BayesFlow is a library for amortized Bayesian inference with neural networks.
⋅ Multi-backend via Keras 3: Use PyTorch, TensorFlow, or JAX.
⋅ Modern nets: Flow matching, diffusion, consistency models, normalizing flows, transformers
⋅ Built-in diagnostics and plotting
🔗 https://github.com/bayesflow-org/bayesflow
⋅ Multi-backend via Keras 3: Use PyTorch, TensorFlow, or JAX.
⋅ Modern nets: Flow matching, diffusion, consistency models, normalizing flows, transformers
⋅ Built-in diagnostics and plotting
🔗 https://github.com/bayesflow-org/bayesflow
Comments
am trying to develop options for probabilistic firewalls
Q: what is/are the best security measure(s) that you are aware of to help stop or mitigate probabilistic injection ?
the simplest form of probabilistic injection is a ‘prompt injection’
In the context of amortized inference, it’s been shown that many of the algorithms we use are susceptible to adversarial attacks, and this can be mitigated by regularizing wrt Fisher information.
📝 Paper by @mackelab.bsky.social:
https://arxiv.org/abs/2305.14984
https://arxiv.org/html/2409.11445v1
how can your solution offer any protection for this ?
https://arxiv.org/abs/2406.20053
this paper offers no solution for the plethora of probabilistic hacks that have taken place since then …
i could provide many papers for your review if you would like , mathprompt is a good one for ex.
do you have a recent robust examination?
☮️