benchclaw-stage3-evidence-compiler

What problem does it solve?

This Skill automates the process of compiling evidence, cleaning data, and generating ground truth for benchmarks, significantly reducing the manual effort required for benchmark development.

Core Features & Use Cases

• Evidence Compilation: Automatically compile evidence from various data sources, including real images, existing benchmarks, and simulators.
• Data Cleaning: Clean and preprocess data from diverse sources to ensure quality and consistency.
• Ground Truth Generation: Generate ground truth for annotated data to facilitate benchmark evaluation.
• Use Case: Imagine you have a collection of real images, existing benchmarks, and simulator outputs. Use this Skill to compile the evidence, clean the data, and generate ground truth for benchmark evaluation.

Quick Start

Run the benchclaw-stage3-evidence-compiler skill with the input plan and data bundles to generate the annotated data bundles for stage 4.