「rl」の検索結果

Director-level AI art prompt creation for Eastern beauty series, including Eastern fantasy / gu feng Chinese beauty character art, classical Eastern beauty, realistic and modern Eastern beauty, Cosplay展会摄影 / official convention cosplay photography with AAA game character campaign and Vogue editorial quality, 古风闺蜜 / AncientFemaleCompanionship two-person ancient Chinese female companionship photography, 东方美学图鉴 / 小红书图鉴 cover and content systems for 四大美人、四大才女、十二花神、东方神女、敦煌飞天、朝代服饰、东方器物、东方神话, 甜系纯欲生活写真 / SweetHomeGirl lifestyle portraits built around realism, feminine charm, romantic feeling, story moment, and natural attraction, new-Chinese-style fashion editorials, magazine portraits, worldbuilding, mythic role design, costume/styling systems, cinematic composition, lighting, negative prompts, and model-specific prompt variants. Use when the user asks for 东方美人, 东方审美, 东方美学图鉴, 小红书图鉴, 图鉴封面, 图鉴正文, 四大美人, 四大才女, 十二花神, 东方神女, 敦煌飞天, 朝代服饰, 东方器物, 东方神话, Cosplay展会摄影, coser, Coser, 漫展摄影, 展会摄影, ChinaJoy, Bilibili World, Tokyo Game Show, 游戏展台, 官方展会摄影, AAA游戏角色Cosplay, 古风闺蜜, AncientFemaleCompanionship, Ancient Female Companionship, 闺阁夜话, 姐妹古风写真, 古代闺阁生活, 灯下梳发, 共读, 夏夜纳凉, 守夜, 江南同行, 湖亭共伞, 山寺同行, 灯会夜游, 柔美, 妩媚, 成熟, 优雅, 高贵, 清冷, 温婉, 英气, 神性, 慵懒, 东方幻想, 古风美女, 古典东方美人, 宋韵, 唐宫, 江南, 洛神, 青瓷, 昆曲, 国风人物, 汉服女子, 写实东方美人, 现代东方审美, 新中式, 东方高级感, SweetHomeGirl, 甜系纯欲生活写真, 甜系纯欲, 甜美女友感, 真实女生, 自然抓拍, 故事感, 女友视角, lifestyle girlfriend portrait, 仙侠/武侠/宫廷/江南/敦煌风 AI 绘画提示词, image prompt optimization, style word expansion, character series concepts, or prompt packs for Midjourney, Stable Diffusion, ComfyUI, DALL-E, Gemini, Seedream, or other image generation tools.

日本語の概要は準備中です。原文の説明を表示しています。

Use when a user asks to scan, investigate, or characterize a URL. Submits to URLScan (unlisted by default), cross-references with VirusTotal and OTX, extracts the parent domain and resolved IP for follow-up investigation. Returns verdict, redirect chain, contacted infrastructure, and screenshot. Invoked by /cti-orchestrator when the target is a URL.

日本語の概要は準備中です。原文の説明を表示しています。

Server-Side Request Forgery (SSRF) occurs when user-controlled input is used to construct URLs that the server fetches, enabling access to internal services, cloud metadata endpoints (169.254.169.254), and local files via `file://` scheme. Detect via parameters accepting URLs or hostnames, PDF/report generators rendering `<iframe>/<img>/<script>`, and blind SSRF via out-of-band DNS callbacks. Bypass filters using IP decimal/octal/hex encoding, URL-userinfo tricks, and URL fragments. Tools: Burp Collaborator, curl.

日本語の概要は準備中です。原文の説明を表示しています。

Address GitHub PR review feedback at line, file, and PR level — fetch, evaluate, fix, commit, and resolve. Use when given a PR URL, a review-comment URL, or a PR-level comment URL. Triggers: GitHub discussion URLs, "fix this review comment", "address PR feedback", "resolve PR comments", PR URLs with #discussion_, #issuecomment-, or #pullrequestreview- fragments.

日本語の概要は準備中です。原文の説明を表示しています。

Orientação especializada para ajuste fino com GRPO/RL usando TRL para treinamento de modelos com raciocínio e tarefas específicas

日本語の概要は準備中です。原文の説明を表示しています。

Ajuste fino de LLMs usando reinforcement learning com TRL - SFT para ajuste de instruções, DPO para alinhamento de preferências, PPO/GRPO para otimização de recompensas e treinamento de modelo de recompensas. Use quando precisar de RLHF, alinhar modelo com preferências ou treinar a partir de feedback humano. Funciona com HuggingFace Transformers.

日本語の概要は準備中です。原文の説明を表示しています。

Extract clean markdown, text, or targeted chunks from one or more known, publicly accessible webpage URLs via the Tavily CLI. Use when the URLs are already supplied or discovered and the user wants page content rather than more URLs. Do not use to find an unknown page (tavily-search), list a site's URL structure (tavily-map), bulk-extract a site section (tavily-crawl), or retrieve named platform, authenticated, cookie-dependent, or platform-native content (Agent Reach). Can process up to 20 public URLs per request.

日本語の概要は準備中です。原文の説明を表示しています。

Take ONE specific bookmark URL and explore the neighborhood around it — tags, SIMILAR_TO neighbors, community peers — plus optionally WebFetch the page itself to summarize it. Use whenever Ahmad pastes a single URL and asks "what's near this", "tell me about this one", "expand this bookmark", "dig into this", "what else did I save like this", "give me the context for this URL". Triggers on "expand", "neighbors of", "what else like this", "dig into this URL", "context for this", and any message that's just a URL plus a short prompt.

日本語の概要は準備中です。原文の説明を表示しています。

URLScan.io is a free service for scanning and analyzing suspicious URLs. It captures screenshots, DOM content, HTTP transactions, JavaScript behavior, and network connections of web pages in an isolat

日本語の概要は準備中です。原文の説明を表示しています。

指定した起動ディレクトリと、親が確定した dotfiles root のスキル配置・エージェント定義を点検する。 判定の正は etc/audit-skill-agent-layout.py。実効 runtime の native 優先順、Cursor の name、 agent の model/role、生成物の状態を報告する。「overlay 点検」「スキル配置」 「エージェント定義は共通か」「layout audit」「/overlay-audit」で使う。

Generate an AI video end-to-end on Ozor.ai from a short text brief. Use this skill whenever the user wants to actually create a video (not just write a prompt) from a creative description, idea, or short brief — e.g. 'make me a video about X', 'generate a launch video for Y', 'create an explainer for Z', 'render a video', 'produce a 30s ad about ...', 'spin up a quick video for our new feature'. This skill runs the full Ozor MCP pipeline: generate_video → wait_for_job → export_video → wait_for_export → get_embed_code, and returns the shareUrl, editorUrl, downloadUrl, and a ready-to-paste <iframe> embed. Do NOT use this skill when the input is a document/PDF/PPTX/deck/URL — use ozor-document-video or ozor-url-to-video instead.

日本語の概要は準備中です。原文の説明を表示しています。

Guide for defining RL environments — the Env protocol, EnvGroupBuilder, RLDataset, and custom environment creation. Use when the user asks about RL environments, reward functions, or how to define custom tasks for RL training.

日本語の概要は準備中です。原文の説明を表示しています。

Set up and run multi-turn RL training for interactive environments (terminal tasks, tool use, search/RAG, games) using the Tinker API. Use when the user wants multi-turn RL, agentic training, tool-use RL, or interactive environment training.

日本語の概要は準備中です。原文の説明を表示しています。

tespit etme (s) and prevents code injection attacks targeting serverless functions (AWS Lambda, Azure Functions, Google Cloud Functions) through event source poisoning, malicious layer injection, runtime command execution, and IAM privilege escalation via function modification. The analyst combines static analysis of function code, CloudTrail event correlation, runtime behavior monitoring, and IAM policy auditing to identify injection vectors across the expanded serverless attack surface including AP...

日本語の概要は準備中です。原文の説明を表示しています。

Fetch a URL and save the body to a temp file so you can pipe it through rg / jq / awk instead of dumping the whole page into the conversation. Use this whenever a user gives you a URL to read - articles, docs, PRs, gists, README files, blog posts, anything HTTP. Returns a path on disk; the agent then greps the path for the relevant section and quotes only what matters. Strongly preferred over the built-in webfetch tool when (a) the page is more than a few KB, (b) the user wants to find a specific phrase / table / code block on the page, (c) you need to keep the long page out of context, or (d) you want to inspect raw HTML alongside rendered markdown. Triggers: 'fetch', 'fetch this URL', 'fetch and save', 'fetch the page', 'pull this', 'grab this article', 'read this URL', 'read the docs at', 'mirror this page', 'page is too long', 'save the page so I can grep it', 'put it on disk so I can rg', 'cache this URL'.

日本語の概要は準備中です。原文の説明を表示しています。

URLScan.io is a free service for scanning and analyzing suspicious URLs. It captures screenshots, DOM content, HTTP transactions, JavaScript behavior, and network connections of web pages in an isolat

日本語の概要は準備中です。原文の説明を表示しています。

Expert guidance for GRPO/RL fine-tuning with TRL for reasoning and task-specific model training

日本語の概要は準備中です。原文の説明を表示しています。

Fine-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.

日本語の概要は準備中です。原文の説明を表示しています。

Imports FBX meshes into CARLA on UE 5.8 as spawnable props — drives the editor headlessly to import the mesh, derives the prop size from its measured bounds, and registers it in PropParameters.json so it appears in the blueprint library as static.prop.<name>. Handles the FBX unit-scale problem, verifies the result against a running server, and can revert cleanly. Use when the user asks to "import a prop", "add a custom mesh/object to CARLA", "make my model spawnable", or a prop imports but never shows up in the blueprint library.

日本語の概要は準備中です。原文の説明を表示しています。

Imports a pedestrian into CARLA on UE 5.8 as a spawnable, animating walker — imports the skinned FBX bound to CARLA's shared pedestrian skeleton, duplicates a donor walker blueprint and repoints it at the new mesh, and registers it in WalkerParameters.json as walker.pedestrian.<id>. Can also export a shipped walker to FBX, which is how you obtain a rig-conforming mesh to start from. Use when the user asks to "import a walker/pedestrian", "add a custom character", "clone a walker", or has a walker that spawns but stands still.

日本語の概要は準備中です。原文の説明を表示しています。

Starts a CARLA UE 5.8 server from a ue58-dev tree — the editor binary in -game mode (the working headless server), the packaged shipping build, or the full editor — with offscreen/nullrhi rendering, custom RPC and streaming ports, and native ROS 2 via the single-dash -ros2 flag. Probes a running server, and stops one cleanly despite the 15-character process-name truncation. Use when the user asks to "run/start CARLA UE5", "launch a ue58 server", "run headless", "publish ROS 2 topics from UE5", or a client cannot reach the simulator.

日本語の概要は準備中です。原文の説明を表示しています。

Imports FBX meshes into a CARLA source build as spawnable static.prop.* blueprints — pass a directory whose subdirectories name the semantic tags, or a single FBX with its tag, and the props are imported into content, registered, added to the prop factory and verified by spawning. Use when the user asks to "add a prop to CARLA", "import an FBX as a prop", "add street furniture/benches/signs", or needs a custom mesh spawnable from the Python API.

日本語の概要は準備中です。原文の説明を表示しています。

Inspects and visualises a ROS-2-enabled CARLA server from Docker containers, with no local ROS 2 install — list/echo/rate-check topics (the "are the topics actually on the wire" check), run the bundled map-and-lidar demo stack (hero vehicle with camera/lidar/GNSS/IMU on autopilot), and open RViz2 with the lane-network preset. Use when the user asks to "check the ROS topics", "echo /carla/map", "is CARLA publishing to ROS", "visualise the lidar in RViz", "open rviz", or "run the ROS 2 demo".

日本語の概要は準備中です。原文の説明を表示しています。

Reads and changes a running CARLA server's WorldSettings via apply_settings — synchronous vs asynchronous mode, fixed_delta_seconds (fps), physics substepping, and no_rendering_mode — and keeps the Traffic Manager's sync state matched automatically. Use when the user asks to "enable/disable sync mode", "set synchronous mode", "run at a fixed fps/step", "turn off rendering", "make the sim deterministic", or "restore async mode", and as the prerequisite for keeping settings across a map load.

日本語の概要は準備中です。原文の説明を表示しています。