Skip to content
@sunblaze-ucb

sunblaze-ucb

Popular repositories Loading

  1. exploitgym exploitgym Public

    ExploitGym is a large-scale, realistic benchmark built from real-world vulnerabilities designed to evaluate AI agents' ability to develop exploits.

    C 1.1k 145

  2. cybergym cybergym Public

    CyberGym is a large-scale, high-quality cybersecurity evaluation framework designed to rigorously assess the capabilities of AI agents on real-world vulnerability analysis tasks.

    Python 918 113

  3. Intuitor Intuitor Public

    [ICLR 2026] Learning to Reason without External Rewards

    Python 424 44

  4. vero vero Public

    Lean 106 15

  5. rl-generalization rl-generalization Public

    Modifiable OpenAI Gym environments for studying generalization in RL

    Python 91 14

  6. verina verina Public

    Verina (Verifiable Code Generation Arena) is a high-quality benchmark enabling a comprehensive and modular evaluation of code, specification, and proof generation as well as their compositions.

    Lean 79 12

Repositories

Showing 10 of 56 repositories
  • exploitgym Public

    ExploitGym is a large-scale, realistic benchmark built from real-world vulnerabilities designed to evaluate AI agents' ability to develop exploits.

    sunblaze-ucb/exploitgym's past year of commit activity
    C 1,114 Apache-2.0 145 8 7 Updated Sep 30, 2026
  • sunblaze-ucb/superred-modules's past year of commit activity
    Python 2 0 0 1 Updated Sep 27, 2026
  • superred Public
    sunblaze-ucb/superred's past year of commit activity
    Python 11 MIT 0 1 0 Updated Sep 27, 2026
  • FedAgent Public

    [NeurIPS 2026] FedAgent: A Library for Decentralized Agent Learning. 🏆 Best Paper Award at the AAAI'26 TrustAgent Workshop, 🏆 Outstanding Paper Award in the AAAI'26 PerFM Workshop.

    sunblaze-ucb/FedAgent's past year of commit activity
    Python 16 Apache-2.0 2 0 2 Updated Sep 26, 2026
  • progent Public

    Progent: Securing AI Agents with Privilege Control

    sunblaze-ucb/progent's past year of commit activity
    Python 52 MIT 14 1 0 Updated Sep 24, 2026
  • cybergym-e2e Public

    CyberGym-E2E is a large-scale benchmark built from real-world vulnerabilities in widely used open-source projects to evaluate AI agents' end-to-end cybersecurity capabilities, from discovering vulnerabilities to generating proof-of-concept to writing patches.

    sunblaze-ucb/cybergym-e2e's past year of commit activity
    Shell 78 Apache-2.0 25 0 2 Updated Sep 3, 2026
  • cybergym Public

    CyberGym is a large-scale, high-quality cybersecurity evaluation framework designed to rigorously assess the capabilities of AI agents on real-world vulnerability analysis tasks.

    sunblaze-ucb/cybergym's past year of commit activity
    Python 918 Apache-2.0 113 6 3 Updated Aug 28, 2026
  • vero Public
    sunblaze-ucb/vero's past year of commit activity
    Lean 106 Apache-2.0 15 5 0 Updated Aug 17, 2026
  • verina Public

    Verina (Verifiable Code Generation Arena) is a high-quality benchmark enabling a comprehensive and modular evaluation of code, specification, and proof generation as well as their compositions.

    sunblaze-ucb/verina's past year of commit activity
    Lean 79 Apache-2.0 12 0 0 Updated Aug 17, 2026
  • sunblaze-ucb/exploitgym-results's past year of commit activity
    2 6 0 4 Updated Jul 16, 2026