adapter

Convert external benchmarks into the BenchFlow framework with routing and parity verification.

317|40|Updated Jan 21, 2025
One-click install
npx skills add https://github.com/benchflow-ai/benchflow --skill adapter-benchflow-ai
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: adapter
Source: https://github.com/benchflow-ai/benchflow/tree/main/.agents/skills/adapter
Command: npx skills add https://github.com/benchflow-ai/benchflow --skill adapter-benchflow-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill solves the fragmentation of benchmark formats by providing a structured, verifiable path to porting external benchmarks into the BenchFlow ecosystem.

Core Features & Use Cases

  • Benchmark Routing: Classifies benchmarks into native, translated, or as-is layers to ensure the most efficient integration.
  • Parity Verification: Automates side-by-side and reward-distribution parity checks to ensure converted benchmarks behave exactly like their originals.
  • Use Case: When you need to port an existing research benchmark into BenchFlow, use this Skill to scaffold the directory, generate the converter, and run the parity gate to confirm the conversion is accurate.

Quick Start

Use the adapter skill to classify the benchmark source at the provided path and generate the necessary scaffold for conversion.

Frequently Asked Questions about adapter

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

FAQPage Schema
How do I convert external benchmarks into the BenchFlow framework?

To convert external benchmarks into BenchFlow, you use a multi-layer routing process that classifies the source and generates a deterministic adapter. This scaffolds the directory and ensures high-fidelity reproduction of the original benchmark behavior.

What is benchmark parity verification and how does it work?

Benchmark parity verification automates side-by-side and reward-distribution checks to ensure converted benchmarks behave exactly like their originals. It validates that the structural conversion maintains the original scoring logic and task format.

How do I standardize diverse task formats for evaluation benchmarking?

You standardize diverse evaluation task formats by routing them through native, translated, or as-is layers. This classification ensures the most efficient integration path and generates structural validation for the specific benchmark.

Can I use automated adapters for as-is benchmark formats?

Yes, automated adapters can process as-is benchmark formats through the routing layer. The system classifies the benchmark source and applies the appropriate integration layer to facilitate verifiable porting into the BenchFlow environment.

What is the best way to scaffold a benchmark converter directory?

The best way to scaffold a benchmark converter directory is to use an automated generation process that classifies the source format and creates the necessary structural files. This ensures the conversion path is deterministic and verifiable.

Why does my benchmark scoring logic change after conversion?

If benchmark scoring logic changes after conversion, the structural validation or parity gate may have failed. Running side-by-side parity checks verifies that the deterministic adapter accurately reproduces the original behavior.