Pith. sign in

REVIEW 1 cited by

Progressive Ensemble Distillation: Building Ensembles for Efficient Inference

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2302.10093 v2 pith:HQRS7RUR submitted 2023-02-20 cs.LG

Progressive Ensemble Distillation: Building Ensembles for Efficient Inference

classification cs.LG
keywords ensemblemodelsinferenceb-distilcostdistillationensemblesmodel
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

We study the problem of progressive ensemble distillation: Given a large, pretrained teacher model $g$, we seek to decompose the model into smaller, low-inference cost student models $f_i$, such that progressively evaluating additional models in this ensemble leads to improved predictions. The resulting ensemble allows for flexibly tuning accuracy vs. inference cost at runtime, which is useful for a number of applications in on-device inference. The method we propose, B-DISTIL , relies on an algorithmic procedure that uses function composition over intermediate activations to construct expressive ensembles with similar performance as $g$ , but with smaller student models. We demonstrate the effectiveness of B-DISTIL by decomposing pretrained models across standard image, speech, and sensor datasets. We also provide theoretical guarantees in terms of convergence and generalization.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Gedanken Experiments of Entanglement in Particle Physics: Interactions, Operators and Bell Inequalities in Flavor Space

    hep-ph 2025-12 conditional novelty 4.0

    Mass-identification, kaon-decay, and weak-mixing observables can be cast as spin-like operators whose correlations violate a Bell-type bound (0.44+0.90=1.34>1) in an idealized Gedanken framework.