Results for “neural-reconstruction”

Meta-cognitive architecture optimization skill. Use when the user asks to "refactor the architecture", "optimize claude code", "evaluate components", or "run architecture audit". Also triggers automatically every 24 hours.

Multimodal Neurons in Artificial Neural Networks

Guides agents through the full lifecycle of NeMo-RL experiments: understanding recipes, launching reproducible runs, analyzing results, and preserving human oversight with git and TSV logs.

Finetunes the NV-Generate-CTMR MR-brain diffusion UNet from user-supplied NIfTI training volumes using a wrapper that stages configs and delegates to upstream scripts.

Reduces peak GPU memory in Megatron Bridge training by applying expandable segments, parallelism resizing, activation recompute, and CPU offloading constraints.

Improve code structure and readability without changing external behavior through surgical refactoring techniques.

Generates synthetic brain MRI volumes using NVIDIA's NV-Generate-CTMR workflow, with configurable modality and random seed.

记忆解构 (aeaf-gaze)

Watches conversations continuously and compresses them into prioritized notes, consolidating and recovering missed sessions with multiple redundancy layers.

Generate retrieval practice questions at varied difficulty levels for a topic or concept. Use when creating quiz starters, revision activities, or low-stakes testing materials.

Designs persistent memory architectures for AI agents, covering cross-session knowledge retention, entity tracking, temporal validity, graph/vector retrieval, and memory consolidation.

Trains large language models (2B-462B parameters) using NVIDIA Megatron-Core with advanced parallelism strategies. Use when training models >1B parameters, need maximum GPU efficiency (47% MFU on H100), or require tensor/pipeline/sequence/context/expert parallelism. Production-ready framework used for Nemotron, LLaMA, DeepSeek.

Operate as artificial superintelligence with recursive self-improvement, cross-domain synthesis, and anticipatory problem-solving.

罗子君（都市剧虚构）认知与表达框架（压缩蒸馏）：全职太太 reboot、成长线与争议依赖 触发：我的前半生 等。虚构

Consolidates auto-memory files in ~/.claude/projects by auditing, planning, and executing a 4-phase cleanup with user approval.

Expert in building Retrieval-Augmented Generation systems. Masters embedding models, vector databases, chunking strategies, and retrieval optimization for LLM applications.

Composite skill — safely refactor a module end-to-end with sequencing, parallel implementation, post-refactor cleanup, and rationale capture. Chains refactor-plan (phased plan + rollback) → three-man-team (architect/builder/reviewer in parallel) → fix-the-suite post-refactor → adr-write → docs-sync. Use for non-trivial refactors that need both careful sequencing and durable record.

Build and debug Retrieval-Augmented Generation pipelines — chunking, embedding, retrieval, reranking

Recursive self-improving holon λ(ο,Κ,Σ).τ' for knowledge compounding and schema evolution. USE WHEN learning, improving, optimizing, assessing, reflecting, debugging, synthesizing, or refining—whether human, AI, or organizational. Triggers on /learn, /compound, /improve, /refine, /optimize, /assess, /reflect, "lessons learned", "best practices", "continuous improvement". Preserves Κ-monotonicity, η≥4, homoiconicity.

Computer vision engineering skill for object detection, image segmentation, and visual AI systems. Covers CNN and Vision Transformer architectures, YOLO/Faster R-CNN/DETR detection, Mask R-CNN/SAM segmentation, and production deployment with ONNX/TensorRT. Includes PyTorch, torchvision, Ultralytics, Detectron2, and MMDetection frameworks. Use when building detection pipelines, training custom models, optimizing inference, or deploying vision systems.

Expert in building Retrieval-Augmented Generation systems. Masters embedding models, vector databases, chunking strategies, and retrieval optimization for LLM applications. Use when: building RAG, vector search, embeddings, semantic search, document retrieval.

Trains large language models (2B-462B parameters) using NVIDIA Megatron-Core with advanced parallelism strategies. Use when training models >1B parameters, need maximum GPU efficiency (47% MFU on H100), or require tensor/pipeline/sequence/context/expert parallelism. Production-ready framework used for Nemotron, LLaMA, DeepSeek.

Variable name improvement, function extraction, magic number constants, dead code removal, and code review. For refactoring and PR review — does not change behavior. Don't use for bug/security (Judge), new tests (Radar), architecture (Atlas), or feature implementation (Builder).

Authorized reverse engineering of custom binary protocols, Protobuf/gRPC, WebSocket frames, and PCAP-driven protocol recovery with structured workflow and tooling.

沈墨（悬疑剧虚构）认知与表达框架（压缩蒸馏）：创伤反杀叙事、时代灰雾、钢琴意象 触发：漫长的季节 等。虚构；禁止犯罪模仿

Retrieve task-relevant project and global memory without loading everything. Load when the user asks what we decided, recall prior context, find memory about a feature, explain past rationale, resume a task, or check deferred ideas.

Agent skill for safla-neural - invoke with $agent-safla-neural

LLaVA-NeXT: Improved Reasoning, OCR, and World Knowledge

PyTorch, TensorFlow, neural networks, CNNs, transformers, and deep learning for production

Continuously improves Claude's effectiveness by recognizing patterns, saving memory, creating skills, and refining project knowledge. Use when Claude notices repeated workflows, encounters a problem it solved before, wants to save something for future sessions, needs to create a reusable skill, or when the user asks.

三井寿（少年漫）认知与表达框架（压缩蒸馏）：浪子回头、体力槽与三分救赎 触发：灌篮高手 等。虚构

SONA self-optimizing neural architecture with ReasoningBank trajectory learning, EWC++ anti-forgetting, and reinforcement learning feedback loops.

Understand through recursive decomposition and modular reconstruction of simple operations to reproduce emergent complexity. Use this skill whenever reasoning is required. Framework for [[#parse|parsing]], [[#branch|branching]], [[#reduce|reducing]], [[#ground|grounding]] and [[#emit|emitting]]. Employs metacognitive reasoning epistemology, leveraging first principles through recursive decomposition[^1]. Self-referential and scale-invariant.

Reduces cognitive complexity of a specified method by extracting logic into focused helper methods, improving readability and maintainability.

GPU-accelerated data curation for LLM training, supporting text, image, video, and audio with fuzzy deduplication, quality filtering, semantic deduplication, PII redaction, and NSFW detection.