AIRE-Prune: Asymptotic Impulse-Response Energy for State Pruning in State Space Models

Apurba Prasad Padhy, Fernando Camacho & Saibal Mukhopadhyay · 2026

Read AIRE-Prune: Asymptotic Impulse-Response Energy for State Pruning in State Space Models by Apurba Prasad Padhy, Fernando Camacho & Saibal Mukhopadhyay (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-00534.txt.

Description

Apurba Prasad Padhy and colleagues prune state space models after training with AIRE-Prune, cutting each layer's state size by minimising output-energy distortion. Each state gets an energy score, its total impulse-response energy over an infinite horizon, normalised for cross-layer comparison. The approach extends modal truncation from single systems to stacks, favouring asymptotic response energy over worst-case gain. Across benchmarks it finds redundancy in SISO and MIMO models, pruning 60.8% on average with only a small accuracy drop and no retraining, while cutting compute.

Catalog record

CREATORApurba Prasad Padhy, Fernando Camacho & Saibal Mukhopadhyay
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE55392
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602.00534

Browse the full library · About the Tower of Babel