All authors
Claude Skills by hiyenwong
github.com/hiyenwong9,934 skills5 installs19,223 views
- Stress Testing Concept Erasure Large Language Model AgentsSkill derived from arXiv:2607.17890 - Stress Testing Concept Erasure with Large Language Model AgentsVotes: 0GitHub stars: 3
- Stress Testing Concept Erasure With Large LanguageDerived from arXiv:2607.17890 - Stress Testing Concept Erasure with Large Language Model AgentsVotes: 0GitHub stars: 3
- Structure Development In Listsorting Transformers**arXiv ID:** 2501.18666 **Authors:** Einar Urdshals, Jasmina Urdshals **Published:** 2025-01-30T15:56:25Z **Abstract:** We study how a one-layer attention-only transformer develops relevant structures while learning to sort lists of numbers. At the end of training, the model organizes its attention heads in two main modes that we refer to as vocabulary-splitting and copy-suppression. Both represent simpler modes than having multiple heads handle overlapping ranges of numbers. Interestingly, ...Votes: 0GitHub stars: 3
- Super Weights Llms Selective Training FailureShows 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 training, parameter pruning, critical parameters, selective training.Votes: 0GitHub stars: 3
- Symbolic Discovery Of Optimization Algorithms**arXiv ID:** 2302.06675 **Authors:** Xiangning Chen, Chen Liang, Da Huang, Esteban Real, Kaiyuan Wang, Yao Liu, Hieu Pham, Xuanyi Dong, Thang Luong, Cho-Jui Hsieh, Yifeng Lu, Quoc V. Le **Published:** 2023-02-13T20:27:30Z **Abstract:** We present a method to formulate algorithm discovery as program search, and apply it to discover optimization algorithms for deep neural network training. We leverage efficient search techniques to explore an infinite and sparse program space. To bridge the la...Votes: 0GitHub stars: 3
- Teaching Llms To Explain WhyAlignment training methodology that focuses on teaching models to explain their reasoning rather than just taking correct actions, based on Anthropic's research.Votes: 0GitHub stars: 3
- Teaching Llms To Self EvolveFramework for teaching LLMs to self-evolve by cultivating core meta-skills with reinforcement learning, enabling autonomous capability expansion through iterative self-improvement cycles.Votes: 0GitHub stars: 3
- Text And Code Embeddings By Contrastive Pre TrainiSkill for AI agent capabilitiesVotes: 0GitHub stars: 3
- Tf Engram A Train Free Engram With Ssd Backed Memory For Large Language ModelsLarge Language Models (LLMs) store factual knowledge and domain-specific patterns implicitly in dense Transformer parameters, making knowledge expansion costly through pretraining, fine-tuning, retrie. Based on arXiv:2607.07388.Votes: 0GitHub stars: 3
- The Instruction Hierarchy Training Llms To PrioritSkill for AI agent capabilitiesVotes: 0GitHub stars: 3
- The Key To Going Linear Analysis Driven Transformer LinearizationThe quadratic cost of causal self-attention severely bottlenecks long-context transformer inference. While numerous post hoc linearization pipelines exist, it is difficult to identify which components. Based on arXiv:2607.07706.Votes: 0GitHub stars: 3
- Tokenize Once Recommend Anywhere Unified Item Tokenization For Multidomain Llmbased Recommendation**arXiv ID:** 2511.12922 **Authors:** Yu Hou, Won-Yong Shin **Published:** 2025-11-17T03:18:04Z **Abstract:** Large language model (LLM)-based recommender systems have achieved high-quality performance by bridging the discrepancy between the item space and the language space through item tokenization. However, existing item tokenization methods typically require training separate models for each item domain, limiting generalization. Moreover, the diverse distributions and semantics across ite...Votes: 0GitHub stars: 3
- Towards Foundation Models For Consensus Rank Aggregation**arXiv ID:** 2603.15218 **Authors:** Yijun Jin, Simon Klüttermann, Chiara Balestra, Emmanuel Müller **Published:** 2026-03-16T12:55:54Z **Abstract:** Aggregating a consensus ranking from multiple input rankings is a fundamental problem with applications in recommendation systems, search engines, job recruitment, and elections. Despite decades of research in consensus ranking aggregation, minimizing the Kemeny distance remains computationally intractable. Specifically, determining an optimal ...Votes: 0GitHub stars: 3
- Tracing Agentic Failure From The Flow Of SuccessTracing 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...Votes: 0GitHub stars: 3
- Training Free Relaxed Speculative DecodingPractical 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 controlled capability-speed trade-offs. Activation: speculative decoding, LLM inference acceleration, lossy speculation, training-free, draft verification.Votes: 0GitHub stars: 3
- Training Free Relaxed Speculative DeodingA Practical Investigation of Training-free Relaxed Speculative DecodingVotes: 0GitHub stars: 3
- Transformer Guided Swarm Intelligence For Frugal NDerived from arXiv:2607.11826 - Transformer-Guided Swarm Intelligence for Frugal Neural Architecture SearchVotes: 0GitHub stars: 3
- Transformer Revolution Sidpp Dynamic ProcessingFramework for interpreting Transformers as Sequence-level Interactive Dynamic Parallel Processing (SIDPP) systems that construct prompt-dependent transformations during inference, with potential neural correlates in cerebral cortex processing.Votes: 0GitHub stars: 3
- Transformerguided Swarm Intelligence For Frugal Neural Architecture Search**arXiv ID:** 2607.11826 **Authors:** Romain Amigon **Published:** 2026-07-13T17:18:24Z **Abstract:** Neural Architecture Search (NAS) has automated the design of deep learning models but traditionally requires massive computational resources, often measured in thousands of GPU-days. In this paper, we propose a frugal and memetic NAS framework designed to democratize architecture design on consumer-grade hardware. Our approach combines the global macro-search capabilities of an autoregressive...Votes: 0GitHub stars: 3
- Travel Oriented Reasoning Large Language Model ViaDerived from arXiv:2606.29254 - Travel-Oriented Reasoning Large Language Model via Domain-Specific Knowledge GraphsVotes: 0GitHub stars: 3
- Trl Fine TuningFine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.Votes: 0GitHub stars: 3
- Ulps Uncertainty Aware Llm Policy ShapingUncertainty-Aware LLM-Guided Policy Shaping for sparse-reward RL. Integrates calibrated LLM into RL training loop with uncertainty-modulated behavioral guidance.Votes: 0GitHub stars: 3
- Ultrax Refining Pre Training Data At Scale WithUltraX: 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 data expansion and mo... Activation: llm, alignment, control, optimization, promptVotes: 0GitHub stars: 3
- Uncertainty Aware Llm Guided Policy ShapingUncertainty-aware LLM-guided policy shaping for sparse-reward RL using A* oracle trajectories and entropy-based blending with MC dropout uncertainty estimation.Votes: 0GitHub stars: 3
- Understanding Textual Emotion Through Emoji Prediction**arXiv ID:** 2508.10222 **Authors:** Ethan Gordon, Nishank Kuppa, Rigved Tummala, Sriram Anasuri **Published:** 2025-08-13T22:17:00Z **Abstract:** This project explores emoji prediction from short text sequences using four deep learning architectures: a feed-forward network, CNN, transformer, and BERT. Using the TweetEval dataset, we address class imbalance through focal loss and regularization techniques. Results show BERT achieves the highest overall performance due to its pre-training adv...Votes: 0GitHub stars: 3
- Urbands A Graph Guided Llm Multi Agent System ForUrbanDS: A Graph-Guided LLM Multi-Agent System for Data-Intensive Urban TasksVotes: 0GitHub stars: 3
- Value Iteration Networks**arXiv ID:** 1602.02867 **Authors:** Aviv Tamar, Yi Wu, Garrett Thomas, Sergey Levine, Pieter Abbeel **Published:** 2016-02-09T05:44:36Z **Abstract:** We introduce the value iteration network (VIN): a fully differentiable neural network with a `planning module' embedded within. VINs can learn to plan, and are suitable for predicting outcomes that involve planning-based reasoning, such as policies for reinforcement learning. Key to our approach is a novel differentiable approximation of the v...Votes: 0GitHub stars: 3
- Veragrid Agent Tool Augmented Llm Power FlowTool-augmented LLM methodology for solving distribution optimal power flow problems by integrating with numerical solvers like VeraGrid.Votes: 0GitHub stars: 3
- Verify Repair Repeat Stop Robust Stopping Noisy Verify Repair Loops Llm AgentsSkill derived from arXiv:2607.17641 - Verify, Repair, Repeat, or Stop? Robust Stopping for Noisy Verify-Repair Loops in LLM AgentsVotes: 0GitHub stars: 3
- VllmServes LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency/throughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints, quantization (GPTQ/AWQ/FP8), and tensor parallelism.Votes: 0GitHub stars: 3
- We Break Llms Self Loops Fine Grained Reasoning Control Activation SteeringSkill derived from arXiv:2607.18100 - Can We Break LLMs Out of Self-Loops? Fine-Grained Reasoning Control with Activation SteeringVotes: 0GitHub stars: 3
- Webgpt Improving The Factual Accuracy Of LanguageSkill for AI agent capabilitiesVotes: 0GitHub stars: 3
- When Llms Develop Languages Symbolic CommunicationDerived from arXiv:2606.29354 - When LLMs Develop Languages: Symbolic Communication for Efficient Multi-Agent ReasoningVotes: 0GitHub stars: 3
- When Llms Over Answer Measuring And Mitigating QuaDerived from arXiv:2607.17063 - When LLMs Over-Answer: Measuring and Mitigating Quality Issues in LLM-Based Hardware Description Language Question AnsweringVotes: 0GitHub stars: 3
- Who Grades The Grader Co Evolving Evaluation MetriWho 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...Votes: 0GitHub stars: 3
- Why Language Models HallucinateSkill for AI agent capabilitiesVotes: 0GitHub stars: 3
- Win By Silence Deletion Non Monotonicity AutonomouWin 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 LLM-gene...Votes: 0GitHub stars: 3
- Workflow As Knowledge Semantic Persistence For Llm MediatedWorkflow 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 address many execut... Activation: llm, retrieval, control, tool use, policyVotes: 0GitHub stars: 3
- Worldcuparena Fine Grained Evaluation Language Models Deep Research Agents Football ForecSkill derived from arXiv:2607.18084 - WorldCupArena: Fine-Grained Evaluation of Language Models and Deep-Research Agents on Football ForecVotes: 0GitHub stars: 3
- Zero Shot Kinematic Bci DecodingZero-shot handwriting BCI decoding methodology via conserved kinematic representations. Aligns intracortical neural activity to imagined kinematics for open-vocabulary character decoding without per-character training.Votes: 0GitHub stars: 3
- Nn Quantum State EncodingNeural network encoding methodology for quantum state preparation: trains classical neural network to map input data directly to quantum circuit parameters, avoiding per-instance variational optimization. Achieves 0.992 fidelity on unseen data with 5000x runtime reduction. Use when designing QML data loading pipelines, quantum state preparation, neural-encoded quantum circuits, or amplitude encoding optimization.Votes: 0GitHub stars: 3
- Noise Directed Adaptive RemappingNoise-directed adaptive remapping methodology for integer optimization — encoding qubit-based problems into qudit representations with noise-aware adaptation. Use when optimizing quantum integer optimization on NISQ hardware, converting qubit encodings to qudit representations, mitigating hardware noise through adaptive remapping, or solving scheduling/resource allocation problems with quantum qudit systems.Votes: 0GitHub stars: 3
- Noise Enhanced Quantum KernelsNoise-enhanced quantum kernel methods for analog quantum computing. Implements analog and hybrid quantum kernels with noise-induced performance improvements for quantum machine learning. Activation: noise quantum kernel, analog quantum kernel, quantum kernel noiseVotes: 0GitHub stars: 3
- Non Hermitian Conscious Preconscious SubliminalNon-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 consciousness, GNW landscape, conscious preconscious subliminal, potential well consciousness, complex-valued GNW, non-Hermitian Hamiltonian consciousness.Votes: 0GitHub stars: 3
- Non Hermitian Gnw ConsciousnessNon-Hermitian Potential Well Formalism for modeling the Global Neuronal Workspace (GNW) consciousness framework. Uses nonlinear Schrödinger-type equation in imaginary time with non-Hermitian, non-normal Hamiltonian and Lotka-Volterra-type term to reproduce subliminal-preconscious-conscious hierarchy. Maps conscious access to bound-state emergence in complex-valued landscape. Use for consciousness modeling, neural field theory, GNW theory, and quantum-inspired cognitive dynamics.Votes: 0GitHub stars: 3
- Non Markovian Kerr Feedback QrcProves unbounded computational superiority of Kerr nonlinear feedback over Gaussian linear reservoirs in continuous-variable quantum reservoir computing. Single Kerr mode with feedback depth D replaces up to ~100 linear modes. Use when: CV-QRC design, non-Markovian reservoir computing, Kerr nonlinear optics, cross-time nonlinear correlations, quantum reservoir capacity analysis, Gaussian limitations, time-delay feedback.Votes: 0GitHub stars: 3
- Non Unitary Qml Fisher EfficiencyNon-unitary quantum machine learning via Linear Combination of Unitaries (LCU) framework, with Fisher efficiency transitions and threshold-dependent parameter scaling in medical imaging tasks. Use when: implementing non-unitary quantum layers, benchmarking quantum vs classical performance across domains, analyzing Fisher information efficiency in QML, or designing quantum circuits for medical image classification.Votes: 0GitHub stars: 3
- Nonlinear Mas Optimal ControlNonlinear Multi-Agent Systems Optimal Control - 非线性多智能体系统分布式最优控制。核心技术：HJB方程分布式近似、私有信息结构、保密协作控制。激活词：MAS optimal control, multi-agent control, 非线性最优控制, HJB distributed.Votes: 0GitHub stars: 3
- Nuclear Lattice VqeVariational Quantum Eigensolver (VQE) framework for nuclear lattice effective field theory. Computes ground state energies of light nuclei (2H, 3H, 4He) using Gray code encoding with symmetry reduction for compact qubit representation. Keywords: nuclear physics, VQE, variational quantum eigensolver, nuclear lattice EFT, Gray code encoding, Jordan-Wigner, light nuclei, deuterium, tritium, helium-4.Votes: 0GitHub stars: 3
- Off Switch Dual Use GramGradient-Routed Auxiliary Modules (GRAM) methodology from Anthropic/AE Studio research (Jul 8, 2026) — training a single LLM with removable, category-specific knowledge compartments that can be toggled on/off post-training without retraining, enabling surgical dual-use knowledge control.Votes: 0GitHub stars: 3