This is a minimal, reproducible set of steps to run a baseline vs RL comparison for Lighthouse using Honggfuzz, and inspect the outputs you care about.

• OS: Linux or macOS (x86_64 or Apple Silicon). On macOS, use Docker Desktop or Colima.
• Docker: Engine 20+ with BuildKit; able to run docker build and docker run.
  • Resources: 4+ CPU, 8–16 GB RAM, 20+ GB free disk (images + corpora + logs).
• Make + Bash available on host.
• Network: Access to GitHub, Docker registries, and crates.io (prefetch steps will pull crates inside the container).

Repository context: https://github.com/RiemaLabs/alma

Build the Lighthouse image and prefetch dependencies (Docker‑based runs):

cd eth2fuzz
make lighthouse
make prefetch-lighthouse

Run a 60s baseline vs RL comparison (2 threads, 10s segments):

make cov-attestation fuzzer=honggfuzz T=60 RT=60 S=10 n=2 tag=demo

Show detailed fields (new_units / mutated_new):

SHOW_VERBOSE=1 make cov-attestation fuzzer=honggfuzz T=60 RT=60 S=10 n=2 tag=demo

After running the command, you will see a summary like this:

--- SUMMARY (target=lighthouse_attestation) ---
Baseline: mutated_new_cov +35, cov: 6%
RL:       mutated_new_cov +48, cov: 7%

• Baseline kept samples:
  • eth2fuzz/workspace/logs/demo/base/libfuzzer_corpus/lighthouse_attestation
• RL kept samples:
  • eth2fuzz/workspace/logs/demo/rl/libfuzzer_corpus/lighthouse_attestation
• Honggfuzz logs (coverage%, new_units):
  • Baseline: eth2fuzz/workspace/logs/demo_base/rl/hfuzz/logs/lighthouse_attestation.log
  • RL: eth2fuzz/workspace/logs/demo/rl/hfuzz/logs/lighthouse_attestation.log

• Focus on mutated_new_cov (coverage‑contributing samples). This is the most reliable indicator of persistent improvement.
• Equal coverage% is common on mature corpora; RL gains often appear as a few extra kept samples rather than large % jumps.

Remove all logs:

make clean-logs

Remove logs for a specific tag:

make clean-tag TAG=demo

Remove temp only (merge_tmp, rl_input, etc.):

make clean-tmp

MIT — see LICENSE

This repository builds upon the open‑source beacon‑fuzz/eth2fuzz work led by Sigma Prime (and contributors), and the broader Ethereum community. Thank you for the foundations and prior art.