# Synthetic Dataset

After prompt injection attacks neared saturation on public benchmarks, an optimizer that runs a swarm of collaborating agents to search for attack payloads pushed success back up. The study built 100 synthetic tool-use scenarios whose attack goal is a random tool call unrelated to the user task; against the strongest attacker, Gemini 4 Argon, the most robust defender, Gemini 4 Argon, resisted only 66% of attacks at the full query budget and 87% at 150 queries.

After prompt injection attacks neared saturation on public benchmarks, an optimizer that runs a swarm of collaborating agents to search for attack payloads pushed success back up. The study built 100 synthetic tool-use scenarios whose attack goal is a random tool call unrelated to the user task; against the strongest attacker, Gemini 4 Argon, the most robust defender, Gemini 4 Argon, resisted only 66% of attacks at the full query budget and 87% at 150 queries.

Sparse autoencoders (SAEs) are widely used to pull interpretable features from biological foundation-model embeddings, and reproducibility across random seeds has served as the trust test. On embeddings with known concepts, single SAE fits reproduced only 27–46% of their features yet covered most concepts, while a consensus over five fits raised reproducibility to 78–98% at single-fit purity. The consensus covered fewer lineages than single fits, and all SAEs missed weak cell states.

Sparse autoencoders (SAEs) are widely used to pull interpretable features from biological foundation-model embeddings, and reproducibility across random seeds has served as the trust test. On embeddings with known concepts, single SAE fits reproduced only 27–46% of their features yet covered most concepts, while a consensus over five fits raised reproducibility to 78–98% at single-fit purity. The consensus covered fewer lineages than single fits, and all SAEs missed weak cell states.

Synthetic electronic health records have long traded clinical fidelity against formal privacy guarantees. SynthEHR-DP combines a mixed-type encoder, a temporal transformer encoder and a conditional variational autoencoder, trained under differentially private stochastic gradient descent with Rényi privacy accounting, then applies a post-hoc clinical constraint layer that spends no privacy budget. On MIMIC-IV at a total budget of ε=8 and δ=10⁻⁵, membership inference AUC is 0.52, downstream TSTR AUROC is 0.786, 4.2% below the non-private TRTR oracle of 0.821, and the clinical validity score is 0.961.

Synthetic electronic health records have long traded clinical fidelity against formal privacy guarantees. SynthEHR-DP combines a mixed-type encoder, a temporal transformer encoder and a conditional variational autoencoder, trained under differentially private stochastic gradient descent with Rényi privacy accounting, then applies a post-hoc clinical constraint layer that spends no privacy budget. On MIMIC-IV at a total budget of ε=8 and δ=10⁻⁵, membership inference AUC is 0.52, downstream TSTR AUROC is 0.786, 4.2% below the non-private TRTR oracle of 0.821, and the clinical validity score is 0.961.