Search papers, labs, and topics across Lattice.
This study investigates the identifiability of physical parameters in latent world models through controlled interventions in the POKEWORLD environment, revealing how prediction objectives influence which physical quantities are internalized. The research identifies that while inputs limit knowledge, the targets of predictions significantly dictate what is retained in the latent representation, with notable findings such as stiffness being captured only when touch is predicted. Additionally, the study highlights that certain parameters, like drag, exhibit high recoverability but plateau in performance under deterministic prediction objectives, suggesting a complex relationship between data structure and parameter acquisition.
Prediction objectives shape the physical knowledge embedded in latent representations, revealing that some parameters are only captured under specific forecasting conditions.
A central premise of latent world models is that predicting the future forces a representation to internalize the physics of its environment. Which physical quantities does a trained latent actually contain, and what decides this? We answer with controlled interventions in POKEWORLD, an interactive environment whose visually identical objects hide mass, drag, and contact stiffness. A certificate-gated protocol first certifies each parameter as recoverable from raw observations, then measures whether it enters the latent, so a null result can be attributed to the objective rather than to the environment. The resulting identifiability map has two organizing mechanisms and one frontier. Inputs limit what can be known, while prediction targets decide what is retained. Stiffness enters the latent only when touch is forecast ($R^2=0.50$, compared with $-0.02$ when the same signal is merely fused into the input), and under single-step prediction a vision-only latent discards even perfectly visible object state. Drag marks the frontier. It carries a recoverability certificate of 0.89 yet plateaus near 0.13 under every deterministic prediction objective we test, while a supervised head on the same trunk reaches 0.45. Parameters whose readout is slow and ratio-type under the sensed coordinates fall outside what these objectives acquire. On RH20T, an input-target factorial across scaling curves reproduces both mechanisms across two robots and 4,258 episodes. Every arm missing information or prediction pressure stays flat over a fivefold data range, and only the full multimodal objective forecasts force beyond a persistence baseline, with held-out gains that grow with scale. Objective structure determines which physical parameters a latent acquires, and additional data improves only the parameters it already acquires.
CMU ML