Results for “reward-functions”

v4.1.0 reward function redesign to fix overtrading and DSR dominance

Use when tuning RL reward function weights, fixing reward hacking, or addressing P&L gradient issues

Fix HOLD bias in RL reward function. Trigger when: (1) model learns to always HOLD, (2) trade rate is too low (<10%), (3) slippage penalty exceeds typical price moves.

Provides SQL window function examples for ranking, aggregation, lag/lead, value extraction, frame specifications, and advanced analytics.

Use to design support incentives that improve behaviour without destroying the metric — pairing pay with guardrails, naming gaming modes, and choosing measures that survive Goodhart pressure. Trigger for "incentive plan", "agent bonus scheme", "SPIFF design", "pay for QA score", "what metric should we bonus", CSAT incentives, or reviewing whether a comp change is driving gaming.

Add account state (P&L, win rate, drawdown) to RL observations + drawdown penalty in rewards. Trigger when: (1) model needs account awareness, (2) training should penalize drawdowns, (3) upgrading obs_dim 5300→5600.

Refines task prompts through iterative questioning about scope, deliverables, and constraints, then copies the final markdown to the clipboard.

Train and fine-tune transformer language models using TRL (Transformers Reinforcement Learning) with support for SFT, DPO, GRPO, KTO, RLOO, and reward model training via CLI commands.

Design motivation systems for products and teams using Autonomy, Mastery, and Purpose (AMP) to drive intrinsic engagement.

Volume-based PUMP token reward system with day-indexed epochs, pro-rata distribution, accumulator tracking, and cross-program claiming on Solana.

Agent validation overhaul: reward weight overrides, fitness decline gate, pinned data, staged experiments

Design in-game economies including currencies, progression systems, loot tables, monetization, and resource flow. Use when building the economic backbone of a game. Also trigger for "game economy", "loot table", "drop rates", "currency system", "progression design", "monetization strategy", or "reward system".

Diagnoses social threats and rewards across five domains to improve organizational, team, and product experiences.

Use this skill for feature flags, gradual rollout, kill switches, metrics, experiment gates, rollback. Trigger when the task involves programming work related to Feature Flag Rollout, implementation, audits, debugging, strategy, or validation.

Build quiz and assessment funnels that generate qualified leads through interactive scorecards, covering concept hooks, question design, dynamic results, and automated follow-up sequences.

Frame jobs, triggers, desired outcomes, anxieties, and product opportunities.

Fine-tune and align language models using reinforcement learning with TRL, including SFT, DPO, PPO, GRPO, and reward model training.

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).

Influence Functions in Deep Learning

Nir Eyal engagement loop design — trigger, action, variable reward, investment

Classifies code review findings to choose the smallest correct response, preventing one-patch-per-comment churn.

Reviews animation and motion code against a high craft bar derived from Emil Kowalski's design engineering philosophy. Default to flagging; approval is earned.

Scores deal margin and risk, routes discount approval to the right human approver, and redlines terms against commercial policy for per-deal review.

Refactors Swift code toward declarative functional programming with immutability, value types, pure functions, transformations, and closure-based APIs.

Guides users through participating in the SOMA decentralized training network, covering data submission, model training, reward claiming, and strategic optimization.

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

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.

Interact with the Stakr protocol to create ERC-4626 vaults, add and modify multi-reward staking programs, and manage agent-owned vaults.

当设定个人或工作目标，希望目标本身能带来持续的内在动力和满足感时

Defines the architecture and orchestration primitives for multi-agent teams, covering role allocation, conflict resolution, reward routing, and the operational lifecycle of agent collectives.

当寻求个人行动或项目在更宏大系统中的意义，或需要将日常工作与长远价值连接时

Read-only query methods for PumpFun claims — unclaimed token rewards, creator vault balances, volume accumulators, distributable fees, and current-day token previews across Pump and PumpAMM programs.

Design motivation systems using Autonomy, Mastery, and Purpose (AMP) for products and teams. Use when the user mentions "intrinsic motivation", "gamification isnt working", "team incentives", "autonomy", "mastery", "purpose-driven", "employee engagement", or "reward systems". Also trigger when designing onboarding progression systems, fixing broken gamification, or building team structures that sustain high performance. Covers why carrot-and-stick fails and how to build progress systems. For habit-forming product loops, see hooked-ux. For retention behavior design, see improve-retention.

Reference for MariaDB window functions and the OVER clause, covering ranking, value navigation, and ordered-set functions with syntax and common pitfalls.

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