# Reading Why: Born-Measured Attention with Exact Word Evidence on Light and Qubits

DevFeed: [Reading Why: Born-Measured Attention with Exact Word Evidence on Light and Qubits](<https://devfeed.tech/articles/reading-why-born-measured-attention-with-exact-word-evidence-on-light-and-qubits-117060.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.05355>)

Author: Teerawat Tongloy, Jiraporn Pongsopa, Rangsan Jomtarak

Published: 2026-11-04T05:00:00Z

Content type: news

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Transformer](<https://devfeed.tech/topics/transformer.md>), [Quantum Computing](<https://devfeed.tech/topics/quantum-computing.md>)

Tags: [attention](<https://devfeed.tech/tags/attention.md>), [circuit](<https://devfeed.tech/tags/circuit.md>), [class](<https://devfeed.tech/tags/class.md>), [coherent](<https://devfeed.tech/tags/coherent.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [error](<https://devfeed.tech/tags/error.md>), [gradient](<https://devfeed.tech/tags/gradient.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [ibm](<https://devfeed.tech/tags/ibm.md>), [inference](<https://devfeed.tech/tags/inference.md>), [input](<https://devfeed.tech/tags/input.md>), [language-model](<https://devfeed.tech/tags/language-model.md>), [layer](<https://devfeed.tech/tags/layer.md>), [light](<https://devfeed.tech/tags/light.md>), [measurement](<https://devfeed.tech/tags/measurement.md>), [natural-language](<https://devfeed.tech/tags/natural-language.md>), [optics](<https://devfeed.tech/tags/optics.md>), [physics-optics](<https://devfeed.tech/tags/physics-optics.md>), [quant-ph](<https://devfeed.tech/tags/quant-ph.md>), [qubits](<https://devfeed.tech/tags/qubits.md>), [tests](<https://devfeed.tech/tags/tests.md>), [word](<https://devfeed.tech/tags/word.md>)

## AI overview

The paper presents a transformer built from Born measurements that provides exact word-level evidence under fixed attention denominators: integrated-gradient contributions sum to each class probability above its no-information share, with reported numerical error at or below 10⁻¹¹. On e-SNLI, its 76.98% test accuracy is comparable to a same-size softmax transformer. Deleting the top-ranked evidence word changes its prediction in 45% of pairs, versus 32% for the softmax model using the same explanation method; attention rankings change predictions in 32%, despite better alignment with annotator-highlighted words. The model's weights compile to optical and qubit implementations. On IBM's ibm_kingston, Hadamard tests reproduced predictions and deletion behavior on two short pairs, but the word-level evidence did not yet survive hardware measurement; the abstract says it analyzes the shots and calibration needed.

## Source excerpt

arXiv:2611.05355v1 Announce Type: cross Abstract: When a language model answers, we can usually only estimate why. We build a transformer in which every step is a Born measurement - attention $|\langle q|k\rangle|^4$ normalised over each row, a measured product in place of the feed-forward nonlinearity, and a Born readout. With its denominators held at their values for the input, each class probability is a polynomial in the word amplitudes, so its integrated-gradient evidence is computed by a finite quadrature with no quadrature error: the words' contributions sum to the class probability above its no-information share, to $10^{-11}$ or better. On e-SNLI natural language inference, with about 98k parameters outside the embedding, the model reaches $76.98 \pm 0.17$% test accuracy, level with a softmax transformer of the same size ($76.62 \pm 0.36$%). Deleting the word its evidence ranks first changes the answer in 45% of test pairs, out of the 53% any single deletion can change, against 32% for the softmax model explained the same way and 40% by its attention x gradient. Attention can look more like the words annotators highlighted, yet accounts for the answer less well: deleting the word it ranks first changes the Born Transformer's answer in 32% of pairs. The same weights compile exactly to optoelectronic optics and to qubits, and a coherent attention layer becomes one circuit equal to the model (852 CNOTs on 23 qubits). On IBM's ibm_kingston, reading only the attention overlaps by Hadamard tests, the device gave the model's answers on two short test pairs and kept or changed them under deletion as the model does; the evidence itself did not yet survive the device, and we work out the shots and calibration that measuring it on hardware would need.