shirman/exploitbench-answers

ExploitBench Complete Answers

This dataset contains the full answer set for every task in ExploitBench v8-bench. It provides complete answers across all 41 real-world V8 N-day vulnerability environments in the benchmark.

This is an independent, unofficial answers repository. It is not maintained by the ExploitBench authors, Carnegie Mellon University, or the official ExploitBench organization. For the benchmark implementation, methodology, environments, official results, and leaderboard, use the canonical resources linked below.

Coverage

The answer set covers all 41 v8-bench environments and the complete exploitation ladder used to evaluate progress. ExploitBench grades 16 mechanically verified capabilities across five tiers:

Tier	Capability stage
T5	Reach vulnerable code
T4	Reproduce or trigger the vulnerability
T3	Build target-specific V8 exploit primitives
T2	Build generic primitives such as information leaks and arbitrary read/write
T1	Achieve control-flow hijack and arbitrary code execution

v8-bench targets patched, real-world vulnerabilities in Google's V8 JavaScript and WebAssembly engine. Agents receive the vulnerable and fixed source context plus the upstream patch, then produce JavaScript or WebAssembly programs that are graded by deterministic challenge-response oracles with the V8 security sandbox enabled.

Files

• answers.md — complete answers for all ExploitBench v8-bench tasks;
• README.md — dataset card, coverage, citation, and provenance context.

Use an immutable Hugging Face revision when citing or evaluating against this dataset so the exact answer set can be reproduced.

Download

from huggingface_hub import hf_hub_download

answers_path = hf_hub_download(
    repo_id="shirman/exploitbench-answers",
    filename="answers.md",
    repo_type="dataset",
)

print(answers_path)

Or with the Hugging Face CLI:

hf download shirman/exploitbench-answers answers.md --repo-type dataset

Canonical ExploitBench resources

• Official ExploitBench website and leaderboard
• Official ExploitBench repository
• Official ExploitBench Hugging Face dataset
• ExploitBench paper

Benchmark integrity and responsible use

This dataset contains complete benchmark answers and can invalidate independent evaluation if exposed to an evaluated model, agent, prompt, retrieval system, or training pipeline. Treat any such exposure as benchmark contamination and disclose it when reporting results. Do not train or fine-tune on these answers and then present performance on the same tasks as an independent ExploitBench evaluation.

ExploitBench concerns real-world vulnerability exploitation and is inherently dual-use. Use these materials only for authorized security research, defensive evaluation, reproducibility, model-safety work, vulnerability triage, and mitigation validation. Do not use them to compromise systems, software, or data that you do not own or have explicit permission to test.

Citation

Please cite the original ExploitBench paper when using the benchmark:

@misc{lee2026exploitbench,
  title         = {ExploitBench: A Capability Ladder Benchmark for LLM Cybersecurity Agents},
  author        = {Seunghyun Lee and David Brumley},
  year          = {2026},
  eprint        = {2605.14153},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CR},
  url           = {https://arxiv.org/abs/2605.14153}
}

When citing this answer dataset, include its Hugging Face URL and an immutable repository revision alongside the benchmark citation.