pith. machine review for the scientific record. sign in

arxiv: 2604.27092 · v1 · submitted 2026-04-29 · 💻 cs.AI · physics.optics

Recognition: unknown

End-to-end autonomous scientific discovery on a real optical platform

Shuxing Yang , Fujia Chen , Rui Zhao , Junyao Wu , Yize Wang , Haiyao Luo , Ning Han , Qiaolu Chen
show 5 more authors
Yuze Hu Wenhao Li Mingzhu Li Hongsheng Chen Yihao Yang
Authors on Pith no claims yet

Pith reviewed 2026-05-07 09:46 UTC · model grok-4.3

classification 💻 cs.AI physics.optics
keywords autonomous scientific discoveryLLM agentoptical platformbilinear interactionTransformer attentionphysical mechanismagentic system
0
0 comments X

The pith

An LLM-based agent autonomously proposes and validates a new optical bilinear interaction analogous to Transformer attention.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents Qiushi Engine, an LLM agentic system designed for fully autonomous scientific discovery on a physical optical platform. It maintains long research trajectories across thousands of reasoning and measurement steps, successfully reproducing prior experiments and converting abstract theory into observables. The central result is the agent's independent identification of optical bilinear interaction, a mechanism that enables pairwise light-based computations and mirrors the bilinear structure of attention operations. This outcome matters because it supplies the first reported case of an AI system generating and experimentally confirming a nontrivial physical mechanism without human direction at each step.

Core claim

Through an extended open-ended investigation, Qiushi Engine proposes optical bilinear interaction as a physical process in which light fields interact bilinearly to perform pairwise computations, directly analogous to the attention mechanism in Transformers. The system then designs and executes experiments on a real optical platform that confirm the predicted properties, establishing the mechanism as previously unreported and experimentally viable.

What carries the argument

optical bilinear interaction, the physical process in which optical fields interact in a bilinear manner to enable pairwise computations analogous to attention operations.

If this is right

  • Optical bilinear interaction opens a route to high-speed, energy-efficient hardware that implements pairwise computation directly in light.
  • The same agent architecture can convert abstract coherence-order theory into measurable experimental signatures.
  • Autonomous reproduction of transmission-matrix experiments succeeds on non-original hardware setups.
  • Long-horizon agent stability is maintained across 145.9 million tokens without trajectory collapse.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the autonomy claim holds, similar agent systems could be deployed in other lab domains such as chemistry or materials to hunt for new mechanisms.
  • The structural analogy between the discovered optical process and attention suggests that optical platforms could be engineered to run attention-like operations at the speed of light.
  • Success in this setting implies that agent memory and dual-layer planning may generalize to other experimental physics tasks requiring thousands of iterative measurements.

Load-bearing premise

The agent's entire research trajectory, mechanism proposal, and validation proceed without human-shaped prompts, embedded prior knowledge, or post-hoc interpretation that retroactively frames the outcome as novel.

What would settle it

Publication of the complete unedited logs of the 3,242 LLM calls and 1,242 tool calls, followed by an independent run on the same platform that either fails to propose the same mechanism or yields experimental data inconsistent with bilinear interaction.

Figures

Figures reproduced from arXiv: 2604.27092 by Fujia Chen, Haiyao Luo, Hongsheng Chen, Junyao Wu, Mingzhu Li, Ning Han, Qiaolu Chen, Rui Zhao, Shuxing Yang, Wenhao Li, Yihao Yang, Yize Wang, Yuze Hu.

Figure 2
Figure 2. Figure 2: Autonomous reproduction of a high-impact transmission-matrix experiment on a non￾original optical platform. a, End-to-end research trajectory. Starting from a minimal prompt, the target paper and a basic description of the optical platform, Qiushi Engine develops the reproduction from paper interpretation to experimental execution, data analysis and report generation. b, Agent trajectory across 50 view at source ↗
Figure 3
Figure 3. Figure 3: Autonomous experimental validation of an abstract coherence-order theory on a real optical platform. a, End-to-end research trajectory. Starting from a minimal prompt, the target theory paper and a basic platform description, Qiushi Engine develops the study from theoretical interpretation to view at source ↗
read the original abstract

Scientific research has long been human-led, driving new knowledge and transformative technologies through the continual revision of questions, methods and claims as evidence accumulates. Although large language model (LLM)-based agents are beginning to move beyond assisting predefined research workflows, none has yet demonstrated end-to-end autonomous discovery in a real physical system that produces a nontrivial result supported by experimental evidence. Here we introduce Qiushi Discovery Engine, an LLM-based agentic system for end-to-end autonomous scientific discovery on a real optical platform. Qiushi Engine combines nonlinear research phases, Meta-Trace memory and a dual-layer architecture to maintain adaptive and stable research trajectories across long-horizon investigations involving thousands of LLM-mediated reasoning, measurement and revision actions. It autonomously reproduces a published transmission-matrix experiment on a non-original platform and converts an abstract coherence-order theory into experimental observables, providing, to our knowledge, the first observation of this class of coherence-order structure. More importantly, in an open-ended study involving 145.9 million tokens, 3,242 LLM calls, 1,242 tool calls, 163 research notes and 44 scripts, Qiushi Engine proposes and experimentally validates optical bilinear interaction, a physical mechanism structurally analogous to a core operation in Transformer attention. This AI-discovered mechanism suggests a route towards high-speed, energy-efficient optical hardware for pairwise computation. To our knowledge, this is the first demonstration of an AI agentic system autonomously identifying and experimentally validating a nontrivial, previously unreported physical mechanism, marking a milestone for research-level autonomous agents.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The manuscript introduces the Qiushi Discovery Engine, an LLM-based agentic system for end-to-end autonomous scientific discovery on a real optical platform. It claims the system autonomously reproduces a published transmission-matrix experiment, converts abstract coherence-order theory into experimental observables (providing the first observation of this class of structure), and, in an open-ended 145.9-million-token run involving 3,242 LLM calls, 1,242 tool calls, 163 research notes and 44 scripts, proposes and experimentally validates a new physical mechanism termed optical bilinear interaction that is structurally analogous to a core operation in Transformer attention. The work presents this as the first demonstration of an AI agentic system autonomously identifying and validating a nontrivial, previously unreported physical mechanism with potential implications for high-speed optical hardware.

Significance. If the autonomy, novelty, and experimental claims hold, the result would constitute a milestone for research-level autonomous agents by demonstrating long-horizon operation on physical hardware that yields a new, hardware-relevant physical mechanism. The external grounding provided by independent optical-platform measurements and the reported scale of the investigation (millions of tokens and thousands of calls) are notable strengths that distinguish this from purely simulated or human-guided workflows. The conversion of theory to observables and the suggestion of an optical route to pairwise computation add applied relevance.

major comments (3)
  1. [Abstract / open-ended study description] Abstract and the open-ended study description: The central claim that the agent autonomously proposed optical bilinear interaction, its experimental observables, and the structural analogy to Transformer attention-matrix multiplication rests on an unverified research trajectory. No excerpts from the 3,242 LLM calls, Meta-Trace memory contents, decision tree, or pre- versus post-discovery literature search are supplied, leaving open the possibility that the mechanism or analogy was shaped by human-designed prompts or applied post-hoc.
  2. [Experimental validation] Experimental validation section: The manuscript reports experimental validation of the proposed mechanism on the real optical platform yet supplies no detailed methodology, error bars, data exclusion criteria, or raw experimental traces. This absence prevents independent assessment of whether the validation steps contain post-hoc selection or interpretation, directly affecting the strength of the physical evidence for the new mechanism.
  3. [Discussion of mechanism analogy] Section on the attention analogy: The assertion that optical bilinear interaction is 'structurally analogous to a core operation in Transformer attention' is stated without a formal mathematical mapping, quantitative similarity metric, or side-by-side comparison of the bilinear form to the attention matrix multiplication, making the analogy difficult to evaluate independently of the authors' interpretation.
minor comments (1)
  1. [Methods] The abstract cites precise counts (145.9M tokens, 3,242 LLM calls) but the methods do not clarify the exact accounting rules for tokens, calls, research notes, or scripts, which would aid reproducibility.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their insightful and constructive comments on our manuscript. We have carefully addressed each major point below with point-by-point responses. Where revisions strengthen transparency and rigor, we have updated the manuscript and supplementary materials accordingly.

read point-by-point responses

  1. Referee: Abstract / open-ended study description: The central claim that the agent autonomously proposed optical bilinear interaction, its experimental observables, and the structural analogy to Transformer attention-matrix multiplication rests on an unverified research trajectory. No excerpts from the 3,242 LLM calls, Meta-Trace memory contents, decision tree, or pre- versus post-discovery literature search are supplied, leaving open the possibility that the mechanism or analogy was shaped by human-designed prompts or applied post-hoc.

    Authors: We agree that detailed evidence of the autonomous trajectory is essential to substantiate the claims. While the complete raw log of 3,242 LLM calls is too voluminous for direct inclusion in the main text, we have prepared an expanded supplementary information file containing: (i) representative excerpts from key LLM calls showing the agent's step-by-step reasoning that led to proposing the bilinear interaction; (ii) summaries of Meta-Trace memory states at decision nodes; (iii) the full decision tree of the open-ended exploration; and (iv) the pre- and post-discovery literature search records. These materials demonstrate that the mechanism and analogy emerged from the agent's general exploration directives without targeted human prompting for this specific result. The system prompts remain unchanged from the original submission and are provided verbatim in the SI. We have added explicit cross-references to this SI in the revised abstract and methods sections. revision: partial

  2. Referee: Experimental validation section: The manuscript reports experimental validation of the proposed mechanism on the real optical platform yet supplies no detailed methodology, error bars, data exclusion criteria, or raw experimental traces. This absence prevents independent assessment of whether the validation steps contain post-hoc selection or interpretation, directly affecting the strength of the physical evidence for the new mechanism.

    Authors: We fully concur that comprehensive experimental documentation is required for independent verification. In the revised manuscript we have substantially expanded the Experimental Validation section to include: a complete description of the optical platform configuration and control parameters; the predefined measurement protocol and statistical analysis pipeline; error bars on all quantitative plots derived from repeated trials; explicit data exclusion criteria (signal-to-noise threshold and outlier rejection rules established before data collection); and representative raw time-series traces for both the bilinear-interaction and control experiments. These additions are supported by new figures and tables in the main text plus full raw datasets deposited in the supplementary repository. All validation runs were performed with a fixed protocol to preclude post-hoc selection. revision: yes

  3. Referee: Section on the attention analogy: The assertion that optical bilinear interaction is 'structurally analogous to a core operation in Transformer attention' is stated without a formal mathematical mapping, quantitative similarity metric, or side-by-side comparison of the bilinear form to the attention matrix multiplication, making the analogy difficult to evaluate independently of the authors' interpretation.

    Authors: We appreciate this observation and have strengthened the presentation. A new subsection has been added to the Discussion that supplies: (i) the explicit mathematical mapping showing that the measured optical bilinear interaction corresponds to the outer-product form underlying scaled dot-product attention (Q K^T V); (ii) a quantitative similarity metric defined as the normalized Frobenius distance between the interaction matrix and the attention matrix under identical input statistics; and (iii) a side-by-side tabular comparison of the algebraic structure, computational complexity, and hardware implications. These additions allow readers to assess the analogy on formal grounds independent of our narrative interpretation. revision: yes

Circularity Check

0 steps flagged

No significant circularity; experimental validation is externally grounded

full rationale

The paper's derivation chain consists of an LLM agent proposing a mechanism followed by independent physical measurements on a real optical platform that reproduce a published experiment and observe a coherence-order structure. These measurements constitute external empirical evidence rather than a reduction of the result to the agent's prompts, architecture, or prior self-citations by construction. No equations, fitted parameters renamed as predictions, or load-bearing self-citations are quoted that would make the claimed discovery equivalent to its inputs. The autonomy of the proposal is a separate empirical question about prompt history, but does not create a self-definitional or fitted-input circularity in the reported chain.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the premise that the LLM agent architecture can sustain coherent, adaptive research over thousands of steps without collapsing into repetitive or human-guided loops, with the physical optical platform serving as the sole source of new evidence.

axioms (1)
  • domain assumption LLM-based agents equipped with Meta-Trace memory and dual-layer architecture can maintain stable, non-circular research trajectories across long-horizon investigations.
    Invoked to justify the system's ability to perform 3,242 LLM calls and 1,242 tool calls while producing nontrivial results.

pith-pipeline@v0.9.0 · 5616 in / 1550 out tokens · 88198 ms · 2026-05-07T09:46:59.424784+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

42 extracted references · 4 canonical work pages · 2 internal anchors

  1. [1]

    P., Ho, S., Kohli, P., Lasenby, J., Leskovec, J., Liu, T.- Y., Manrai, A., Marks, D., Ramsundar, B., Song, L., Sun, J., Tang, J., Veličković, P., Welling, M., Zhang, L., Coley, C

    Wang, H., Fu, T., Du, Y., Gao, W., Huang, K., Liu, Z., Chandak, P., Liu, S., Van Katwyk, P., Deac, A., Anandkumar, A., Bergen, K., Gomes, C. P., Ho, S., Kohli, P., Lasenby, J., Leskovec, J., Liu, T.- Y., Manrai, A., Marks, D., Ramsundar, B., Song, L., Sun, J., Tang, J., Veličković, P., Welling, M., Zhang, L., Coley, C. W., Bengio, Y. & Zitnik, M. Scientif...

    2023

  2. [2]

    A., Bradshaw, G

    Langley, P., Simon, H. A., Bradshaw, G. L. & Zytkow, J. M. Scientific Discovery: Computational Explorations of the Creative Process. (MIT Press, 1987). doi:10.7551/mitpress/6090.001.0001

    work page doi:10.7551/mitpress/6090.001.0001 1987

  3. [3]

    & Buchanan, B

    Waltz, D. & Buchanan, B. G. Automating Science. Science 324, 43–44 (2009)

    2009

  4. [4]

    A., MacKnight, R., Kline, B

    Boiko, D. A., MacKnight, R., Kline, B. & Gomes, G. Autonomous chemical research with large language models. Nature 624, 570–578 (2023)

    2023

  5. [5]

    The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery

    Lu, C., Lu, C., Lange, R. T., Foerster, J., Clune, J. & Ha, D. The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery. Preprint at https://doi.org/10.48550/arXiv.2408.06292 (2024)

    work page internal anchor Pith review doi:10.48550/arxiv.2408.06292 2024

  6. [6]

    R., Ektefaie, Y., Kondic, J

    Gao, S., Fang, A., Huang, Y., Giunchiglia, V., Noori, A., Schwarz, J. R., Ektefaie, Y., Kondic, J. & Zitnik, M. Empowering biomedical discovery with AI agents. Cell 187, 6125–6151 (2024)

    2024

  7. [7]

    M., Cox, S., Schilter, O., Baldassari, C., White, A

    Bran, A. M., Cox, S., Schilter, O., Baldassari, C., White, A. D. & Schwaller, P. Augmenting large language models with chemistry tools. Nat. Mach. Intell. 6, 525–535 (2024)

    2024

  8. [8]

    P., Dupont, E., Ruiz, F

    Romera-Paredes, B., Barekatain, M., Novikov, A., Balog, M., Kumar, M. P., Dupont, E., Ruiz, F. J. R., Ellenberg, J. S., Wang, P., Fawzi, O., Kohli, P. & Fawzi, A. Mathematical discoveries from program search with large language models. Nature 625, 468–475 (2024)

    2024

  9. [9]

    Towards an AI co-scientist

    Gottweis, J., Weng, W.-H., Daryin, A., Tu, T., Palepu, A., Sirkovic, P., Myaskovsky, A., Weissenberger, F., Rong, K., Tanno, R., Saab, K., Popovici, D., Blum, J., Zhang, F., Chou, K., Hassidim, A., Gokturk, B., Vahdat, A., Kohli, P., Matias, Y., Carroll, A., Kulkarni, K., Tomasev, N., Guan, Y., Dhillon, V., Vaishnav, E. D., Lee, B., Costa, T. R. D., Penad...

    work page internal anchor Pith review doi:10.48550/arxiv.2502.18864 2025

  10. [10]

    L., Pak, J

    Swanson, K., Wu, W., Bulaong, N. L., Pak, J. E. & Zou, J. The Virtual Lab of AI agents designs new SARS-CoV-2 nanobodies. Nature 646, 716–723 (2025)

    2025

  11. [11]

    C., Lo, K., Soldaini, L., Feldman, S., D’Arcy, M., Wadden, D., Latzke, M., Sparks, J., Hwang, J

    Asai, A., He, J., Shao, R., Shi, W., Singh, A., Chang, J. C., Lo, K., Soldaini, L., Feldman, S., D’Arcy, M., Wadden, D., Latzke, M., Sparks, J., Hwang, J. D., Kishore, V., Tian, M., Ji, P., Liu, S., Tong, H., Wu, B., Xiong, Y., Zettlemoyer, L., Neubig, G., Weld, D. S., Downey, D., Yih, W., Koh, P. W. & Hajishirzi, H. Synthesizing scientific literature wit...

    2026

  12. [12]

    & Xie, W

    Zhao, W., Wu, C., Fan, Y., Qiu, P., Zhang, X., Sun, Y., Zhou, X., Zhang, S., Peng, Y., Wang, Y., Sun, X., Zhang, Y., Yu, Y., Sun, K. & Xie, W. An agentic system for rare disease diagnosis with traceable reasoning. Nature 651, 775–784 (2026)

    2026

  13. [13]

    T., Yamada, Y., Hu, S., Foerster, J., Ha, D

    Lu, C., Lu, C., Lange, R. T., Yamada, Y., Hu, S., Foerster, J., Ha, D. & Clune, J. Towards end-to-end automation of AI research. Nature 651, 914–919 (2026)

    2026

  14. [14]

    M., Gusev, V

    Burger, B., Maffettone, P. M., Gusev, V. V., Aitchison, C. M., Bai, Y., Wang, X., Li, X., Alston, B. M., Li, B., Clowes, R., Rankin, N., Harris, B., Sprick, R. S. & Cooper, A. I. A mobile robotic chemist. Nature 583, 237–241 (2020)

    2020

  15. [15]

    J., Rendy, B., Fei, Y., Kumar, R

    Szymanski, N. J., Rendy, B., Fei, Y., Kumar, R. E., He, T., Milsted, D., McDermott, M. J., Gallant, M., Cubuk, E. D., Merchant, A., Kim, H., Jain, A., Bartel, C. J., Persson, K., Zeng, Y. & Ceder, G. An autonomous laboratory for the accelerated synthesis of inorganic materials. Nature 624, 86–91 (2023)

    2023

  16. [16]

    T., Ayme, J.-F., Simaei, E., Fellowes, T., Clowes, R., Kotopanov, L., Shields, C

    Dai, T., Vijayakrishnan, S., Szczypiński, F. T., Ayme, J.-F., Simaei, E., Fellowes, T., Clowes, R., Kotopanov, L., Shields, C. E., Zhou, Z., Ward, J. W. & Cooper, A. I. Autonomous mobile robots for exploratory synthetic chemistry. Nature 635, 890–897 (2024)

    2024

  17. [17]

    P., Baird, S

    Tom, G., Schmid, S. P., Baird, S. G., Cao, Y., Darvish, K., Hao, H., Lo, S., Pablo-Garcí a, S., Rajaonson, E. M., Skreta, M., Yoshikawa, N., Corapi, S., Akkoc, G. D., Strieth-Kalthoff, F., Seifrid, M. & Aspuru-Guzik, A. Self-Driving Laboratories for Chemistry and Materials Science. Chem. Rev. 124, 9633–9732 (2024). 16

    2024

  18. [18]

    P., Lagendijk, A., Lerosey, G

    Mosk, A. P., Lagendijk, A., Lerosey, G. & Fink, M. Controlling waves in space and time for imaging and focusing in complex media. Nat. Photonics 6, 283–292 (2012)

    2012

  19. [19]

    Wetzstein, G., Ozcan, A., Gigan, S., Fan, S., Englund, D., Soljačić, M., Denz, C., Miller, D. A. B. & Psaltis, D. Inference in artificial intelligence with deep optics and photonics. Nature 588, 39–47 (2020)

    2020

  20. [20]

    J., Tait, A

    Shastri, B. J., Tait, A. N., Ferreira de Lima, T., Pernice, W. H. P., Bhaskaran, H., Wright, C. D. & Prucnal, P. R. Photonics for artificial intelligence and neuromorphic computing. Nat. Photonics 15, 102–114 (2021)

    2021

  21. [21]

    Cao, H., Mosk, A. P. & Rotter, S. Shaping the propagation of light in complex media. Nat. Phys. 18, 994–1007 (2022)

    2022

  22. [22]

    M., Lerosey, G., Carminati, R., Fink, M., Boccara, A

    Popoff, S. M., Lerosey, G., Carminati, R., Fink, M., Boccara, A. C. & Gigan, S. Measuring the Transmission Matrix in Optics: An Approach to the Study and Control of Light Propagation in Disordered Media. Phys. Rev. Lett. 104, 100601 (2010)

    2010

  23. [23]

    Springer Series in Statistics, Springer New York, New York, NY, https://doi.org/10.1007/978-0-387-68276-1

    Marshall, A. W., Olkin, I. & Arnold, B. C. Inequalities: Theory of Majorization and Its Applications. (Springer New York, New York, NY, 2011). doi:10.1007/978-0-387-68276-1

    work page doi:10.1007/978-0-387-68276-1 2011

  24. [24]

    Coherence for vectorial waves and majorization

    Luis, A. Coherence for vectorial waves and majorization. Opt. Lett. 41, 5190 (2016)

    2016

  25. [25]

    & Plenio, M

    Streltsov, A., Adesso, G. & Plenio, M. B. Colloquium : Quantum coherence as a resource. Rev. Mod. Phys. 89, 041003 (2017)

    2017

  26. [26]

    Guo, C., Miller, D. A. B. & Fan, S. Transport Measurements of Majorization Order for Wave Coherence. Phys. Rev. Lett. 135, 053801 (2025)

    2025

  27. [27]

    C., Skirlo, S., Prabhu, M., Baehr-Jones, T., Hochberg, M., Sun, X., Zhao, S., Larochelle, H., Englund, D

    Shen, Y., Harris, N. C., Skirlo, S., Prabhu, M., Baehr-Jones, T., Hochberg, M., Sun, X., Zhao, S., Larochelle, H., Englund, D. & Soljačić, M. Deep learning with coherent nanophotonic circuits. Nat. Photonics 11, 441–446 (2017)

    2017

  28. [28]

    T., Veli, M., Luo, Y., Jarrahi, M

    Lin, X., Rivenson, Y., Yardimci, N. T., Veli, M., Luo, Y., Jarrahi, M. & Ozcan, A. All-optical machine learning using diffractive deep neural networks. Science 361, 1004–1008 (2018). 17

    2018

  29. [29]

    & Englund, D

    Hamerly, R., Bernstein, L., Sludds, A., Soljačić, M. & Englund, D. Large-Scale Optical Neural Networks Based on Photoelectric Multiplication. Phys. Rev. X 9, 021032 (2019)

    2019

  30. [30]

    S., Liu, J., Wright, C

    Feldmann, J., Youngblood, N., Karpov, M., Gehring, H., Li, X., Stappers, M., Le Gallo, M., Fu, X., Lukashchuk, A., Raja, A. S., Liu, J., Wright, C. D., Sebastian, A., Kippenberg, T. J., Pernice, W. H. P. & Bhaskaran, H. Parallel convolutional processing using an integrated photonic tensor core. Nature 589, 52–58 (2021)

    2021

  31. [31]

    McMahon, P. L. The physics of optical computing. Nat. Rev. Phys. 5, 717–734 (2023)

    2023

  32. [32]

    & Fleury, R

    Momeni, A., Rahmani, B., Mallé jac, M., Del Hougne, P. & Fleury, R. Backpropagation-free training of deep physical neural networks. Science 382, 1297–1303 (2023)

    2023

  33. [33]

    N., Kaiser, L

    Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L. & Polosukhin, I. Attention is All you Need. in Advances in Neural Information Processing Systems vol. 30 (Curran Associates, Inc., 2017)

    2017

  34. [34]

    & Zhang, B.-T

    Kim, J.-H., Jun, J. & Zhang, B.-T. Bilinear attention networks. in Proceedings of the 32nd International Conference on Neural Information Processing Systems 1571–1581 (Curran Associates Inc., Red Hook, NY, USA, 2018)

    2018

  35. [35]

    Vellekoop, I. M. & Mosk, A. P. Focusing coherent light through opaque strongly scattering media. Opt. Lett. 32, 2309 (2007)

    2007

  36. [36]

    & Fan, S

    Guo, C. & Fan, S. Majorization Theory for Unitary Control of Optical Absorption and Emission. Phys. Rev. Lett. 130, 146202 (2023)

    2023

  37. [37]

    & Fan, S

    Guo, C. & Fan, S. Unitary control of partially coherent waves. I. Absorption. Phys. Rev. B 110, 035430 (2024)

    2024

  38. [38]

    & Fan, S

    Guo, C. & Fan, S. Unitary control of partially coherent waves. II. Transmission or reflection. Phys. Rev. B 110, 035431 (2024)

    2024

  39. [39]

    & Wolf, E

    Mandel, L. & Wolf, E. Coherence Properties of Optical Fields. Rev. Mod. Phys. 37, 231–287 (1965)

    1965

  40. [40]

    & Gigan, S

    Rafayelyan, M., Dong, J., Tan, Y., Krzakala, F. & Gigan, S. Large-Scale Optical Reservoir Computing for Spatiotemporal Chaotic Systems Prediction. Phys. Rev. X 10, 041037 (2020). 18

    2020

  41. [41]

    & Chu, D

    Zhang, Z., You, Z. & Chu, D. Fundamentals of phase-only liquid crystal on silicon (LCOS) devices. Light Sci. Appl. 3, e213–e213 (2014)

    2014

  42. [42]

    & Boyd, R

    Bolduc, E., Bent, N., Santamato, E., Karimi, E. & Boyd, R. W. Exact solution to simultaneous intensity and phase encryption with a single phase-only hologram. Opt. Lett. 38, 3546–3549 (2013). 19 Figure 1. Qiushi Engine architecture. a, Dual-layer multi-agent architecture. Qiushi Engine consists of a core research agent system, a support research agent sys...

    2013