JetCoRD: Reliability-Gated Correction of Jet Taggers over Two Frozen Teachers
Abstract
Knowledge distillation for jet taggers is usually cast as compression toward a near-oracle teacher. We instead train a small reliability-gated correction layer jointly over two frozen teachers --- a GN2-type ATLAS flavour tagger (3 classes) and CMS ParT (10 classes) --- reading only their frozen penultimate embeddings and logits, so it adds negligible marginal cost. A single per-sample reliability signal reweights the distillation loss, drives a prototype-based repair of the teachers' errors, and anchors the inference-time gate that mixes teacher and student logits. Zero-initialised, the 82\,k-parameter student starts as a bit-exact copy of the teacher --- a provable invariant on the training distribution --- and departs only where the data support it. Under a statistically-grounded evaluation, with binomial, Poisson and paired uncertainties on every table, it reproduces both teachers at ≥ 99% agreement and matches their overall accuracy while redistributing it toward their weakest classes: the JetClass per-class accuracy of H → 4q, QCD and H → bb rises by +0.8, +0.4 and +0.2 percentage points (|z|>7), offset by comparable drops elsewhere. Oracle heads on the frozen embedding recover a budget consistent with zero in both domains, and a conformal-risk-control analysis certifies the false-accept rate. This full re-evaluation supersedes an earlier, incorrectly-evaluated set of numbers; code and models are available at https://github.com/sysu17363020/JetCoRD.
Related
- Autonomous Thermal Switching Outperforms the Frozen Optimum with Vanishing Correlations
- Circular orbits and particle collisions in the space-time of boson stars and their frozen states
- Frozen-In Gravitational Fields
- Glassy dynamics of metropolitan human mobility frozen near the gravitational equilibrium
- Microscopic Investigation of α Clustering for the α Decay Based on the Density-Constrained Frozen Hartree-Fock Calculations
BibTeX
@misc{gao2026jetcord,
title = {{JetCoRD: Reliability-Gated Correction of Jet Taggers over Two Frozen Teachers}},
author = {Liu-Long Gao and Zheng-Kun Huang and Xiao-Wei Jiang and Jia-Feng Li and Gong-Xing Sun},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2607.15531},
eprint = {2607.15531},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2607.15531},
}
RIS
TY - UNPB
TI - JetCoRD: Reliability-Gated Correction of Jet Taggers over Two Frozen Teachers
AU - Liu-Long Gao
AU - Zheng-Kun Huang
AU - Xiao-Wei Jiang
AU - Jia-Feng Li
AU - Gong-Xing Sun
PY - 2026
DA - 2026/11/03
PB - arXiv
DO - 10.48550/arxiv.2607.15531
UR - https://arxiv.org/abs/2607.15531
AB - Knowledge distillation for jet taggers is usually cast as compression toward a near-oracle teacher. We instead train a small reliability-gated correction layer jointly over two frozen teachers --- a GN2-type ATLAS flavour tagger (3 classes) and CMS ParT (10 classes) --- reading only their frozen penultimate embeddings and logits, so it adds negligible marginal cost. A single per-sample reliability signal reweights the distillation loss, drives a prototype-based repair of the teachers' errors, and anchors the inference-time gate that mixes teacher and student logits. Zero-initialised, the 82\,k-parameter student starts as a bit-exact copy of the teacher --- a provable invariant on the training distribution --- and departs only where the data support it. Under a statistically-grounded evaluation, with binomial, Poisson and paired uncertainties on every table, it reproduces both teachers at ≥ 99% agreement and matches their overall accuracy while redistributing it toward their weakest classes: the JetClass per-class accuracy of H → 4q, QCD and H → bb rises by +0.8, +0.4 and +0.2 percentage points (|z|>7), offset by comparable drops elsewhere. Oracle heads on the frozen embedding recover a budget consistent with zero in both domains, and a conformal-risk-control analysis certifies the false-accept rate. This full re-evaluation supersedes an earlier, incorrectly-evaluated set of numbers; code and models are available at https://github.com/sysu17363020/JetCoRD.
ER -