3 papers
cs.LG2026
Mechanistic Evidence for Spectral Structures in Prior-Data Fitted Networks
Kaustubh Sharma, Srijan Tiwari, Ojasva Nema +1
Prior-Data Fitted Networks (PFNs) enable amortized Bayesian inference in a single forward pass, yet their internal representations remain opaque. It is unknown whether PFNs encode…
cs.LG2026
Structural Disentanglement in Bilinear MLPs via Architectural Inductive Bias
Ojasva Nema, Kaustubh Sharma, Aditya Chauhan +1
Selective unlearning and long-horizon extrapolation remain fragile in modern neural networks, even when tasks have underlying algebraic structure. In this work, we argue that these…
cs.LG2025
Decoupled-Value Attention for Prior-Data Fitted Networks: GP Inference for Physical Equations
Kaustubh Sharma, Simardeep Singh, Parikshit Pareek
Prior-data fitted networks (PFNs) are a promising alternative to time-consuming Gaussian process (GP) inference for creating fast surrogates of physical systems. PFN reduces the co…