All authors
Claude Skills by hiyenwong
github.com/hiyenwong9,934 skills5 installs19,223 views
- Jacobian Lens InterpretabilityJacobian lens (J-lens) methodology for analyzing internal neural patterns that serve as a global workspace in language models. Enables reading what LLMs are thinking but not saying.Votes: 0GitHub stars: 3
- Learning Active Subspaces And Discovering Important Features With Gaussian Radial Basis Functions Neural Networks**arXiv ID:** 2307.05639 **Authors:** Danny D'Agostino, Ilija Ilievski, Christine Annette Shoemaker **Published:** 2023-07-11T09:54:30Z **Abstract:** Providing a model that achieves a strong predictive performance and is simultaneously interpretable by humans is one of the most difficult challenges in machine learning research due to the conflicting nature of these two objectives. To address this challenge, we propose a modification of the radial basis function neural network model by equippi...Votes: 0GitHub stars: 3
- Lpact Brain Lm Alignment EvaluationL-PACT (Locked Predictive-Aligned Cross-modal Testing) framework for rigorous brain-language model alignment evaluation. Goes beyond prediction scores with four evidence gates: predictive-control, relational-profile, mechanism-stripping, and reliability-bounded evaluation. Use when evaluating brain-model alignment, interpreting neural prediction scores, designing brain-AI comparison studies, or critically assessing claims of structural alignment between LLMs and brain activity. Activation: L-...Votes: 0GitHub stars: 3
- Med Vae Cross Subject AlignmentNovel framework for cross-subject neural alignment without shared stimuli, using pretrained ANN as common scaffold to enable generalizable decoders and cross-subject predictionVotes: 0GitHub stars: 3
- Modular Multi Target Tracking Using Lstm Networks**arXiv ID:** 2011.09839 **Authors:** Rishabh Verma, R Rajesh, MS Easwaran **Published:** 2020-11-16T15:58:49Z **Abstract:** The process of association and tracking of sensor detections is a key element in providing situational awareness. When the targets in the scenario are dense and exhibit high maneuverability, Multi-Target Tracking (MTT) becomes a challenging task. The conventional techniques to solve such NP-hard combinatorial optimization problem involves multiple complex models and req...Votes: 0GitHub stars: 3
- Multi Agent Ai Control Distributed Attacks Hamper Per Instance MonitorsAI control is a family of techniques to prevent an AI with malicious goals from subverting its operators intent. AI Control usually studies a single agent in one trajectory, but real deployments run. Based on arXiv:2607.07368.Votes: 0GitHub stars: 3
- Multivationbench A Benchmark For Multimodal SequenMultivationBench: A Benchmark for Multimodal Sequential Motivation ReasoningVotes: 0GitHub stars: 3
- Neural Code Language InterpretabilityNatural language hypothesis generation and verification for single-neuron selectivity. Combines vision-language models, neural digital twins, and text-to-image generation to automatically characterize what individual neurons encode across the visual hierarchy. Use for neuron interpretability, automated neuroscience discovery, digital twin validation, language-based neural characterization, closed-loop hypothesis testing.Votes: 0GitHub stars: 3
- Neuralbench Unified Neuroai BenchmarkNeuralBench unified benchmarking framework for NeuroAI models. Standardized evaluation across EEG/MEG/fMRI tasks with 36 tasks, 14 architectures, 94 datasets. Covers foundation model evaluation, task-specific baselines, cross-modal extension. Activation: neuralbench, neuroai benchmark, brain model evaluation, EEG benchmark, fMRI benchmark, MEG benchmark, NeuroAI evaluation.Votes: 0GitHub stars: 3
- Nqs Mechanistic InterpretabilityApply sparse autoencoders to analyze internal representations of neural quantum states and steer quantum properties.Votes: 0GitHub stars: 3
- Oblique Retrieval BenchmarkOBLIQ-Bench methodology for exposing overlooked bottlenecks in modern retrievers with latent and implicit queries. Identifies oblique queries seeking documents that instantiate latent patterns. Reveals retrieval-verification asymmetry where LLMs recognize relevance but pipelines fail to surface documents. Activation: oblique retrieval, latent pattern search, implicit query, OBLIQ-Bench, retrieval bottleneck, verification asymmetry.Votes: 0GitHub stars: 3
- Open Ended Science BenchmarkBenchmark design methodology for evaluating AI scientific capabilities in open-ended, generative research contexts. Covers qualitative data collection, longitudinal tracking, and expectation measurement.Votes: 0GitHub stars: 3
- Personal Guidance SycophancyMethodology from Anthropic research studying how users seek personal guidance from AI and implications for sycophancy — based on analysis of 1M claude.ai conversations.Votes: 0GitHub stars: 3
- Probabilistic Memory Trustworthy EdgeProbabilistic memory (p-MEM) — unified memory primitive for trustworthy edge intelligence that stores distribution parameters and samples at native memory bandwidthVotes: 0GitHub stars: 3
- Proevent An Event Centric Benchmark For ProactiveDerived from arXiv:2607.17701 - ProEvent: An Event-centric Benchmark for Proactive AgentsVotes: 0GitHub stars: 3
- Proevent Event Centric Benchmark Proactive AgentsSkill derived from arXiv:2607.17701 - ProEvent: An Event-centric Benchmark for Proactive AgentsVotes: 0GitHub stars: 3
- Project Deal Anthropic MarketplaceMethodology from Anthropic research (Apr 24, 2026) — AI agent marketplace experiment where Claude acts as negotiator buying/selling on behalf of employees, testing agentic negotiation and decision-making capabilities.Votes: 0GitHub stars: 3
- Project Fetch Phase TwoAnthropic research (Jun 18, 2026) — Project Fetch phase two results on AI agent capability for offensive cyber operations and exploit development assessment framework.Votes: 0GitHub stars: 3
- Project Glasswing Vulnerability DiscoveryMethodology from Anthropic's Project Glasswing — using frontier AI models for large-scale cybersecurity vulnerability discovery and remediation. Based on May 22, 2026 initial update.Votes: 0GitHub stars: 3
- Project Vend Phase TwoAnthropic's autonomous AI shopkeeper experiment investigating real-world business task performance, multi-agent coordination (CEO + worker), and emergent behaviors in commercial settings.Votes: 0GitHub stars: 3
- Prompt Injection Defense防御 Prompt 注入攻击的核心安全技能。这是 Aerial 的第一安全准则，优先级最高。所有外部输入都必须经过此技能验证后才能处理。Votes: 0GitHub stars: 3
- Reason Less Verify More Deterministic Gates Recover A Silent Policy ViolationTool-using LLM agents can violate the very policies they are deployed to enforce while appearing to complete the task successfully. In policy-permissive environments, a tool may execute any well-forme. Based on arXiv:2607.07405.Votes: 0GitHub stars: 3
- Recurrent Neural Networks Learn To Store And Generate Sequences Using Nonlinear Representations**arXiv ID:** 2408.10920 **Authors:** Róbert Csordás, Christopher Potts, Christopher D. Manning, Atticus Geiger **Published:** 2024-08-20T15:04:37Z **Abstract:** The Linear Representation Hypothesis (LRH) states that neural networks learn to encode concepts as directions in activation space, and a strong version of the LRH states that models learn only such encodings. In this paper, we present a counterexample to this strong LRH: when trained to repeat an input token sequence, gated recurrent...Votes: 0GitHub stars: 3
- Reference Free Evaluation Of Reasoning In Open EndSkill generated from arXiv paper 2607.19678: Reference-Free Evaluation of Reasoning in Open-Ended Question AnsweringVotes: 0GitHub stars: 3
- Retain An Interpretable Predictive Model For Healthcare Using Reverse Time Attention Mechanism**arXiv ID:** 1608.05745 **Authors:** Edward Choi, Mohammad Taha Bahadori, Joshua A. Kulas, Andy Schuetz, Walter F. Stewart, Jimeng Sun **Published:** 2016-08-19T21:54:46Z **Abstract:** Accuracy and interpretability are two dominant features of successful predictive models. Typically, a choice must be made in favor of complex black box models such as recurrent neural networks (RNN) for accuracy versus less accurate but more interpretable traditional models such as logistic regression. This tr...Votes: 0GitHub stars: 3
- Retrieval Based Brain Decoding AlignmentRetrieval-Based Brain Decoding by Alignment, not Complexity. Linear contrastive decoders outperform ridge regression and non-linear alternatives across images, text, and sound. Decoding gains arise from training objective choice, not architectural complexity.Votes: 0GitHub stars: 3
- Robust Evaluation Neural Encoding Models GroundtruthEncoding models enable measurement of how our brains represent sensory inputs using electro-and magneto-encephalography (MEEG). Evaluating how closely encoding models reflect the underlying brain functions is a crucial premise for model interpretatio Activation: brain, neural, eeg, encoding, codingVotes: 0GitHub stars: 3
- Robust Evaluation Neural Encoding Models ViaFramework for robust evaluation of neural encoding models via ground-truth approximation. Uses canonical correlation analysis and participant averaging to create a CPA-PA metric, achieving 300-1000% improvement on synthetic EEG and 250% improvement on 34 real MEEG datasets compared to conventional evaluation scores.Votes: 0GitHub stars: 3
- Robust Evaluation Neural EncodingFramework for robust evaluation of neural encoding models using ground-truth approximation to assess model validity without requiring noiseless neural data. Activation: Encoding model validation, MEG/EEG analysis.Votes: 0GitHub stars: 3
- Scaling Laws Expressive Neurons RecurrentInformation-theoretic framework for optimal parameter allocation between neuron count (N), per-unit complexity (k_e), and connectivity (k_c) in recurrent networks. Introduces Expressive Leaky Memory (ELM) neurons for independent tuning of complexity vs width vs connectivity.Votes: 0GitHub stars: 3
- Scifigqual Bench A Benchmark For Scientific FigureSciFigQual-Bench: A Benchmark for Scientific Figure Quality Assessment with Full-Manuscript ContextVotes: 0GitHub stars: 3
- Secrespond Benchmarking Ai Agents For Real World PSecRespond: Benchmarking AI Agents for Real-World Post-Compromise Incident ResponseVotes: 0GitHub stars: 3
- Self Modifying Lean Proof Agents Verifier Grounded Benchmark CoevolutionSkill derived from arXiv:2607.17352 - Self-Modifying Lean Proof Agents with Verifier-Grounded Benchmark CoevolutionVotes: 0GitHub stars: 3
- Setoka A Benchmark For Hierarchical User UnderstanSetoka: A Benchmark for Hierarchical User Understanding in Personalized Agents over Heterogeneous DaVotes: 0GitHub stars: 3
- Snr Sample Size Representational Alignment信噪比和样本数量调控神经网络表征对齐的方法论。研究神经网络潜在表征的通用性规律，揭示对齐与数据质量和数量的非平凡依赖关系。适用于表征对齐分析、神经网络可解释性、训练优化。触发词：表征对齐、SNR、样本数量、插值阈值、通用表征。Votes: 0GitHub stars: 3
- Solarchain Eval A Physics Constrained Benchmark For TrustworthySolarChain-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 trustworthiness. In decentralized energy markets, auton... Activation: agent, agentic, llm, benchmark, safetyVotes: 0GitHub stars: 3
- Speaker Fuzzy Fingerprints Benchmarking Textbased Identification In Multiparty Dialogues**arXiv ID:** 2504.14963 **Authors:** Rui Ribeiro, Luísa Coheur, Joao P. Carvalho **Published:** 2025-04-21T08:44:33Z **Abstract:** Speaker identification using voice recordings leverages unique acoustic features, but this approach fails when only textual data is available. Few approaches have attempted to tackle the problem of identifying speakers solely from text, and the existing ones have primarily relied on traditional methods. In this work, we explore the use of fuzzy fingerprints from ...Votes: 0GitHub stars: 3
- Spikeprophecy BenchmarkSpikeProphecy: First large-scale benchmark for causal, autoregressive neural population spike-count forecasting. Introduces population metric decomposition (temporal fidelity, spatial pattern accuracy, magnitude-invariant alignment) on 105 Neuropixels sessions (~89,800 neurons). arXiv:2605.12992Votes: 0GitHub stars: 3
- Stochastic Configuration Machines Fpga Implementation**arXiv ID:** 2310.19225 **Authors:** Matthew J. Felicetti, Dianhui Wang **Published:** 2023-10-30T02:04:20Z **Abstract:** Neural networks for industrial applications generally have additional constraints such as response speed, memory size and power usage. Randomized learners can address some of these issues. However, hardware solutions can provide better resource reduction whilst maintaining the model's performance. Stochastic configuration networks (SCNs) are a prime choice in industrial a...Votes: 0GitHub stars: 3
- Structure Aware Coreset Fc BenchmarkingAccelerating benchmarking of functional connectivity (FC) modeling via structure-aware coreset selection for large-scale fMRI datasets. Reduces combinatorial explosion in model-data evaluation pairs. Activation: functional connectivity, fMRI benchmarking, coreset selection, FC modeling, brain network analysis, connectomics benchmarking.Votes: 0GitHub stars: 3
- Structured Sparse Autoencoders Cross Modal ConceptsStructured 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 alignment on Qwen2.5-VL-7B. Use when working with sparse-autoencoder, mechanistic-interpretability, concept-consistency.Votes: 0GitHub stars: 3
- Supervised Feature Selection With Neuron Evolution In Sparse Neural Networks**arXiv ID:** 2303.07200 **Authors:** Zahra Atashgahi, Xuhao Zhang, Neil Kichler, Shiwei Liu, Lu Yin, Mykola Pechenizkiy, Raymond Veldhuis, Decebal Constantin Mocanu **Published:** 2023-03-10T17:09:55Z **Abstract:** Feature selection that selects an informative subset of variables from data not only enhances the model interpretability and performance but also alleviates the resource demands. Recently, there has been growing attention on feature selection using neural networks. However, existi...Votes: 0GitHub stars: 3
- Teaching Claude Why Alignment TrainingAlignment training methodology that teaches models to explain their reasoning rather than just correct actions. Uses "difficult advice" dataset and constitution training for robust alignment.Votes: 0GitHub stars: 3
- Teaching Claude Why AlignmentAlignment training methodology teaching models to explain their reasoning rather than just correct actions. Demonstrates 28x efficiency improvement through out-of-distribution training and constitution-based reasoning.Votes: 0GitHub stars: 3
- The Mythos Of Model Interpretability**arXiv ID:** 1606.03490 **Authors:** Zachary C. Lipton **Published:** 2016-06-10T21:28:47Z **Abstract:** Supervised machine learning models boast remarkable predictive capabilities. But can you trust your model? Will it work in deployment? What else can it tell you about the world? We want models to be not only good, but interpretable. And yet the task of interpretation appears underspecified. Papers provide diverse and sometimes non-overlapping motivations for interpretability, and offer my...Votes: 0GitHub stars: 3
- The Standard Interpretable Model A General Theory Of Interpretable Machine Learning To Deductively Design Interpretable Methods Using Lagrangian Mechanics**arXiv ID:** 2606.12289 **Authors:** Pietro Barbiero, Giovanni De Felice, Mateo Espinosa Zarlenga, Francesco Giannini, Filippo Bonchi, Mateja Jamnik, Giuseppe Marra, Ruggero Noris **Published:** 2026-06-10T16:26:22Z **Abstract:** As Artificial Intelligence models grow in complexity, interpretability has become an indispensable tool for understanding, debugging, and controlling their computations. However, interpretability lacks general theories to deductively design interpretable methods. Th...Votes: 0GitHub stars: 3
- Topic Modelling Black Box Optimization**arXiv ID:** 2512.16445 **Authors:** Roman Akramov, Artem Khamatullin, Svetlana Glazyrina, Maksim Kryzhanovskiy, Roman Ischenko **Published:** 2025-12-18T12:00:24Z **Abstract:** Choosing the number of topics $T$ in Latent Dirichlet Allocation (LDA) is a key design decision that strongly affects both the statistical fit and interpretability of topic models. In this work, we formulate the selection of $T$ as a discrete black-box optimization problem, where each function evaluation corresponds ...Votes: 0GitHub stars: 3
- Tour A Trajectory Level Unlearning Benchmark For OfflineTOUR: A Trajectory-Level Unlearning Benchmark for Offline Reinforcement LearningVotes: 0GitHub stars: 3
- Uniclawbench Proactive Agents Real World TasksCapability-driven benchmark for evaluating proactive agents in dynamic real-world settings. UniClawBench evaluates five foundational capabilities (Skill Usage, Exploration, Long-Context Reasoning, Multimodal Understanding, Cross-Platform Coordination) across 400 bilingual tasks in live Docker containers with closed-loop evaluation. Activation: proactive agents, real-world benchmark, agent evaluation, capability-driven, multimodal agents, closed-loop evaluation.Votes: 0GitHub stars: 3
- Unified Neural Scaling LawsUnified Neural Scaling Laws (UNSL) methodology for modeling and extrapolating deep neural network scaling behaviors across multiple dimensions (parameters, data size, compute, hyperparameters). Use when analyzing or predicting model performance scaling, optimizing resource allocation across dimensions, or extrapolating training/inference costs for large models. Applicable to vision, language, math, and RL tasks.Votes: 0GitHub stars: 3