When Attention Collapses: Residual Evidence Modeling for Compositional Inference

TL;DR

This paper introduces residual evidence modeling with evidence depletion to prevent slot collapse in attention-based compositional inference, improving the separation of latent components in complex observations.

Contribution

The paper proposes a novel residual evidence modeling technique that addresses the structural failure mode of slot collapse in attention models under additive superposition.

Findings

Evidence depletion reduces slot collapse by up to an order of magnitude.

The method generalizes beyond synthetic benchmarks to real-world audio and gravitational-wave data.

Standard attention fails in additive superposition scenarios, while evidence depletion succeeds.

Abstract

Compositional inference - the decomposition of observations into an unknown number of latent components - is central to perception and scientific data analysis. Attention-based models perform well when components are approximately separable, as in object-centric vision. Under additive superposition, however - where multiple components contribute to every observation - we identify a structural failure mode we term slot collapse: multiple slots converge to the same dominant component while weaker ones remain unrepresented. We trace this to a general limitation: attention is memoryless with respect to explained evidence. All slots repeatedly operate on the same input without accounting for what has already been explained, so gradients are dominated by the strongest component, inducing shared fixed points across slots. As a result, attention fails to enforce non-redundant allocation under additive superposition. We address this by introducing residual evidence modeling, instantiated via evidence depletion - a minimal modification combining multiplicative depletion with an attention bias. Controlled ablations show that parallel attention, sequential processing alone, and loss-based regularization fail to resolve collapse; evidence depletion, which adds residual state to sequential attention, consistently succeeds. Across synthetic benchmarks and real-world audio mixtures (FUSS), evidence depletion reduces slot collapse by up to an order of magnitude, generalizing beyond synthetic settings. On gravitational-wave source inference for the ESA/NASA LISA mission, under identical architectures, data, and losses, standard attention fails while evidence depletion prevents collapse and enables multi-source posterior estimation. These results show that under additive superposition, residual evidence tracking is the operative ingredient for preventing collapse and enabling compositional inference.