---
_id: '7152'
abstract:
- lang: eng
text: "Property prediction for molecules and materials is bottlenecked by label
scarcity, and in this regime the operative failure mode is not a low mean but
an unpredictable tail: individual training runs collapse. The three dominant molecular
representations fail in complementary ways?physicochemical descriptors are exact
but fixed, graph neural networks learn topology but degenerate under sparse supervision,
and language models supply semantic context but cannot compute exact quantities?
yet combining them is itself the learning problem, because modality-level scalar
gating commits the whole model to one trust weight per source. We introduce TAME,
a tri-modal encoder whose element-wise Mixture-of-Experts router assigns an independent
modality mixture to every hidden coordinate, stabilized by a balance?entropy regularizer,
a closed-form gate initialization (? = log Ï\x84 ) that is intended to remove
router burn-in, and two-stage self-supervised graph pretraining. On scaffold-split
BACE (â\x89\x88 1.5k molecules, 100 seeds) we evaluate two fusion topologies,
each against its own graph-only control. In both, adding the text and descriptor
experts contracts the seed-to-seed distribution of the threshold-free metrics?ROC-AUC
Ï\x83 falls five-fold in the flat topology and 1.6-fold in the hierarchical one,
whose control was already the tighter of the two?and the collapse tail reaching
ROC-AUC â\x89\x88 0.45 that the weaker single-modality configurations carry is
absent from the fused models while accuracy is retained. Each comparison is matched
on everything but the fusion stage, isolating its effect. The router allocates
a distinct mixture to each coordinate, which no scalar gate can represent, and
the entropy weight moves that allocation continuously between graded mixing and
a near-binary regime in which each coordinate is claimed outright by one modality.
Fused-representation alignment tracks encoder quality without supervision, shifting
from text to topology once pretraining lifts the graph embedding out of rank-1
collapse. The same three experts map onto crystal graphs, composition?structure
descriptors and literature-mined text. These three modalities recur unchanged
across molecular and inorganic chemistry, so TAME is a single interpretable recipe
for turning volatile low-data predictors into reliable, self-explaining ones,
wherever data are scarce and every failed prediction costs a real experiment."
author:
- first_name: Robert P.
full_name: Schiller, Robert P.
last_name: Schiller
- first_name: Christoph
full_name: Weisser, Christoph
id: '264138'
last_name: Weisser
orcid: 0000-0003-0616-1027
orcid_put_code_url: https://api.orcid.org/v2.0/0000-0003-0616-1027/work/226396719
- first_name: Klaus-Robert
full_name: Müller, Klaus-Robert
last_name: Müller
- first_name: Christian
full_name: Ochsenfeld, Christian
last_name: Ochsenfeld
- first_name: Parastoo
full_name: Semnani, Parastoo
last_name: Semnani
citation:
alphadin: 'Schiller, Robert P. ; Weisser, Christoph ; Müller,
Klaus-Robert ; Ochsenfeld, Christian
; Semnani, Parastoo: TAME: Element-wise
Mixture-of-Experts Fusion for Reliable and Interpretable Molecular Property Prediction,
2026'
ama: 'Schiller RP, Weisser C, Müller K-R, Ochsenfeld C, Semnani P. TAME: Element-Wise
Mixture-of-Experts Fusion for Reliable and Interpretable Molecular Property Prediction.;
2026. doi:10.26434/chemrxiv.15008270/v2'
apa: 'Schiller, R. P., Weisser, C., Müller, K.-R., Ochsenfeld, C., & Semnani,
P. (2026). TAME: Element-wise Mixture-of-Experts Fusion for Reliable and Interpretable
Molecular Property Prediction. https://doi.org/10.26434/chemrxiv.15008270/v2'
bibtex: '@book{Schiller_Weisser_Müller_Ochsenfeld_Semnani_2026, title={TAME: Element-wise
Mixture-of-Experts Fusion for Reliable and Interpretable Molecular Property Prediction},
DOI={10.26434/chemrxiv.15008270/v2},
author={Schiller, Robert P. and Weisser, Christoph and Müller, Klaus-Robert and
Ochsenfeld, Christian and Semnani, Parastoo}, year={2026} }'
chicago: 'Schiller, Robert P., Christoph Weisser, Klaus-Robert Müller, Christian
Ochsenfeld, and Parastoo Semnani. TAME: Element-Wise Mixture-of-Experts Fusion
for Reliable and Interpretable Molecular Property Prediction, 2026. https://doi.org/10.26434/chemrxiv.15008270/v2.'
ieee: 'R. P. Schiller, C. Weisser, K.-R. Müller, C. Ochsenfeld, and P. Semnani,
TAME: Element-wise Mixture-of-Experts Fusion for Reliable and Interpretable
Molecular Property Prediction. 2026.'
mla: 'Schiller, Robert P., et al. TAME: Element-Wise Mixture-of-Experts Fusion
for Reliable and Interpretable Molecular Property Prediction. 2026, doi:10.26434/chemrxiv.15008270/v2.'
short: 'R.P. Schiller, C. Weisser, K.-R. Müller, C. Ochsenfeld, P. Semnani, TAME:
Element-Wise Mixture-of-Experts Fusion for Reliable and Interpretable Molecular
Property Prediction, 2026.'
date_created: 2026-09-10T07:31:07Z
date_updated: 2026-09-11T08:11:21Z
doi: 10.26434/chemrxiv.15008270/v2
language:
- iso: eng
main_file_link:
- open_access: '1'
oa: '1'
publication_status: published
status: public
title: 'TAME: Element-wise Mixture-of-Experts Fusion for Reliable and Interpretable
Molecular Property Prediction'
type: working_paper
user_id: '220548'
year: '2026'
...