'preference optimization' 검색 결과

Choose a model training method and configuration from paired response preferences or individual positive and negative ratings. Covers DPO, ORPO, KTO, and SimPO, including pair construction, repeated training rounds, and troubleshooting learning settings.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Choose a training method for paired response preferences
• Select a method for individual positive and negative ratings
• Review learning settings when DPO training loses general capability

Guide language model training with SimPO using pairs of preferred and rejected answers. Covers configuration, learning rate tuning, and memory troubleshooting for preference alignment, which adjusts model responses toward human preferences without a separate reference model.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Training on preferred and rejected answer pairs
• Exploring an alternative to DPO without a reference model
• Tuning preference training for math and code tasks

Train language models with SimPO using pairs of preferred and rejected answers. Configure preference training without a separate reference model, adapt settings for base or instruction models, and address unstable training, memory limits, or lost capabilities.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Training a base model with preferred and rejected answers
• Further tuning an instruction model with preference data
• Adjusting training settings for math and code tasks

Audits and optimizes the five global Claude layers (User Preferences, Styles, Global Memory, Skills, MCP Connectors) using the Global Layer Scorecard (six dimensions, anchored 1-5 rubrics). Detects eight cross-layer failure modes and produces evolutionary recommendations (Promotion, Demotion, Codification, Skill Extraction). Use when user says "audit my global setup," "optimize my preferences," "review my Claude configuration," "check my cross-project setup," "are my preferences working," "clean up my global memory," or "what should be in my preferences vs my project." Also use when a user has 3+ Projects and wants to improve their shared foundation. Do NOT use for single-Project audits, Project Memory optimization, or full-stack audits (use rootnode-project-audit, rootnode-memory-optimization, or rootnode-full-stack-audit respectively, if available). Run on Opus 5 or Sonnet 5 at `high` effort (both defaults); depth reduces on legacy models.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Rebalances context across Memory, Custom Instructions, knowledge files, and User Preferences in Claude Projects. Audits Memory for redundancy, staleness, and misplacement; prescribes optimization including Codification of stable Memory patterns into explicit User Preferences or Project CI rules. Use when user says "optimize my memory," "what should be in my memory," "trim my knowledge files," "reduce my context usage," "rebalance my project," "what should be in my preferences vs memory," or asks whether something belongs in Memory, a knowledge file, or User Preferences. Also trigger on symptom-phrased: "my project feels bloated," "Claude keeps forgetting things," "my context window keeps hitting limits." Activate whenever Memory-layer balance is the primary concern. Do NOT use for full Project audits (use rootnode-project-audit if available), single-prompt evaluation (use rootnode-prompt-validation if available), or global-layer audits that don't touch Project Memory (use rootnode-global-audit if available).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Guide language model training with TRL, from learning to follow instructions to matching human preferences. Covers method selection, training data, reward models that score responses, evaluation, and adjustments for memory limits or unstable training.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Train a model on instruction and response examples
• Align responses using preferred and rejected answer pairs
• Build a training pipeline using human feedback

Guide language model training with TRL, from learning to follow instructions to matching human preferences. Covers method selection, training data, reward models that score responses, evaluation, and adjustments for memory limits or unstable training.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Train a model on instruction and response examples
• Align responses using preferred and rejected answer pairs
• Build a training pipeline using human feedback

Simple Preference Optimization for LLM alignment. Reference-free alternative to DPO with better performance (+6.4 points on AlpacaEval 2.0). No reference model needed, more efficient than DPO. Use for preference alignment when want simpler, faster training than DPO/PPO.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Simple Preference Optimization para alinhamento de LLMs. Alternativa sem modelo de referência ao DPO com melhor desempenho (+6.4 pontos no AlpacaEval 2.0). Sem modelo de referência necessário, mais eficiente que DPO. Use para alinhamento de preferências quando quer treinamento mais simples e rápido que DPO/PPO.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Set up and run Direct Preference Optimization (DPO) training on preference datasets using the Tinker API. Use when the user wants to train with preference data, chosen/rejected pairs, or DPO.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

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.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Decide whether an LLM needs additional training or whether retrieval and better prompts will suffice. Match available examples, preferences, or verifiable results to a training method, then assess model size and memory requirements before starting.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Comparing fine-tuning with retrieval and prompt changes
• Choosing a training method for available data
• Assessing model size and memory requirements

Align language models with human preferences using Direct Preference Optimization — no reward model, no RL loop.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Use this skill to create, configure, or tune Field Service Lightning scheduling policies — including work rules (pass/fail filters) and service objectives (weighted ranking criteria). Covers the four default policies, custom policy design, work rule type selection, and objective weighting strategy. NOT for running or tuning the Optimizer itself — use admin/fsl-scheduling-optimization-design. NOT for resource skills and preferences — use admin/fsl-resource-management. NOT for resource capacity records — use admin/fsl-capacity-planning.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

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.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Simple Preference Optimization for LLM alignment. Reference-free alternative to DPO with better performance (+6.4 points on AlpacaEval 2.0). No reference model needed, more efficient than DPO. Use for preference alignment when want simpler, faster training than DPO/PPO.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Review past dsh agent sessions to identify recurring tasks and test proposed skill or memory improvements. This skill guides preview runs, validation on reserved tasks, staged proposals, adoption with backups, and scheduled nightly cycles.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Extract recurring tasks from past agent sessions
• Test proposed skill and memory improvements
• Review staged changes before adopting them

Review past Claude Code sessions and turn recurring tasks into proposed updates to skills and project memory. Replay tasks, compare scores on separate validation tasks, and review staged edits before applying them with backups.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Consolidate repeated Claude Code corrections into reusable rules
• Review proposed changes to SKILL.md and CLAUDE.md
• Try the cycle without provider calls using mock

Use past local Codex sessions to propose updates to a selected skill and project memory file. Recurring tasks are replayed, changes are checked on separate validation tasks, and proposals are staged for review before adoption.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Turn recurring Codex requests into skill updates
• Consolidate feedback into skills and project memory
• Compare validation scores before applying changes

Design persistent agent memory that retains knowledge across sessions and retrieves relevant facts when needed. Compare storage approaches, track people and changing information, and plan memory consolidation and evaluation based on the kinds of questions the agent must answer.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Retain user preferences across conversations
• Track people and entities consistently across sessions
• Retrieve facts valid at a specific point in time

Fine-tune models on Microsoft Foundry using SFT (supervised), DPO (preference), or RFT (reinforcement with graders). Covers dataset preparation, training job submission, deployment, and evaluation. USE FOR: fine-tune, SFT, DPO, RFT, training data, grader, distillation, fine-tuned model, training job, large file upload, calibrate grader, deploy fine-tuned model, evaluate fine-tuned model. DO NOT USE FOR: general model deployment without fine-tuning (use deploy-model), agent creation (use agents), prompt optimization without training (use prompt-optimizer).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Fine-tune models on Microsoft Foundry using SFT (supervised), DPO (preference), or RFT (reinforcement with graders). Covers dataset preparation, training job submission, deployment, and evaluation. USE FOR: fine-tune, SFT, DPO, RFT, training data, grader, distillation, fine-tuned model, training job, large file upload, calibrate grader, deploy fine-tuned model, evaluate fine-tuned model. DO NOT USE FOR: general model deployment without fine-tuning (use deploy-model), agent creation (use agents), prompt optimization without training (use prompt-optimizer).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Generates 3D conformer ensembles using RDKit ETKDGv3 with knowledge-enhanced distance geometry, MMFF94/UFF force-field optimization, CREST + GFN2-xTB semi-empirical refinement, and macrocycle-aware torsion preferences. Provides explicit decision rules for single vs ensemble conformer use, RMSD pruning, energy windows, conformer count, and force-field choice. Use when preparing 3D ligands for docking, generating descriptor input for 3D QSAR, or sampling macrocycle/peptide conformational ensembles.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Generates 3D conformer ensembles using RDKit ETKDGv3 with knowledge-enhanced distance geometry, MMFF94/UFF force-field optimization, CREST + GFN2-xTB semi-empirical refinement, and macrocycle-aware torsion preferences. Provides explicit decision rules for single vs ensemble conformer use, RMSD pruning, energy windows, conformer count, and force-field choice. Use when preparing 3D ligands for docking, generating descriptor input for 3D QSAR, or sampling macrocycle/peptide conformational ensembles.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.