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.
Hybrid Quantum Neighborhood Selection (HQNS) — resource-efficient framework for large-scale combinatorial optimization. Decomposes dense QUBO into bounded-width quantum subproblems via stochastic frontier selection. Preserves 99.99% solution quality while…
Quantum Koopman method — data-driven framework for simulating nonlinear dynamics on quantum computers. Embeds nonlinear dynamics into learned linear representation via Koopman observables, then implements evolution using shallow quantum circuits with parallel…
Universal purification dynamics of monitored Clifford circuits methodology — shows purification reduces to exactly solvable Markovian death process, bypassing replica trick. Computes universal scaling functions for all Renyi entropies.
Dual-level trainable variational quantum solver methodology for interior-point optimal power flow (OPF) — uses early solver-generated trajectories to accelerate convergence, reducing variational updates by up to 95%. Combines VQLS parameter trajectory…
Pauli-structured preconditioning methodology for quantum linear system solvers. Based on arXiv:2606.01733 — reduces normalization overhead via Pauli expansion regrouping.
Qolumbina benchmark infrastructure for controlled Quantum Software Testing (QST) experiments on scalable quantum programs — curates 40 programs from open-source repos with systematic selection, refactoring, specifications, and standardized interfaces.
Quantum probability framework for understanding confirmation bias as optimal evidence selection - square-root probability spaces, matrix-valued observations, and evolutionary advantages in sequential hypothesis testing.
Quantum linear system solvers with complexity independent of the condition number, using block encoding input model for Ax=b solutions.
Quantum computing methodology for minimax estimation of high-order functionals (Rényi/Tsallis entropy) — achieves optimal sample complexity O(α) vs classical O(α²)
Methodology for analyzing sample complexity in quantum PAC-learning models where concepts are functions acting on quantum states.
Quantum probabilistic local differential privacy methodology - structural properties, sample complexity bounds, and hypothesis testing applications for privacy-preserving quantum information processing.
Compiler-driven sub-microsecond feedback control stack for scalable trapped-ion quantum experiments. Covers Python-embedded DSL, six-stage transpilation pipeline (CFG, SSA, liveness, register allocation), cross-board synchronization with <700ns latency, and…
Feature- and learning-guided quantum compiler pass tuning methodology — moves beyond static circuit features to capture how circuits react to compiler optimizations, enabling more effective pass sequence selection.
RISC-V vector extension architecture for scalable quantum control processors (QCP) with 128-qubit single-instruction addressing, halt-resume mid-circuit measurement protocol, and parameterized rotation support.
Tilted XOR games methodology — variant of XOR nonlocal games where winning condition depends on XOR plus one output bit. Shows quantum value approximation is RE-complete, dramatically increasing expressive power over standard XOR games.
Pattern formation in multimode open quantum systems via GKSL master equation — extends Turing instabilities and mode competition to dissipative quantum systems with parametric driving and nonlinear damping.
Physics-informed LLM framework for general quantum control combining symbolic reasoning with optimization. Covers analytic control ansätze generation, parameter refinement through feedback, QCTRL-BENCH benchmark design, inference-time scaling, and pulse…
Von Neumann algebra framework for controllability of bilinear systems on infinite-dimensional Hilbert spaces. Uses operator affiliation theory to prove existence of time-optimal controls and define dynamical Lie algebras for unbounded operators. Applicable to…
AI 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.
Hypergraph neural networks have shown powerful capability in modeling higher-order relations, yet their predictive uncertainty remains underexplored. Unlike pairwise graphs, uncertainty in hypergraphs. Based on arXiv:2607.07330.
The inverse design of physical systems governed by partial differential equations is computationally demanding due to the high dimensionality and non-convexity of design spaces. Generative models for. Based on arXiv:2607.07682.
Agents acting on our behalf in the real world (e.g. placing phone calls) must learn online from costly, often irreversible interactions rather than cheap simulator steps. Two things follow. First, dep. Based on arXiv:2607.07435.
Long-horizon LLM multi-agent systems face reliability risks invisible to infrastructure monitoring. We propose the ADE Predictive Reliability Framework (ADE-PRF), enabling proactive health trajectory. Based on arXiv:2607.07689.
Autonomous agents promise substantial gains in speed, scale, and labor efficiency, but their failures can impose abrupt and often irreversible costs. The central challenge for agentic automation is th. Based on arXiv:2607.07397.
Product data scientists often ask LLM-based agents to help with recurring execution tasks such as cleaning data, writing SQL, choosing statistical tests, and formatting results. Reusable skill files a. Based on arXiv:2607.07504.
Tool utilization enables Large Language Model (LLM) agents to interact with the real world and resolve complex tasks. However, existing agent frameworks predominantly rely on static toolsets composed. Based on arXiv:2607.07321.
The optimization of long-horizon agents increasingly relies on reflection-based mechanisms, where a large language model (LLM) acts as an optimizer to diagnose agent failures and improve agent policie. Based on arXiv:2607.07702.
In agentic scientific machine learning (SciML), large language model (LLM) agents can discover surrogate models and select one by an automated score, typically an error metric. A low error, however, d. Based on arXiv:2607.07379.
Reinforcement learning from verifiable rewards (e.g. GRPO) is the engine behind todays reasoning models, yet it grades only the final answer. On hard problems this trains models to write more rather. Based on arXiv:2607.07690.
Limited memory language models (LMLMs) externalize factual knowledge during pretraining to a knowledge base (KB), rather than memorizing it in their weights. During generation, the model then fetches. Based on arXiv:2607.07707.
Speculative decoding accelerates LLM inference by drafting multiple tokens and verifying them in parallel. Block-parallel drafters such as DFlash further improve drafting efficiency by predicting an e. Based on arXiv:2607.07409.
Reliable confidence estimation is essential for deploying large language models (LLMs) in confidence-aware systems, where downstream decisions such as retrieval, tool use, and adaptive computation dep. Based on arXiv:2607.07626.
One-shot pruning methods like Wanda and SparseGPT apply the same sparsity ratio to every layer of a transformer, ignoring known variation in layer importance. We propose PALS (Percentile-Aware Layerwi. Based on arXiv:2607.07557.
Rigorous content moderation is crucial for online advertising but leads to millions of daily rejections. This scale renders manual rectification infeasible, particularly for video advertisements. Howe. Based on arXiv:2607.07318.
Does RL post-training merely amplify primitive skills already latent in a base model, or can it compose primitive skills into new higher-level strategies? We study this question in a fully observable. Based on arXiv:2607.07646.
Many reasoning tasks are not well described by a single left-to-right chain: a solver may need to pursue a plausible branch, observe delayed failure, and return to the latest prefix that can still be. Based on arXiv:2607.07492.
Linear attention models allow a fixed state size and a fixed amount of compute per token. However, due to their limited state size, linear attention models fall behind in long-context recall compared. Based on arXiv:2607.07386.
The 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.
We address the problem of efficiently sampling multimodal probability distributions, where standard Markov Chain Monte Carlo methods often suffer from poor mixing and mode trapping. To mitigate these. Based on arXiv:2607.07519.
Nonlinear least-squares optimization is central to regression, physics-informed neural networks, and other machine-learning tasks. Such problems have a natural geometric interpretation, model predicti. Based on arXiv:2607.07623.