Emerging ethical approaches have attempted to filter pretraining material, but such approaches have been ad hoc and failed to take context into account. We offer an approach to filtering grounded in law, which has directly addressed the tradeoffs in filtering material.

The23 proposed SSCM does coverthe case of non-zero variance, but currently the identifiability proof is only shown in a24 specific case. Inour simulations under non-zero variance settings, we neverobserved that the procedure converged25 to wrong solutions, suggesting that the non-zero-variance case is also identifiable. For the fMRI and cellular data, the null hypothesis was rejected at significance level 0.01. Regarding causal28 structure variation, for fMRI data, it is well-known that neural connectivities may change across different external29 stimuliorintrinsicstates. Forcellular32 data, causal structure may be different across conditions/interventions.(0)Theyare different.

We thank the reviewers for the thorough feedbacks. Based on those, we have made numerous improvements. Original code is for decrete actions.) IM baseline with the random object. The plot is similar to "tool" in Figure 1 and we omit it due to space constraints. Rev. #1 suggested that the environments could be solved by classic planning methods.

Details of the layers can be found in the supplement Table S.2. We will add another, larger figure illustrating the18 U-shape and thethree parts ofthearchitecture. Thesleepstage23 scores indeed show human interpretable patterns even on short timescales. We regard it as a big41 advantage thatwedidnotextensivelytune ourarchitec-42 ture to the tasks. Wewill improvethe colour palette as suggested.

We thank the reviewer for these references, they are very relevant6 and will be added to the manuscript. Rev #1 also suggests to add experiments in order to compare UMNNs with7 thismethod. Instead,weargue18 that other neural architectures for density estimation do so in a way that "leads to a cap on the expressiveness" of19 thetransformations inthenon-asymptotic case(finitenumber ofneurons). Then, this pass must be evaluated backward again in order to38 obtainthelog-likelihood derivative. BothNAFandB-NAFprovideamethod tomakethiscomputation numerically39 stable, however both fail at not increasing the size of the computation graph of the log-likelihood derivative, hence40 leadingtoamemoryoverhead.