural networks for semi-automated analysis of naturalistic brain responses to cinematic transitions.

Langue du texte source : anglais

exact-ensemble-controllability-neural-differential-equations

Exact ensemble controllability for neural differential equations via neural interpolation - constructive solution for steering multiple initial states to corresponding target states with a single set of control parameters in neural dynamics systems.

Langue du texte source : anglais

ai-enabled-cyber-threat-mapping

Methodology for mapping real-world AI-enabled cyber attacks onto MITRE ATT&CK framework with AI Risk Enablement Score (ARiES) — identifying patterns in how threat actors weaponize AI for cyber operations.

Langue du texte source : anglais

anthropic-research-fetch

Recurring pipeline for fetching Anthropic research articles and turning reusable methodologies into ai_collection skills. Triggers on scheduled/cron Anthropic research fetches, "fetch anthropic research", extracting methods from anthropic.com/research, or…

Langue du texte source : anglais

cocot-eeg-contrastive-foundation

Contrastive pretraining methodology for EEG foundation models using multiscale convolutional Transformer architecture. Demonstrates contrastive learning as a superior alternative to masked reconstruction pretraining for EEG, which has high noise and…

Langue du texte source : anglais

continuous-metadata-conditioning-peft

Continuous (non-discretized) metadata conditioning for parameter-efficient VL/CLIP adaptation — feed numerical attributes directly into the prompt representation so the embedding space modulates smoothly, while inference stays purely visual (no metadata…

Langue du texte source : anglais

cumulant-order-block-sparse-attention

COBS - block sparse attention selector that stores a compressed SECOND-order (cumulant) statistic per block instead of only first-order (attention mass), closing the gap to dense attention at ~15x less KV-cache read traffic. Use when building/improving…

Langue du texte source : anglais

dendri-cl-single-layer-snn

DendriCL methodology for dendritic in-context learning in single-layer compartmental spiking neural networks. Proves that apical dendritic subthreshold dynamics implement online leaky LMS, collapsing ICL architectural depth to one layer with frozen inference…

Langue du texte source : anglais

deterministic-retrieval-agent-reliability

Methodology from Anthropic's "Paving the way for agents in biology" (Jun 2026). Use when building LLM agents that must query structured scientific/enterprise databases reliably. Core pattern: wrap unreliable LLM API guessing behind a deterministic retrieval…

Langue du texte source : anglais

emergent-generalization-representation-learning

Emergent generalization by representation learning in artificial neural networks. An explicit information bottleneck forcing an RNN to learn a low-dimensional representation is necessary for rotational and out-of-distribution generalization in time-series…

Langue du texte source : anglais

event-based-neural-decoding-neuroprosthetic

Event-based neural decoding framework using spiking GRU with sparse graded spikes for efficient on-device motor control. Achieves >90% decoding accuracy with <1mW power consumption on neuromorphic hardware for neuroprosthetic applications.

Langue du texte source : anglais

foveated-dynamic-token-selection

Foveation-guided dynamic token selection for robust and efficient vision transformers. Inspired by human visual system foveated sampling + eye movements. Use when building efficient ViTs, dynamic token pruning/selection, or robustness-to-noise/adversarial…

Langue du texte source : anglais

ghost-directory-detection-pattern

Methodology for detecting and resolving 'ghost directories' in ai_collection skill synchronization where skill directories exist but SKILL.md files are missing. This pattern addresses a chronic sync failure mode observed in automated research cron jobs.

Langue du texte source : anglais

graph-augmented-tree-search-agent-planning

GATS - eliminate LLM calls during agent planning by combining UCB1 tree search with a layered world model (exact symbolic match / learned statistics / LLM-for-unknown). Deterministic, zero-variance plans, 100% success on stress tests vs LATS/ReAct. Use when…

Langue du texte source : anglais

input-constrained-spatiotemporal-tubes

Input-constrained spatiotemporal tube (STT) control framework for safe navigation of unknown Euler-Lagrange systems in dynamic environments. Provides finite-time reach-avoid-stay guarantees with explicit actuator constraint handling. Approximation-free and…

Langue du texte source : anglais

krylov-complexity-loschmidt-amplitude

Relate Krylov complexity to the Loschmidt amplitude to diagnose quantum dynamics and quantum chaos. Use when analyzing operator growth, OTOCs, Lyapunov exponents, or Krylov-space methods in many-body quantum systems. Triggers: Krylov complexity, Loschmidt…

Langue du texte source : anglais