Tag
This paper introduces strategic interactive oversight (SIO) to study how AI agents can pursue latent objectives while maintaining task performance in debate protocols, emphasizing the need to evaluate oversight beyond verdict correctness.
Introduces Decoupled Latent Optimization (DLO) for full waveform inversion, which relaxes latent optimization into a quadratic-penalty objective, outperforming classical and diffusion-based methods on benchmarks while preserving smoothed-velocity initialization.