Pith. sign in

REVIEW 2 major objections 2 minor 53 references

On photonic chips, shadow tomography error stops falling with more samples and freezes at a hardware-set floor.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-14 22:24 UTC pith:UQOOACND

load-bearing objection Wrong full text is attached; the Hardware Horizon claim is only an abstract, so the experimental phase-transition result cannot be checked. the 2 major comments →

arxiv 2603.12235 v2 pith:UQOOACND submitted 2026-03-12 quant-ph cond-mat.dis-nncs.ETphysics.optics

Transition from Statistical to Hardware-Limited Scaling in Photonic Quantum State Reconstruction

Attila Baumann , Zsolt Kis , J\'anos Koltai , G\'abor Vattay This is my paper
classification quant-ph cond-mat.dis-nncs.ETphysics.optics
keywords classical shadow tomographyintegrated photonicsHardware Horizonquantum state reconstructionNISQunitary ensemblespectral distortionphotonic quantum processors
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Classical shadow tomography is supposed to reconstruct quantum states with error that keeps shrinking as one over square-root of the number of random measurements, but only if the random unitaries are perfect Haar-random draws. Real integrated photonic processors cannot produce that ideal ensemble. This paper reports that reconstruction error on such hardware follows the expected statistical improvement only up to a point, then abruptly stops improving and sits at a fixed floor the authors call the Hardware Horizon. That floor is set by spectral distortions in the actual unitaries the chip realizes, not by the usual counting statistics. A phenomenological error model separates those static coherent distortions from ordinary dynamic decoherence and shows that simply taking more shots cannot push past the floor. The practical upshot is that the usefulness of shadow tomography on near-term photonic hardware is governed by a hardware-parameter scaling law, so active compensation is required if theory-level accuracy is the goal.

Core claim

Reconstruction error in classical shadow tomography on integrated photonic processors undergoes a sharp phase transition: it first scales as the predicted statistical O(M^{-1/2}), then saturates at an intrinsic accuracy floor (the Hardware Horizon) fixed by spectral distortions of the realized unitary group; statistical accumulation alone cannot overcome that floor.

What carries the argument

The Hardware Horizon: the accuracy floor at which error stops obeying statistical scaling and becomes limited by spectral distortions of the physical unitary ensemble, together with a phenomenological error model that separates static coherent spectral distortion from dynamic decoherence.

Load-bearing premise

The saturation is caused by static coherent spectral distortion of the realized unitaries, cleanly separated from other hardware effects by the phenomenological model.

What would settle it

On the same photonic processor, reconstruction error continues to fall as 1/sqrt(M) with no measurable floor, or the measured floor fails to track independently characterized spectral distortions of the unitary ensemble while tracking other systematics instead.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Utility of shadow tomography on NISQ photonic hardware is set by a scaling law that includes hardware parameters, not by sample count alone.
  • Simply increasing the number of random measurements cannot reach theoretically ideal reconstruction accuracy once the Hardware Horizon is reached.
  • Active compensation or calibration strategies that correct spectral distortions of the unitary ensemble become necessary to close the gap to ideal performance.
  • Device characterization of the realized unitary group’s spectral properties becomes a first-class requirement for predicting shadow-tomography accuracy.

Where Pith is reading between the lines

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

  • If the Hardware Horizon is generic to imperfect unitary ensembles, similar floors should appear in other physical platforms (superconducting, trapped-ion) once measurement budgets exceed the statistical regime.
  • Benchmarking protocols for photonic processors may need to report the location of the Horizon (in sample number and error level) alongside usual fidelity metrics.
  • Compensation schemes that actively reshape the effective unitary ensemble could turn the Horizon into a tunable design parameter rather than a hard limit.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 2 minor

Summary. The materials under review present a title and abstract claiming an experimental discovery on integrated photonic processors: classical-shadow reconstruction error undergoes a sharp transition from statistical O(M^{-1/2}) scaling to a hard accuracy floor (the “Hardware Horizon”) set by spectral distortions of the realized unitary ensemble, with a phenomenological model that decouples static coherent spectral distortion from dynamic decoherence. The supplied full manuscript body, however, is an unrelated preprint on LLM multi-agent teams framed as distributed systems (Amdahl’s law, centralized vs. decentralized coordination, consistency conflicts, stragglers, and token-cost tradeoffs). No photonic hardware description, unitary characterization, shadow-tomography protocol, scaling curves, error model derivation, or controls appear in the body.

Significance. If the abstract’s claims were substantiated—an experimentally demonstrated phase transition from statistical to hardware-limited scaling in photonic shadow tomography, plus a cleanly decoupled phenomenological model—they would be of clear interest to the quant-ph and NISQ communities, clarifying practical limits of classical shadows on near-term photonic processors and motivating active compensation. Those claims cannot be assessed from the supplied manuscript, so significance remains conditional on a correct, complete submission.

major comments (2)
  1. Title/abstract vs. full text mismatch: the body is the LLM-teams distributed-systems manuscript (sections on Amdahl’s Law, Experiments 1–2, Figs. 2–5, Table 1, coding benchmarks MathUtils20/DataAnalysis/SVGRendering). None of the load-bearing content required by the abstract—photonic processor, Haar-random vs. realized unitaries, spectral distortion measurements, reconstruction-error vs. M curves, or the phenomenological error model—is present. The central claim of a Hardware Horizon therefore cannot be evaluated.
  2. Absent experimental and theoretical support for the abstract’s mechanism: there are no methods, controls (calibration drift, detector systematics), raw scaling data, error bars, or derivation showing that saturation is caused by static coherent spectral distortion of the unitary group rather than other systematics, and no demonstration that the model cleanly separates that effect from dynamic decoherence. Without these, the asserted hard bound and scaling law involving hardware parameters are unsupported.
minor comments (2)
  1. The abstract uses the coined term “Hardware Horizon” without definition in any matching body text; if a correct manuscript is resubmitted, define it formally against measured spectral quantities.
  2. arXiv identifiers in the materials (2603.12235 vs. body content consistent with 2603.12229) should be reconciled so referees receive the intended paper.

Circularity Check

0 steps flagged

No circularity identifiable: supplied full text is a different paper (LLM teams / distributed systems), so the Hardware Horizon derivation chain cannot be walked; abstract alone shows no definitional loop.

full rationale

The claimed paper (arXiv:2603.12235) asserts an experimental phase transition of classical-shadow reconstruction error from O(M^{-1/2}) to a hardware floor set by spectral distortions of the realized unitary ensemble, plus a phenomenological model that decouples static coherent spectral distortion from dynamic decoherence. Circularity would require showing that this floor or model is defined from, or fitted to and then re-presented as predicting, the same reconstruction-error saturation. The CACHEABLE full manuscript is not that work: it is “Language Model Teams as Distributed Systems” (arXiv:2603.12229), with Amdahl-style speedup experiments, consistency/overhead/straggler measurements, and no photonic unitaries, shadow tomography, spectral characterization, or error-model equations. From the photonic abstract alone, the floor is attributed to measured spectral distortions of realized unitaries rather than defined as the reconstruction error itself, so there is no self-definitional reduction. Residual risk that a phenomenological model was fitted to the same saturation it “explains” cannot be confirmed or refuted without equations and fitting procedure; under the hard rule that circularity may be claimed only with a quoted reduction, that risk is not scored. The attached LLM-teams manuscript is itself an empirical analogy study (preassigned vs decentralized coding tasks vs Amdahl bounds) and does not present a first-principles derivation that collapses to its inputs. Result: no circular steps can be exhibited; score 0.

Axiom & Free-Parameter Ledger

2 free parameters · 4 axioms · 1 invented entities

Abstract-only review of a quantum-photonics experiment. Load-bearing content is experimental and phenomenological; almost no formal axioms are stated. Free parameters and invented entities are inferred from the abstract's claims about a noise floor and error model.

free parameters (2)
  • Hardware noise-floor / spectral-distortion parameters in the phenomenological error model
    Abstract states the floor is determined by spectral distortions of the realized unitary group; those distortion strengths are hardware parameters that must be measured or fitted to set the bound.
  • Relative weights of static coherent distortion vs dynamic decoherence
    The model is said to decouple these mechanisms; their relative contributions are not fixed by theory alone and would be set by data.
axioms (4)
  • domain assumption Ideal classical shadow tomography efficiency assumes a perfect Haar-random unitary ensemble.
    Stated as the theoretical premise the experiment departs from.
  • domain assumption Near-term integrated photonic processors cannot physically realize that ideal ensemble; residual spectral distortions and decoherence remain.
    Core physical premise of the Hardware Horizon claim.
  • domain assumption Reconstruction error initially follows statistical scaling O(M^{-1/2}) before saturating.
    Standard statistical expectation used as the pre-transition baseline.
  • ad hoc to paper A phenomenological error model can cleanly separate static coherent spectral distortion from dynamic decoherence as competing mechanisms.
    Abstract presents this decoupling as derived; without the body it is an uncheckable modeling assumption of the paper.
invented entities (1)
  • Hardware Horizon no independent evidence
    purpose: Name the claimed sharp phase transition / accuracy floor where statistical scaling ends and hardware spectral distortion dominates.
    Terminological framing of the observed saturation; independent evidence would be the experimental transition itself, which is not inspectable here.

pith-pipeline@v1.1.0-grok45 · 20485 in / 2465 out tokens · 25026 ms · 2026-07-14T22:24:01.119131+00:00 · methodology

0 comments
read the original abstract

The theoretical efficiency of classical shadow tomography is predicated on a perfect Haar-random unitary ensemble, yet this mathematical ideal remains physically unattainable in near-term hardware. Here, we report the experimental discovery of a fundamental accuracy bound on integrated photonic processors: a ``Hardware Horizon'' where the reconstruction error undergoes a sharp phase transition. While the error initially obeys the predicted statistical scaling $\mathcal{O}(M^{-1/2})$, it abruptly saturates at a floor determined by the spectral distortions of the realized unitary group. By deriving a phenomenological error model, we decouple the competing mechanisms of static coherent spectral distortion and dynamic decoherence, demonstrating that this intrinsic noise floor imposes a hard bound that statistical accumulation cannot overcome. These findings establish that the utility of shadow tomography on NISQ (noisy intermediate-scale quantum) hardware is defined by a specific scaling law involving hardware parameters, necessitating active compensation strategies to bridge the gap between theoretical purity and the noisy reality of integrated photonics.

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

53 extracted references · 14 linked inside Pith

  1. [1]

    Pearson Education, 1995

    Frederick P Brooks Jr.The Mythical Man-Month: Essays on Software Engineering. Pearson Education, 1995

    1995

  2. [2]

    Computing machinery and intelligence.Mind, 59(236):433, 1950

    Alan Turing. Computing machinery and intelligence.Mind, 59(236):433, 1950

    1950

  3. [3]

    The secret of our success: How culture is driving human evolution, domesticating our species, and making us smarter

    Joseph Henrich. The secret of our success: How culture is driving human evolution, domesticating our species, and making us smarter. Princeton University Press, 2015

    2015

  4. [4]

    Understanding and sharing intentions: The origins of cultural cognition.Behavioral and Brain Sciences, 28(5):675–691, 2005

    Michael Tomasello, Malinda Carpenter, Josep Call, Tanya Behne, and Henrike Moll. Understanding and sharing intentions: The origins of cultural cognition.Behavioral and Brain Sciences, 28(5):675–691, 2005

    2005

  5. [5]

    When should we orchestrate multiple agents? arXiv preprint arXiv:2503.13577, 2025

    Umang Bhatt, Sanyam Kapoor, Mihir Upadhyay, Ilia Sucholutsky, Francesco Quinzan, Katherine M Collins, Adrian Weller, Andrew Gordon Wilson, and Muhammad Bilal Zafar. When should we orchestrate multiple agents? arXiv preprint arXiv:2503.13577, 2025

    Pith/arXiv arXiv 2025

  6. [6]

    A survey on the distributed computing stack.Computer Science Review, 42:100422, 2021

    Cristian Ramon-Cortes, Pol Alvarez, Francesc Lordan, Javier Alvarez, Jorge Ejarque, and Rosa M Badia. A survey on the distributed computing stack.Computer Science Review, 42:100422, 2021

    2021

  7. [7]

    distributed-systems.net, 2023

    Maarten Van Steen and Andrew S Tanenbaum.Distributed Systems. distributed-systems.net, 2023

    2023

  8. [8]

    Understanding Agent Scaling in LLM-Based Multi-Agent Systems via Diversity.arXiv preprint arXiv:2602.03794, 2026

    Yingxuan Yang, Chengrui Qu, Muning Wen, Laixi Shi, Ying Wen, Weinan Zhang, Adam Wierman, and Shangding Gu. Understanding Agent Scaling in LLM-Based Multi-Agent Systems via Diversity.arXiv preprint arXiv:2602.03794, 2026

    arXiv 2026

  9. [9]

    Towards a science of scaling agent systems.arXiv preprint arXiv:2512.08296, 2025

    Yubin Kim, Ken Gu, Chanwoo Park, Chunjong Park, Samuel Schmidgall, A Ali Heydari, Yao Yan, Zhihan Zhang, Yuchen Zhuang, Mark Malhotra, et al. Towards a science of scaling agent systems.arXiv preprint arXiv:2512.08296, 2025

    Pith/arXiv arXiv 2025

  10. [10]

    Multi-agent teams hold experts back.arXiv preprint arXiv:2602.01011, 2026

    Aneesh Pappu, Batu El, Hancheng Cao, Carmelo di Nolfo, Yanchao Sun, Meng Cao, and James Zou. Multi-agent teams hold experts back.arXiv preprint arXiv:2602.01011, 2026

    Pith/arXiv arXiv 2026

  11. [11]

    Agents of chaos.arXiv preprint arXiv:2602.20021, 2026

    Natalie Shapira, Chris Wendler, Avery Yen, Gabriele Sarti, Koyena Pal, Olivia Floody, Adam Belfki, Alex Loftus, Aditya Ratan Jannali, Nikhil Prakash, et al. Agents of chaos.arXiv preprint arXiv:2602.20021, 2026

    Pith/arXiv arXiv 2026

  12. [12]

    Cooperate or collapse: Emergence of sustainable cooperation in a society of LLM agents.Advances in Neural Information Processing Systems, 37:111715–111759, 2024

    Giorgio Piatti, Zhijing Jin, Max Kleiman-Weiner, Bernhard Schölkopf, Mrinmaya Sachan, and Rada Mihalcea. Cooperate or collapse: Emergence of sustainable cooperation in a society of LLM agents.Advances in Neural Information Processing Systems, 37:111715–111759, 2024. 11 PREPRINT

    2024

  13. [13]

    Position: Science is Collaborative—LLM for Science Should Be Too

    Terry Jingchen Zhang, Wenyuan Jiang, Yongjin Yang, Yinya Huang, Sirui Lu, Bernhard Schölkopf, and Zhijing Jin. Position: Science is Collaborative—LLM for Science Should Be Too

  14. [14]

    Chatdev: Communicative agents for software development

    Chen Qian, Wei Liu, Hongzhang Liu, Nuo Chen, Yufan Dang, Jiahao Li, Cheng Yang, Weize Chen, Yusheng Su, Xin Cong, et al. Chatdev: Communicative agents for software development. InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics, pages 15174–15186, 2024

    2024

  15. [15]

    How we built our multi-agent research system

    Anthropic. How we built our multi-agent research system. �������������������������������������� ���������������������������, June 2025. Anthropic Engineering Blog

    2025

  16. [16]

    Machine behaviour.Nature, 568(7753):477–486, 2019

    Iyad Rahwan, Manuel Cebrian, Nick Obradovich, Josh Bongard, Jean-François Bonnefon, Cynthia Breazeal, Jacob W Crandall, Nicholas A Christakis, Iain D Couzin, Matthew O Jackson, et al. Machine behaviour.Nature, 568(7753):477–486, 2019

    2019

  17. [17]

    Evaluating collective behaviour of hundreds of llm agents.arXiv preprint arXiv:2602.16662, 2026

    Richard Willis, Jianing Zhao, Yali Du, and Joel Z Leibo. Evaluating collective behaviour of hundreds of llm agents.arXiv preprint arXiv:2602.16662, 2026

    Pith/arXiv arXiv 2026

  18. [18]

    The Virtual Lab of AI agents designs new SARS-CoV-2 nanobodies.Nature, 646(8085):716–723, 2025

    Kyle Swanson, Wesley Wu, Nash L Bulaong, John E Pak, and James Zou. The Virtual Lab of AI agents designs new SARS-CoV-2 nanobodies.Nature, 646(8085):716–723, 2025

    2025

  19. [19]

    Autonomous chemical research with large language models.Nature, 624(7992):570–578, 2023

    Daniil A Boiko, Robert MacKnight, Ben Kline, and Gabe Gomes. Autonomous chemical research with large language models.Nature, 624(7992):570–578, 2023

    2023

  20. [20]

    Social Agents: Collective Intelligence Improves LLM Predictions

    Aanisha Bhattacharyya, Abhilekh Borah, Yaman Kumar Singla, Rajiv Ratn Shah, Changyou Chen, and Balaji Krishnamurthy. Social Agents: Collective Intelligence Improves LLM Predictions. InThe Fourteenth International Conference on Learning Representations

  21. [21]

    Improving factuality and reasoning in language models through multiagent debate

    Yilun Du, Shuang Li, Antonio Torralba, Joshua B Tenenbaum, and Igor Mordatch. Improving factuality and reasoning in language models through multiagent debate. InForty-first International Conference on Machine Learning, 2024

    2024

  22. [22]

    Chain of agents: Large language models collaborating on long-context tasks.Advances in Neural Information Processing Systems, 37:132208–132237, 2024

    Yusen Zhang, Ruoxi Sun, Yanfei Chen, Tomas Pfister, Rui Zhang, and Sercan Arik. Chain of agents: Large language models collaborating on long-context tasks.Advances in Neural Information Processing Systems, 37:132208–132237, 2024

    2024

  23. [23]

    More agents is all you need.arXiv preprint arXiv:2402.05120, 2024

    Junyou Li, Qin Zhang, Yangbin Yu, Qiang Fu, and Deheng Ye. More agents is all you need.arXiv preprint arXiv:2402.05120, 2024

    Pith/arXiv arXiv 2024

  24. [24]

    How do large language models navigate conflicts between honesty and helpfulness?arXiv preprint arXiv:2402.07282, 2024

    Ryan Liu, Theodore R Sumers, Ishita Dasgupta, and Thomas L Griffiths. How do large language models navigate conflicts between honesty and helpfulness?arXiv preprint arXiv:2402.07282, 2024

    Pith/arXiv arXiv 2024

  25. [25]

    MetaGPT: Meta programming for a multi-agent collaborative framework

    Sirui Hong, Mingchen Zhuge, Jonathan Chen, Xiawu Zheng, Yuheng Cheng, Jinlin Wang, Ceyao Zhang, Zili Wang, Steven Ka Shing Yau, Zijuan Lin, et al. MetaGPT: Meta programming for a multi-agent collaborative framework. InThe Twelfth International Conference on Learning Representations, 2023

    2023

  26. [26]

    Agentverse: Facilitating multi-agent collaboration and exploring emergent behaviors

    Weize Chen, Yusheng Su, Jingwei Zuo, Cheng Yang, Chenfei Yuan, Chi-Min Chan, Heyang Yu, Yaxi Lu, Yi-Hsin Hung, Chen Qian, et al. Agentverse: Facilitating multi-agent collaboration and exploring emergent behaviors. In The Twelfth International Conference on Learning Representations, 2023

    2023

  27. [27]

    Autogen: Enabling next-gen LLM applications via multi-agent conversations

    Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Beibin Li, Erkang Zhu, Li Jiang, Xiaoyun Zhang, Shaokun Zhang, Jiale Liu, et al. Autogen: Enabling next-gen LLM applications via multi-agent conversations. InFirst Conference on Language Modeling, 2024

    2024

  28. [28]

    Embodied LLM agents learn to cooperate in organized teams.arXiv preprint arXiv:2403.12482, 2024

    Xudong Guo, Kaixuan Huang, Jiale Liu, Wenhui Fan, Natalia Vélez, Qingyun Wu, Huazheng Wang, Thomas L Griffiths, and Mengdi Wang. Embodied LLM agents learn to cooperate in organized teams.arXiv preprint arXiv:2403.12482, 2024

    Pith/arXiv arXiv 2024

  29. [29]

    Owl: Optimized workforce learning for general multi-agent assistance in real-world task automation.arXiv preprint arXiv:2505.23885, 2025

    Mengkang Hu, Yuhang Zhou, Wendong Fan, Yuzhou Nie, Bowei Xia, Tao Sun, Ziyu Ye, Zhaoxuan Jin, Yingru Li, Qiguang Chen, et al. Owl: Optimized workforce learning for general multi-agent assistance in real-world task automation.arXiv preprint arXiv:2505.23885, 2025

    Pith/arXiv arXiv 2025

  30. [30]

    Can LLM Agents Really Debate? A Controlled Study of Multi-Agent Debate in Logical Reasoning.arXiv preprint arXiv:2511.07784, 2025

    Haolun Wu, Zhenkun Li, and Lingyao Li. Can LLM Agents Really Debate? A Controlled Study of Multi-Agent Debate in Logical Reasoning.arXiv preprint arXiv:2511.07784, 2025

    arXiv 2025

  31. [31]

    Scaling large language model-based multi-agent collaboration.arXiv preprint arXiv:2406.07155, 2024

    Chen Qian, Zihao Xie, Yifei Wang, Wei Liu, Kunlun Zhu, Hanchen Xia, Yufan Dang, Zhuoyun Du, Weize Chen, Cheng Yang, et al. Scaling large language model-based multi-agent collaboration.arXiv preprint arXiv:2406.07155, 2024

    Pith/arXiv arXiv 2024

  32. [32]

    G-designer: Architecting multi-agent communication topologies via graph neural networks

    Guibin Zhang, Yanwei Yue, Xiangguo Sun, Guancheng Wan, Miao Yu, Junfeng Fang, Kun Wang, Tianlong Chen, and Dawei Cheng. G-designer: Architecting multi-agent communication topologies via graph neural networks. arXiv preprint arXiv:2410.11782, 2024. 12 PREPRINT

    Pith/arXiv arXiv 2024

  33. [33]

    Validity of the single processor approach to achieving large scale computing capabilities

    Gene M Amdahl. Validity of the single processor approach to achieving large scale computing capabilities. In Proceedings of the April 18-20, 1967, Spring Joint Computer Conference, pages 483–485, 1967

    1967

  34. [34]

    Reevaluating Amdahl’s law.Communications of the ACM, 31(5):532–533, 1988

    John L Gustafson. Reevaluating Amdahl’s law.Communications of the ACM, 31(5):532–533, 1988

    1988

  35. [35]

    A general theory of computational scalability based on rational functions.arXiv preprint arXiv:0808.1431, 2008

    Neil J Gunther. A general theory of computational scalability based on rational functions.arXiv preprint arXiv:0808.1431, 2008

    Pith/arXiv arXiv 2008

  36. [36]

    Amdahl’s law in the multicore era.Computer, 41(7):33–38, 2008

    Mark D Hill and Michael R Marty. Amdahl’s law in the multicore era.Computer, 41(7):33–38, 2008

    2008

  37. [37]

    How to make a multiprocessor computer that correctly executes multiprocess programs.IEEE transactions on computers, 100(9):690–691, 1979

    Lamport. How to make a multiprocessor computer that correctly executes multiprocess programs.IEEE transactions on computers, 100(9):690–691, 1979

    1979

  38. [38]

    Linearizability: A correctness condition for concurrent objects.ACM Transactions on Programming Languages and Systems (TOPLAS), 12(3):463–492, 1990

    Maurice P Herlihy and Jeannette M Wing. Linearizability: A correctness condition for concurrent objects.ACM Transactions on Programming Languages and Systems (TOPLAS), 12(3):463–492, 1990

    1990

  39. [39]

    Elsevier, 1996

    Nancy A Lynch.Distributed algorithms. Elsevier, 1996

    1996

  40. [40]

    MapReduce: Simplified data processing on large clusters.Communications of the ACM, 51(1):107–113, 2008

    Jeffrey Dean and Sanjay Ghemawat. MapReduce: Simplified data processing on large clusters.Communications of the ACM, 51(1):107–113, 2008

    2008

  41. [41]

    Improving MapReduce performance in heterogeneous environments

    Matei Zaharia, Andy Konwinski, Anthony D Joseph, Randy H Katz, and Ion Stoica. Improving MapReduce performance in heterogeneous environments. InOsdi, volume 8, page 7, 2008

    2008

  42. [42]

    Using straggler replication to reduce latency in large-scale parallel computing.ACM SIGMETRICS Performance Evaluation Review, 43(3):7–11, 2015

    Da Wang, Gauri Joshi, and Gregory Wornell. Using straggler replication to reduce latency in large-scale parallel computing.ACM SIGMETRICS Performance Evaluation Review, 43(3):7–11, 2015

    2015

  43. [43]

    Energy dissipation in general purpose microprocessors.IEEE Journal of solid-state circuits, 31(9):1277–1284, 1996

    Ricardo Gonzalez and Mark Horowitz. Energy dissipation in general purpose microprocessors.IEEE Journal of solid-state circuits, 31(9):1277–1284, 1996

    1996

  44. [44]

    Beyond Amdahl’s law: An objective function that links multiprocessor performance gains to delay and energy.IEEE Transactions on Computers, 61(8):1110–1126, 2011

    Andrew S Cassidy and Andreas G Andreou. Beyond Amdahl’s law: An objective function that links multiprocessor performance gains to delay and energy.IEEE Transactions on Computers, 61(8):1110–1126, 2011

    2011

  45. [45]

    Characterizing resource allocation heuristics for heterogeneous computing systems

    Shoukat Ali, Tracy D Braun, Howard J Siegel, Anthony A Maciejewski, Noah Beck, Ladislau Bölöni, Muthu- cumaru Maheswaran, Albert I Reuther, James P Robertson, Mitchell D Theys, and Bin Yao. Characterizing resource allocation heuristics for heterogeneous computing systems. InAdvances in Computers, volume 63, pages 91–128. Elsevier, 2005

    2005

  46. [46]

    The implementation of reliable distributed multiprocess systems.Computer Networks (1976), 2(2):95–114, 1978

    Leslie Lamport. The implementation of reliable distributed multiprocess systems.Computer Networks (1976), 2(2):95–114, 1978

    1976

  47. [47]

    Dependable computing and fault-tolerance.Digest of Papers FTCS-15, 10(2):124, 1985

    Jean-Claude Laprie. Dependable computing and fault-tolerance.Digest of Papers FTCS-15, 10(2):124, 1985

    1985

  48. [48]

    Hallucination is inevitable: An innate limitation of large language models.arXiv preprint arXiv:2401.11817, 2024

    Ziwei Xu, Sanjay Jain, and Mohan Kankanhalli. Hallucination is inevitable: An innate limitation of large language models.arXiv preprint arXiv:2401.11817, 2024

    Pith/arXiv arXiv 2024

  49. [49]

    Why do multi-agent LLM systems fail?arXiv preprint arXiv:2503.13657, 2025

    Mert Cemri, Melissa Z Pan, Shuyi Yang, Lakshya A Agrawal, Bhavya Chopra, Rishabh Tiwari, Kurt Keutzer, Aditya Parameswaran, Dan Klein, Kannan Ramchandran, et al. Why do multi-agent LLM systems fail?arXiv preprint arXiv:2503.13657, 2025

    Pith/arXiv arXiv 2025

  50. [50]

    A comparison of eleven static heuristics for mapping a class of independent tasks onto heterogeneous distributed computing systems

    Tracy D Braun, Howard J Siegel, Noah Beck, Ladislau L Bölöni, Muthucumaru Maheswaran, Albert I Reuther, James P Robertson, Mitchell D Theys, Bin Yao, Debra Hensgen, and Richard F Freund. A comparison of eleven static heuristics for mapping a class of independent tasks onto heterogeneous distributed computing systems. Journal of Parallel and Distributed Co...

    2001

  51. [51]

    Static task allocation of concurrent programs for distributed computing systems with processor and resource heterogeneity.Parallel Computing, 20(6):835–851, 1994

    S Selvakumar and C Siva Ram Murthy. Static task allocation of concurrent programs for distributed computing systems with processor and resource heterogeneity.Parallel Computing, 20(6):835–851, 1994

    1994

  52. [52]

    Performance analysis of load balancing algorithms

    Sandeep Sharma, Sarabjit Singh, and Meenakshi Sharma. Performance analysis of load balancing algorithms. World Academy of Science, Engineering and Technology, 38(3):269–272, 2008

    2008

  53. [53]

    Arpaci-Dusseau Books, LLC Madison, WI, USA, 2018

    Remzi H Arpaci-Dusseau and Andrea C Arpaci-Dusseau.Operating Systems: Three Easy Pieces, volume 1. Arpaci-Dusseau Books, LLC Madison, WI, USA, 2018. A Appendix A.1 Empirical setup Agent architectureEach experiment used a team of �∈ {���������} teammate agents, all backed by the same LLM (GPT-5.2, Gemini-3-Flash, or Claude-Sonnet-4-6), with temperature ���...

    2018