---
pretty_name: RTLLM Variants
license: mit
task_categories:
  - text-generation
language:
  - en
tags:
  - verilog
  - rtl
  - benchmark
  - dataset-derivative
size_categories:
  - n<1K
configs:
  - config_name: default
    data_files:
      - split: test
        path: rtllm_v2_mod/viewer/test.jsonl
---

# RTLLM Variants (Community Dataset Derivatives)

This repository hosts **community-maintained derivative variants** based on upstream RTLLM releases. It is intended for benchmarking reproducibility and evaluation workflow integration.

## Important Notice

- This repository is **not an official release** from the RTLLM paper authors.
- Variants in this repository may include prompt wording normalization, interface naming unification, or testbench output standardization for evaluator consistency.
- For official benchmark content, refer to upstream RTLLM.

## Upstream Reference

- Project: https://github.com/hkust-zhiyao/RTLLM
- RTLLM v2 baseline commit used by current variant pipeline:
  - `41b26896e33b536940116a975626455eed3de65e`

## Available Variants

- `rtllm_v2_mod/`
  - `converted/`: benchmark layout
  - `delta_report.json`: machine-readable differences vs upstream baseline
  - `manifest.json`: provenance and generation metadata
  - `README.md`: variant-specific summary

## Intended Usage

- Direct download via Hugging Face dataset API/CLI
- Reproducible benchmark evaluation pipelines that consume RTLLM-style folder structure

## License

Each variant follows upstream licensing constraints and includes required license files where applicable.