1.3천건 · 관련도순
Fornece orientação para interpretar e manipular internals de redes neurais usando nnsight com execução remota NDIF opcional. Use quando precisar executar experimentos de interpretabilidade em modelos massivos (70B+) sem recursos locais de GPU, ou ao trabalhar com qualquer arquitetura PyTorch.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Fornece orientação para pesquisa de interpretabilidade mecanística usando TransformerLens para inspecionar e manipular internals de transformers via HookPoints e caching de ativações. Use ao fazer engenharia reversa de algoritmos de modelos, estudar padrões de atenção ou realizar experimentos de activation patching.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Mescle múltiplos modelos ajustados usando mergekit para combinar capacidades sem retreinar. Use ao criar modelos especializados misturando expertise específica de domínio (math + coding + chat), melhorando performance além de modelos únicos, ou experimentando rapidamente variantes de modelos. Cobre SLERP, TIES-Merging, DARE, Task Arithmetic, mesclagem linear e estratégias de deploy em produção.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Analyze, design, write, revise, translate, and review empirical or technical research papers through an evidence-bounded dialectical workflow. Use when Codex must develop a research question, integrate reference papers with a user's ideas, build phenomenon-mechanism-conditions reasoning, design methods or experiments, restructure or edit a manuscript, translate Chinese academic text into restrained English, answer reviewers, or run multi-round peer-review convergence without fabricating evidence.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Use when diagnosing agent benchmarks or runtime failures from trajectories instead of aggregate pass rates. Builds shared action-observation decision landscapes, identifies productive cores and trap regions, profiles Access/Trap/Repair behavior, and designs conservative trap-triggered recovery experiments.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Runs an autonomous experiment loop on any programming task: you define the goal, metric, and eval command; the agent edits code, measures, and keeps or discards each change until interrupted. Use when optimizing a measurable target (speed, accuracy, size, score) by iteration.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Autonomous experiment loop that optimizes any file by a measurable metric. Inspired by Karpathy's autoresearch. The agent edits a target file, runs a fixed evaluation, keeps improvements (git commit), discards failures (git reset), and loops indefinitely. Use when: user wants to optimize code speed, reduce bundle/image size, improve test pass rate, optimize prompts, improve content quality (headlines, copy, CTR), or run any measurable improvement loop. Requires: a target file, an evaluation command that outputs a metric, and a git repo
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Data Scientist role within the AID methodology. Activate when working on: EDA, feasibility assessment, experiment planning, ML architecture, model development & validation, or production deployment & monitoring. Stack-agnostic — ask the user for their stack at the start of any new project.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Create, revise, and validate publication-ready academic paper figures and tables. Use for LaTeX tables, related-work comparison tables, result tables, notation/dataset/taxonomy tables, precise source-data-driven experiment plots from CSV/JSON/logs, generated conceptual figures such as system overviews/pipelines/architectures/threat models, captions, artifact specs, source-data traceability, and paper-ready PDF/SVG/PNG/LaTeX exports. Do not use for prose-only paper writing, self-review, reviewer response, rebuttal drafting, or external literature-management workflows.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
The original v1 taste-skill, preserved for projects depending on its exact behavior. The current default is `design-taste-frontend` (v2 experimental), which is a substantial rewrite. Use this v1 install name only if you need exact backward compatibility.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Autonomous experiment loop that optimizes any file by a measurable metric. Inspired by Karpathy's autoresearch. The agent edits a target file, runs a fixed evaluation, keeps improvements (git commit), discards failures (git reset), and loops indefinitely. Use when: user wants to optimize code spe...
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Archive a completed change in the experimental workflow. Use when the user wants to finalize and archive a change after implementation is complete.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Start a new OpenSpec change using the experimental artifact workflow. Use when the user wants to create a new feature, fix, or modification with a structured step-by-step approach.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
W&B: log ML experiments, sweeps, model registry, dashboards.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Use for A/B experiments on OpenAmer configs and skills.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Throwaway experiments to validate an idea before build.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Training Details of the Second RL Stage 4 Experiment 5 Ethics and Safety Statement 6 Conclusion, Limitation, and Future Work 7 Author List A Backgroun
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Our large-scale experiment with 19 LLMs reveals that current models degrade documents during delegation: even frontier models (Gemini 3.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Econometrics skill for Difference-in-Differences (DID) analysis. Activates when the user asks about: "difference in differences", "DID", "DiD", "diff-in-diff", "parallel trends", "treatment group", "control group", "pre-treatment", "post-treatment", "policy evaluation", "natural experiment", "staggered DID", "event study regression", "two-way fixed effects DID", "callaway santanna", "sun and abraham", "双重差分", "倍差法", "平行趋势", "处理组", "对照组", "政策评估", "事件研究", "交错DID", "渐进处理"
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Knowledge base from the DoD Test & Evaluation Enterprise Guidebook (OUSD R&E, August 2022). Use for DoD test and evaluation (T&E) across the Adaptive Acquisition Framework — the four test communities (contractor testing, developmental DT&E, operational OT&E, live-fire LFT&E); the five operational events (Operational Demonstration, EOA, OA, IOT&E, FOT&E) and the OTRR gate; the T&E Strategy / TEMP and its IDSK, with the upstream documents that feed it and the test plans that decompose it; the T&E organizations and oversight roles (USD(R&E), DOT&E, TRMC, the Service Executives and OTAs, the T&E WIPT, CDT); and pathway-specific T&E for all six acquisition pathways (Urgent Capability, Middle Tier, Major Capability, Software, Defense Business Systems). Statutory anchors include 10 U.S.C. §4171 (IOT&E) and §4172 (LFT&E/FUSL). This is guidance, not policy: it points outward to DoDI 5000.89 and the DoDI 5000-series for binding detail, and is thin on detailed test statistics/design-of-experiments methodology, cost estimating, the Acquisition of Services pathway, and any CUI/CAC-gated companion guides.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Query PostHog product analytics via REST API — event analytics, HogQL SQL queries, feature flags, experiments, session recordings, insights, persons, cohorts. Use when the user wants to answer an analytics question, toggle/inspect a flag, or pull event data programmatically without opening the PostHog UI or installing the PostHog MCP.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Use when the user asks to "generate subject line variants", "pre-score my subject lines", or "will this subject get truncated / trigger spam filters"; produces a labeled subject + preheader variant set and a per-variant heuristic pre-score card — spam-trigger flags, length/truncation across desktop + mobile, emoji-count, and the inbox preview render (from-name + subject + preheader) — before any test is run. Not for the body copy or CTA — use email-creative-builder; not for the A/B test design or significance read — use send-experiment-designer; not for the profile-weighted EQS or the S1/S2/N1/D1 vetoes — use email-quality-auditor. 邮件主题行生成/主题行预打分/截断与垃圾词检查
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Design or evaluate pricing and packaging using customer value, willingness-to-pay evidence, competitive context, unit economics and experiment design.
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.
Use when the user asks to "design an email A/B test", "set up a multivariate subject/CTA test", "run a send-time test", "build a hold-out group", or "is this email result statistically and practically material?"; produces a falsifiable hypothesis, one-variable-per-cell matrix, sample-size/MDE/duration/power plan, and an effect/uncertainty read from own ESP data. Applies only a precommitted owner-approved action rule; the helper never chooses a business action. Not for EQS/vetoes or writing the email. 邮件AB测试设计/多变量测试/发送时间测试/留出组/显著性判定
한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.