Hugging Face's logo

Spaces:
A-HK
/
agentic-security-lab
Sleeping

App Files Community
agentic-security-lab

Commit History

  1. Upload folder using huggingface_hub

    A-HK
    Verified signature
    a7c990b

  2. Update README: add Round 2 generalization results, procedural eval data, train_continuation.py

    A-HK
    Verified signature
    45ae1a6

  3. Rewrite BLOG.md: add Round 2 generalization results, full technical depth, honest limitations

    A-HK
    Verified signature
    62732ff

  4. Upload folder using huggingface_hub

    A-HK
    Verified signature
    e43736b

  5. Upload folder using huggingface_hub

    A-HK
    Verified signature
    97524be

  6. Add continuation training script: procedural scenarios, generalization eval

    A-HK
    Verified signature
    4bbb603

  7. Rewrite README with full technical implementation details, research references, real training numbers

    A-HK
    Verified signature
    9e217f9

  8. Upload folder using huggingface_hub

    A-HK
    Verified signature
    98dce4a

  9. Upload folder using huggingface_hub

    A-HK
    Verified signature
    23d9813

  10. Clean up: remove build artifacts, debug logs, old notebook, internal docs

    A-HK
    Verified signature
    c4e8937

  11. Upload folder using huggingface_hub

    A-HK
    Verified signature
    6c62fb1

  12. Fix plot crash: filter None values from metrics before np.convolve

    A-HK
    Verified signature
    4728f59

  13. Upload artifacts/baseline_vs_trained.png with huggingface_hub

    A-HK
    Verified signature
    df3e8b5

  14. Upload artifacts/training_health.png with huggingface_hub

    A-HK
    Verified signature
    e695c2f

  15. Upload artifacts/loss_curve.png with huggingface_hub

    A-HK
    Verified signature
    b6d4d4d

  16. Upload artifacts/reward_curve.png with huggingface_hub

    A-HK
    Verified signature
    075873a

  17. Fix plot NoneType bug in rewards/entropy lists"

    A-HK
    Verified signature
    b0e5cb2

  18. Redesigned train_job.py: fix reward signal (100x), DAPO-style GRPO, curriculum learning

    A-HK
    Verified signature
    daceb77

  19. Upload redesigned train_job.py header (full upload via job)

    A-HK
    Verified signature
    5e7e929

  20. Upload folder using huggingface_hub

    A-HK
    Verified signature
    280be27

  1. Upload folder using huggingface_hub

    A-HK
    Verified signature
    9a03233

  2. Upload folder using huggingface_hub

    A-HK
    Verified signature
    22c2a42

  3. Upload folder using huggingface_hub

    A-HK
    Verified signature
    6ff212f

  4. Upload folder using huggingface_hub

    A-HK
    Verified signature
    589f1cb

  5. Delete round2-hackathon-spec.txt

    A-HK
    Verified signature
    ccc27e0

  6. Delete OpenEnv Hackathon - Provided Resources and Guidelines.pdf

    A-HK
    Verified signature
    8abfce0

  7. Add anti-reward-hacking test suite: scan_logs spam, false positive exploitation, conclude gaming, repeat action exploit"

    A-HK
    Verified signature
    8a72c51

  8. Fix train_grpo.py: rename SFT path, add actual GRPO CLI entrypoint with wired-in procedural gen + trajectory filter + adversarial attacker"

    A-HK
    Verified signature
    5214eea

  9. Upload folder using huggingface_hub

    A-HK
    Verified signature
    f876b60

  10. Rewrite README as clear project blog with all features documented"

    A-HK
    Verified signature
    95852d4

  11. Fix: set response_schema for Qwen2.5 tool-calling + add jmespath dep

    A-HK
    Verified signature
    598c4c8

  12. Remove temp file

    A-HK
    Verified signature
    dd45c68

  13. Fix: remove max_prompt_length (removed in TRL 1.x, Unsloth rejects it)

    A-HK
    Verified signature
    93ce912

  14. Remove max_prompt_length from GRPOConfig (removed in TRL 1.x, Unsloth rejects it)"

    A-HK
    Verified signature
    f6cacba

  15. Fix dtype='auto' -> None for Unsloth compatibility (T4 requires float16, Unsloth auto-detects with None)"

    A-HK
    Verified signature
    62e0566

  16. Fix total_mem bug + integrate procedural gen, security APIs, StarPO-S, adversarial attacker into notebook"

    A-HK
    Verified signature
    2b7135f

  17. Add adversarial attacker for self-play curriculum\n\nReference: SPIRAL (arXiv 2506.24119)\nAdaptive attacker that adds decoys, obscures IOCs, tightens deadlines,\nand exploits defender blind spots. Creates automatic difficulty scaling."

    A-HK
    Verified signature
    93a36f8

  18. Add StarPO-S trajectory filtering for GRPO stability\n\nReference: RAGEN (arXiv 2504.20073)\nFilters to keep only informative rollouts, preventing echo trap collapse."

    A-HK
    Verified signature
    60fa6a8

  19. Add real-world security API integrations (OSV.dev, deps.dev, GHSA, npm)\n\nFree, no-auth-required live lookups against security databases.\nAdds genuine information-gathering vs time-pressure tradeoff."

    A-HK
    Verified signature
    581cdcc

  20. Add procedural scenario generation (infinite training variation)\n\nGenerates unique supply-chain incidents from parameterized distributions.\nReference: Self-Evolving Curriculum (arXiv 2505.14970)"

    A-HK
    Verified signature
    1053202

  21. Replace notebook with complete GRPO training pipeline (TRL environment_factory + Unsloth QLoRA)

    A-HK
    Verified signature
    5d6d233

  22. Upload folder using huggingface_hub

    A-HK
    Verified signature
    4f73b89

  1. Upload folder using huggingface_hub

    A-HK
    Verified signature
    bccfde3

  2. Upload folder using huggingface_hub

    A-HK
    Verified signature
    cd43a41

  3. Upload folder using huggingface_hub

    A-HK
    Verified signature
    60f4610

  4. Upload folder using huggingface_hub

    A-HK
    Verified signature
    b25217e

  5. Upload folder using huggingface_hub

    A-HK
    Verified signature
    26bb614

  6. Upload folder using huggingface_hub

    A-HK
    Verified signature
    193594b

  7. Upload folder using huggingface_hub

    A-HK
    Verified signature
    580d6c5

  8. Upload folder using huggingface_hub

    A-HK
    Verified signature
    5080156