All authors
Claude Skills by hiyenwong
github.com/hiyenwong9,934 skills5 installs19,223 views
- Arxiv 2509 25084 Scaling Generalist Data Analytic AgentsData-analytic agents are emerging as a key catalyst for automated scientific discovery. This paper introduces DataMind, a scalable data synthesis and agent training recipe designed to build generalist data-analytic agents. DataMind tackles three key challenges including insufficient data resources, improper training strategy, and unstable code-based multi-turn rollout. Built on DataMind, DataMind-14B achieves state-of-the-art with an average score of 71.16% on multiple data analysis benchmarks, Votes: 0GitHub stars: 3
- Arxiv 2509 26596 Safety Contract Graph Multi Agent Reinforcement Learning ForSafety-Contract Graph MARL framework for network security applications. Combines graph neural networks with multi-agent reinforcement learning under safety constraints for cyber-physical system protection.Votes: 0GitHub stars: 3
- Arxiv 2509 26628 Attention As A Compass Efficient Exploration For Process SupWe introduce AttnRL, a novel PSRL framework which enables efficient exploration for reasoning models. Motivated by observations that steps exhibiting high attention scores correlate with reasoning behaviors, we propose to branch from positions with high values. We develop an adaptive sampling strategy and a one-step off-policy training pipeline. Extensive experiments on mathematical reasoning benchmarks demonstrate consistent improvements.Votes: 0GitHub stars: 3
- Arxiv 2609 30147 Grasp Strategic Planning Agentic AiResearch paper: GRASP: Generating, Revising, and Assessing for Strategic Planning with Agentic AI.Votes: 0GitHub stars: 3
- Arxiv 2609 31590 Agentworld Benchmark Long Horizon CollaborationResearch paper: AgentWorld: Benchmarking Long-Horizon Collaboration of Multi-agent LLMs.Votes: 0GitHub stars: 3
- Arxiv 2609 35936 Embodied Semantic Communication For Collective AutEmbodied Semantic Communication for Collective Autonomous Agents: A Tutorial on Representation, Wireless Delivery, and Closed-Loop Coordination (arXiv: 2609.35936)Votes: 0GitHub stars: 3
- Arxiv 2609 37457 Veriweave Govern Evidence Gated Deterministic RuntVeriWeave Govern: Evidence-Gated Deterministic Runtime Governance for Enterprise AI Agents (arXiv: 2609.37457)Votes: 0GitHub stars: 3
- Arxiv 2609 38482 Panda A Decentralized Architecture With Flexible OPANDA: A Decentralized Architecture with Flexible Orchestration for Scalable, Fault-Tolerant Multi-Agent Systems (arXiv: 2609.38482)Votes: 0GitHub stars: 3
- Arxiv 2609 38662 Collabflow Recursive Self Improvement Of Agent ColCollabFlow: Recursive Self-Improvement of Agent Collaboration (arXiv: 2609.38662)Votes: 0GitHub stars: 3
- Arxiv 2609 39045 Rsigame Autonomous Agentic Game Development With RRSIGame: Autonomous Agentic Game Development with Recursive Self-improvement (arXiv: 2609.39045)Votes: 0GitHub stars: 3
- Arxiv 2609 39143 Refcon Iterative Refinement Contrastive MemoryResearch paper: RefCon: Iterative Refinement and Contrastive Memory Extraction for Context-Evolving Agent. Proposes RefCon combining sequential self-refinement with parallel self-contrast to extract higher-quality memories without gold labels. Achieves 21.6% gain on ACE and 16.6% on ReMe, generalizes across model scales and software engineering tasks.Votes: 0GitHub stars: 3
- Arxiv 2610 01161 My Fault Self Diagnosis As Credit AssignmentResearch paper: My FAULT: Self-Diagnosis as Credit Assignment in Self-Evolving Agentic Reinforcement Learning. Proposes FAULT (Self-Diagnosis-guided Terminal Credit Redistribution) that turns diagnosed errors into explicit step-level credit anchored by terminal outcomes, solving credit assignment in long-horizon agentic RL tasks.Votes: 0GitHub stars: 3
- Arxiv 2610 01756 Sok Decentralized Agent Economic InfrastructureResearch paper: SoK: Decentralized Agent Economic Infrastructure. Systematizes security and economic requirements across the full lifecycle of agent tasks, introducing 'guarantee closure' criterion for end-to-end verification. Examines 12 systems, 5 mechanism families, and exposes recurring failures between verification and settlement in decentralized agent economies.Votes: 0GitHub stars: 3
- Jaxolotl A Unified High Performance Benchmark SuiteJaxolotl: A Unified High-Performance Benchmark Suite for LTL-Based Multi-Task RLVotes: 0GitHub stars: 3
- Learning Meta Skills Agent Harness DesignTest-time AI-for-AI where Builder agents learn meta-skills to construct better execution environments for Target agents with fixed weightsVotes: 0GitHub stars: 3
- Precision Gated Mbrl ExcavatorPrecision-gated model-based RL for real robots - probabilistic dynamics ensemble learned from scratch on hardware with sampling-based MPC, progress reward conditioned on path accuracy (precision gates speed). Use for sample-efficient learning on heavy machinery, precision tracking under actuation dynamics, or direct-on-hardware RL without demonstrations.Votes: 0GitHub stars: 3
- Skill Space Shooting Autonomous Robot PolicyAutonomous robot policy improvement through skill-space shooting that enables robots to improve beyond initial training without human demonstrationVotes: 0GitHub stars: 3
- Ablation Response Fidelity Behavioral ModelsUse when validating input ablations of behavioral/cognitive models. Oracle-calibrated response fidelity.Votes: 0GitHub stars: 3
- Adaptive Fractional State Cortical Dynamics分析皮层异常扩散/层级动力学时用。AF态双分数均场理论。Votes: 0GitHub stars: 3
- Brain Alignment Causal Dissociation Attention Heads检验脑对齐头是否因果重要时用。对齐与计算解离。Votes: 0GitHub stars: 3
- Brain Sad Brain Inspired Safe Autonomous DrivingBrain-inspired safe autonomous driving framework with dynamic fear-oriented constraints on dual-policy for constrained reinforcement learningVotes: 0GitHub stars: 3
- Connectome Synapse Flow Certified MixingUse when analyzing connectome random-walk mixing certificates.Votes: 0GitHub stars: 3
- Connectome Wiring Specificity Null ModelsNested rewired nulls test connectome wiring specificity.Votes: 0GitHub stars: 3
- Fc Spectral Flattening Fmri PretrainingUse when recalibrating FC spectra for fMRI prediction.Votes: 0GitHub stars: 3
- High Rank Connectivity Scaffolds RnnUse when analyzing high-rank RNN connectivity scaffolds.Votes: 0GitHub stars: 3
- Identifiability Guarantees Delayed Physical SystemsTheory-grounded method proving identifiability of structural drivers and drift in stochastic delayed physical systems under permissive assumptionsVotes: 0GitHub stars: 3
- Infant Fmri Deep Learning ReviewDeep learning for infant fMRI: representations, prediction, validation.Votes: 0GitHub stars: 3
- Neurodyn Eeg Neural Dynamics PretrainedUse when inverting EEG into neural mass model parameters.Votes: 0GitHub stars: 3
- Rf Constrained Mei Visual CortexUse when synthesizing most-exciting-inputs for early visual cortex voxels.Votes: 0GitHub stars: 3
- Arxiv 2509 25035 Ultra Fast Language Generation Via Discrete Diffusion DivergWe introduce DiDi-Instruct, a training-based method that initializes from a pre-trained diffusion LLM and distills a few-step student for fast generation. The model matches or surpasses its dLLM teacher and GPT-2 baseline while providing up to 64x acceleration. Based on integral KL-divergence minimization with grouped reward normalization and reward-guided ancestral sampler. Achieves perplexity 62.2 (8 NFEs) to 18.4 (128 NFEs) on OpenWebText.Votes: 0GitHub stars: 3
- Arxiv 2509 26626 Recursive Self Aggregation Unlocks Deep Thinking In Large LaWe propose Recursive Self-Aggregation (RSA), a test-time scaling method inspired by evolutionary methods that combines the benefits of both parallel and sequential scaling. Each step of RSA refines a population of candidate reasoning chains through aggregation of subsets. RSA with Gemini 3 Flash attains performance near the top of the ARC-AGI-2 public leaderboard. RSA also enables Qwen3-4B to compete with larger reasoning models including DeepSeek-R1 and o3-mini.Votes: 0GitHub stars: 3
- Conformal Context Trust Decision TransformerTrust Guided Decision Transformer (TGDT) - use the model's own rolling next-state prediction error, calibrated by split conformal prediction on held-out offline data, to filter unreliable conditioning contexts before applying critic value guidance. Use for long-rollout stability in sequence-mode RL, offline RL with context drift, or any autoregressive policy where conditioning context goes out of distribution.Votes: 0GitHub stars: 3
- Gpt Red Self Play Red TeamingUse when building automated LLM red-teaming systems.Votes: 0GitHub stars: 3
- Hilbert Space Interpretability QtbHilbert-space interpretability framework for Quantum Transformer Blocks: track quantum mutual information, entanglement entropy, participation ratio, and inter-layer fidelity through circuit layers to watch a quantum model think. Use when interpreting variational quantum circuits, building intrinsically interpretable QML, or diagnosing quantum model failures.Votes: 0GitHub stars: 3
- Adaptive Coupling Variance Mean FieldWeight-variance mean-field for adaptive oscillator networks.Votes: 0GitHub stars: 3
- Apst Association Profile Frozen BciUse for cross-session BCI decoding with frozen weights. APST.Votes: 0GitHub stars: 3
- Arithmetic Sync Basin Pulse Coupled OscillatorsUse when analyzing pulse-coupled oscillator sync basins by number theoryVotes: 0GitHub stars: 3
- Compositional Invariance LebesgueNetwork safety verification via local if-and-only-if checks.Votes: 0GitHub stars: 3
- Contest Envelope CodesignControl+estimation co-design via dual-variable gradients.Votes: 0GitHub stars: 3
- Context Modulated Emrnn Memory建模PFC上下文调制记忆时用。低秩门控RNN+键值EM情境检索。Votes: 0GitHub stars: 3
- Dfa Common Mode CollapseMean-error-driven representation collapse in DFA training.Votes: 0GitHub stars: 3
- Entangled Hopfield Projector MemoryUse for entangled-state quantum Hopfield memory.Votes: 0GitHub stars: 3
- Frontier Training Safety CasesUse when writing safety cases for frontier RL training runs.Votes: 0GitHub stars: 3
- Ghost R Sic Twisted ConvolutionUse for ghost r-SIC construction via twisted convolution identity.Votes: 0GitHub stars: 3
- Greens Operator Multitask RnnGreen's operator analysis of task reuse in multitask RNNs.Votes: 0GitHub stars: 3
- Grid Field Microstructure Matched Null HarmonicsMatched-null harmonic analysis shows single grid fields lack local sixfold symmetry.Votes: 0GitHub stars: 3
- Hessian Null Space ContinuationUse when traversing mode-connected solution regions via Hessian null space.Votes: 0GitHub stars: 3
- Hyperreservoir Context Dependent ReadoutContext reservoir bilinearly modulates shared reservoir readout.Votes: 0GitHub stars: 3
- Mts Slds Multi Timescale SwitchingUse when inferring regime-specific neural timescales from population recordings. MTS-SLDS method.Votes: 0GitHub stars: 3
- Neural Structural ReasonerBrain-inspired KG reasoning via layered neuron dynamics.Votes: 0GitHub stars: 3