Commit History

Today

  1. Declare licence: cc-by-nc-4.0 (derivative of tiny-aya-global)

    Verified signature
    a532deb
  2. Model card: link the 72-episode attack gallery

    Verified signature
    756a526
  3. Model card: 208 injected tasks across six splits, 13/208 in the ledger, 0/208 in answers

    Verified signature
    6c69c90
  4. Upload README.md with huggingface_hub

    Verified signature
    2da4364

3 days ago

  1. Model card: second safety sample and payee probe (0 attacker values on 112 injected tasks)

    Verified signature
    898df72

4 days ago

  1. AuditKIT: note the 1.1.6 re-check (identical verdicts)

    Verified signature
    2328f48
  2. Usage: load in fp16 so the snippet does not OOM a 16 GB card

    Verified signature
    c625046
  3. State the oracle call count per split

    Verified signature
    2ad96fb
  4. Embed the before/after GIF above the results table

    Verified signature
    72b3867
  5. Add live demo / demo video / write-up links

    Verified signature
    ba89f17
  6. Upload README.md with huggingface_hub

    Verified signature
    056e692

7 days ago

  1. Upload README.md with huggingface_hub

    Verified signature
    1c63784
  2. Upload README.md with huggingface_hub

    Verified signature
    6767e98
  3. Upload README.md with huggingface_hub

    Verified signature
    978cdeb

8 days ago

  1. docs: link the live demo and the write-up

    Verified signature
    d09b737
  2. docs: measured results, safety split and reproducibility notes

    Verified signature
    ccfbb65