skills.rest
qualit527
qualit527COMMUNITY
Explore all skills by this author →
GitHub

verify-decoder

Audits checkpoint's promised delta_ler improvement for authenticity using onboard sensors and cameras.

3|Updated Apr 20, 2026
Search Tags:#verification#checkpoint#verify#llm-reasoning#autoqec#delta_ler#predecoder
One-click install
npx skills add https://github.com/qualit527/qec-ai-decoder --skill verify-decoder
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: verify-decoder
Source: https://github.com/qualit527/qec-ai-decoder/tree/main/.claude/skills/verify-decoder
Command: npx skills add https://github.com/qualit527/qec-ai-decoder --skill verify-decoder

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Audit a predecoder checkpoint against holdout seeds to ensure the delta_ler improvement is genuine and not a reward-hacking artifact.

Core Features & Use Cases

  • Run independent_eval (3 fair-baseline guards) and interpret borderline cases with LLM reasoning.
  • Read verification_report.json + training.log to diagnose the reliability of a promising delta_ler.
  • Archive a concise diagnostic decision to round_N/decision.md for auditability and governance.

Quick Start

Run the verification workflow on a round directory with the AutoQEC verify command.

Frequently Asked Questions about verify-decoder

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I verify a predecoder checkpoint for reward hacking before deployment?▼

You need a round directory and an environment YAML file to execute the verify CLI command. These inputs allow the workflow to run independent_eval guards, interpret borderline cases, and archive the final decision into round_N/decision.md.

What is delta_ler improvement verification in AutoQEC rounds?▼

The verification process executes independent_eval with three fair-baseline guards and uses LLM reasoning to interpret borderline cases. It then reads verification_report.json and training.log to diagnose the reliability of the promising delta_ler.

How do I diagnose a promising delta_ler result from a predecoder checkpoint?▼

After verification, the workflow archives a concise diagnostic decision into round_N/decision.md. This output provides auditability and governance by documenting whether the predecoder checkpoint passed the delta_ler verification.

When do I need to run independent verification on an AutoQEC round?▼

The verification workflow uses independent_eval with three fair-baseline guards to interpret borderline cases. It reads verification_report.json and training.log to diagnose the delta_ler reliability and rule out reward-hacking artifacts.

Related Skills

Explore More Data & Analytics Skills→
johnlindquistjohnlindquist
24

beads

Track tasks and issues with dependency awareness using a Git-backed issue tracker.

Community
Intermediate
SnakeOSnakeO
9

debug:angular

Diagnose Angular runtime errors using DevTools, ng.probe, and console utilities.

Community
Advanced
anxiety135790anxiety135790

evaluating-llms-harness

Benchmarks language models using standardized academic and code evaluation suites.

Community
Advanced
junhyekhjunhyekh
1

rlxp-reward-review

Reviews reinforcement-learning reward changes for task alignment and approval readiness.

Community
Intermediate