Results for “equal-error-rate”

Benchmarks automatic post-editing (APE) models on WMT'18 SMT, SubEdits, and MLQE-PE datasets, reporting BLEU, ChrF, and TER scores computed with SacreBLEU and TERCOM.

Enforce idiomatic Rust patterns for ownership, error handling, enums, traits, concurrency, and best practices to build safe, performant applications.

Error Recovery Playbook

Diagnose and remove FPS caps so the editor and game run uncapped (or at a target FPS). Use when the user says the editor/game is "locked", "capped", or "stuck" at a frame rate (commonly 60 FPS), asks to "unlock"/"uncap"/"raise" FPS, or wants to set a max FPS. Covers t.MaxFPS, VSync, fixed/smoothed frame rate, and background CPU throttling (EngineSettingsService).

Run Semgrep static analysis scans with automatic language detection, parallel subagent execution, and merged SARIF output. Supports full ruleset coverage or high-confidence security vulnerability filtering.

Optimize Entity Framework Core queries by fixing N+1 problems, choosing correct tracking modes, using compiled queries, and avoiding common performance traps.

Compare coding agents head-to-head on reproducible tasks with pass rate, cost, time, and consistency metrics.

跨 TypeScript、Python、Go 的健壮错误处理模式——涵盖类型化错误、错误边界、重试策略及优雅降级

Integrate FIRST's Exploit Prediction Scoring System (EPSS) API to prioritize vulnerability remediation based on real-world exploitation probability within 30 days.

VaR Unit Mismatch Fix

Prevents silent decimal mismatch bugs across EVM chains by querying token decimals at runtime, caching by chain and token address, and using safe normalization for bots, dashboards, and DeFi tools.

Stop throwing everywhere - handle errors as values using Either and TaskEither for cleaner, more predictable code

**Pattern:**

Runs regression evaluations comparing agent edits against human-approved golden projects, and registers new goldens after human approval.

Score LLM and agent outputs using LLM-as-judge techniques — direct scoring against rubrics or pairwise comparison between two outputs. Includes built-in bias mitigation for position bias, length bias, and self-enhancement bias. Load when the user asks to score an output, judge a response, evaluate against a rubric, compare two outputs, do direct scoring, run pairwise comparison, or says "rate this", "which response is better", "score this against the rubric", "judge this output", "LLM as judge this". Sub-skill of eval-output orchestrator.

Every metric encodes an opinion about which mistake hurts.

Prevents AI-introduced regressions with sandbox-mode API testing, automated bug-check workflows, and patterns that catch blind spots where the same model writes and reviews code.

Stop throwing everywhere - handle errors as values using Either and TaskEither for cleaner, more predictable code

Replace exception-heavy TypeScript code with fp-ts Either and TaskEither for type-safe error handling as values.

GPU-accelerated correlation matrix computation with persistent SQLite caching to eliminate bottleneck at correlation calculation during symbol selection

Detect insider threats by modeling normal user and entity behavior with Elasticsearch, computing anomaly scores, and correlating low-confidence indicators into high-confidence alerts.

SSOT（Single Source of Truth）と実際のファイル/設定の整合性をチェックし、乖離があれば修正するスキル。 「SSOTチェックして」「SSOT整合性チェックして」「SSOT整理して」「SSOTのズレを直して」 「00_SYSTEM更新して」「乖離を修正して」と言った時にトリガーする。 /ssot-check でも呼び出せる。

Test a pipeline stage's skill file by running the stage WITH and WITHOUT the skill on the same input, comparing outputs, and proposing skill edits. Ryan Law principle 3 — recursive self-improvement. Run after any board complaint about a stage, and monthly per core stage.

Provides a structured risk scoring methodology for Data Protection Impact Assessments aligned with ENISA threat taxonomy and ISO 29134. Covers likelihood and severity assessment, risk matrix construction, inherent vs residual risk calculation, and risk appetite thresholds per EDPB WP248rev.01 guidance. Keywords: risk scoring, DPIA risk matrix, likelihood, severity, ENISA, ISO 29134, residual risk, risk appetite.

Generates four improved variations of any installed skill, scores them against a weighted rubric, and applies the winning version while preserving the original.

Head-to-head comparison of coding agents (Claude Code, Aider, Codex, etc.) on custom tasks with pass rate, cost, time, and consistency metrics

Validates the Emotional Attitude Score (EAS) metric by measuring its consistency with human judgment on word-level sentiment polarity, using the AmbGIMT dataset and pairwise score comparisons.

Use this skill to evaluate which DataQ challenges have winning evidence and which don't. Covers documented vs anecdotal evidence and the 8 high-success patterns.

EMC pre-compliance risk analysis for KiCad PCB designs — 18 check categories, 44 rule IDs covering ground planes, decoupling, I/O filtering, switching harmonics, clock routing, differential pair skew, board edge radiation, PDN impedance, return paths, crosstalk, ESD protection, shielding, and magnetic leakage from switching inductors. Produces severity-ranked risk report with pre-compliance test plan. Supports FCC Part 15, CISPR 32, CISPR 25 (automotive), MIL-STD-461G. SPICE-enhanced when available. Use when the user asks about EMC, EMI, radiated/conducted emissions, FCC compliance, CE marking, CISPR, ground plane issues, decoupling strategy, clock routing EMC, switching noise, differential pair skew, or whether their board will pass EMC testing. Also for "will this pass FCC?", "check my EMC", "is my ground plane okay?", "check my decoupling", or "generate an EMC test plan".

Evaluate LLM outputs systematically — benchmarks, automated metrics, human preference, and regression tracking

Evaluates speaker verification models by computing true positive rate at fixed false positive rate thresholds, probing embedding space separation of same-speaker versus different-speaker pairs.

Handle errors as values using fp-ts Either and TaskEither for cleaner, more predictable TypeScript code. Use when implementing error handling patterns with fp-ts.

Stop throwing everywhere - handle errors as values using Either and TaskEither for cleaner, more predictable code

Fixes common PyTorch bugs including percentile calculations, LayerNorm for Conv1d, and buffer edge cases in reinforcement learning and neural network code.

使用seedance2.5模型生成视频,使用 Seedance 2.5 按用户原始提示词生成视频，禁止改写提示词或切换模型，并在生成前补齐时长、比例和检索所得的必要信息后向用户确认原样透传提示词、不润色视频 prompt、不要改写后生成，或显式调用本 Skill 时使用。