arXiv subjects

Peiying Zhu

Publications and source records attributed to Peiying Zhu.

At least 19 recordsLinked to original sources

arXiv2026

When Are Aggregate Agent Traces Diagnosable? Traffic-Governed Interpretation and Calibrated Abstention

Runtime traces can appear transparent, but a closed-loop policy determines which states are visited and which failures become visible. We study a simulated hotel-pricing agent mapping time, inventory, and market state to discrete price actions under varying demand regimes. A fault may leave no aggregate trace when the policy rarely visits affected cells. We treat entry into aggregate-only fault interpretation as a diagnosability decision preceding scoring or localization. A reference-map gate requires repeated clean-policy support; a matched runtime gate then requires joint support in clean and current streams. Signal analysis occurs only after both pass. We calibrate false admission on a disjoint clean stream at the physical-component level and model detection by affected clean traffic rather than nominal cell coverage. In a frozen one-shot heldout, 55/72 (76.4%) regime-component units were reference-admitted, representing 20 physical components; 54/55 passed matched runtime admission, while the rejected unit abstained. Stable false admission was 0/20, with a one-sided exact 95% upper bound of 0.1391, meeting the frozen 0.20 criterion. Across 540 repeated unit-arm rows nested in those 20 clusters, affected clean traffic reduced negative log likelihood by 29.3% relative to cell coverage, a gain of 0.1264 nats per row (cluster-bootstrap 95% interval [0.0593, 0.1918]). Adding mask family and its interaction improved log loss by 0.0015 nats per row (one-sided upper bound 0.0066), below the frozen 0.01 practical-sufficiency margin. A development audit found that exact minimum hitting set and greedy selection chose identical supports in 12/12 scenarios because singleton evidence had resolved the conflicts. The result is a bounded rule for interpreting aggregate agent behavior: first establish exposure, then score change, and abstain when the trace cannot support the claim.

cs.AI↗

arXiv2026

Refuse, Decompose, Refresh: A Claim-Safe Protocol for Closed-Loop AI Evaluation

An AI evaluation can be perfectly reproducible and still support the wrong claim. This risk is acute in closed-loop systems: policy determines visited states, observable components, and which failures leave a measurable trace. We propose a claim-safe protocol with three actions. Refuse: abstain when a clean reference stream or matched runtime comparison lacks support. Decompose: report protocol execution, operational false admission, and structural hypotheses separately rather than as one PASS/FAIL label. Refresh: treat distribution-shift alarms as requests to invalidate and recompute a reference map, not as fault evidence. We instantiate the protocol in an aggregate-only simulator with 24 policy components, three demand regimes, two fault-mask families, and independent development and heldout seeds. The preregistered heldout contains 1,440 cases and 21,600 partition rows. Only 55/72 regime-component units were reference-admitted and 54/55 remained runtime-admitted, making abstention part of the result. Stable false admission was 0/20 represented components, with a one-sided exact 95% upper bound of 0.1391 under a frozen 0.20 rule. Within admitted units, affected clean traffic outpredicted nominal fault-cell fraction: across 540 unit-arm rows nested in 20 component clusters, the cell-minus-traffic negative-log-likelihood difference was 0.1264 nats per row, with a 95% component-cluster interval of [0.0593, 0.1918]. A drift log shows why "null" must be reference-relative: clean fault-null streams triggered 15/15, 0/15, and 14/15 alarms across three regimes, while only the middle regime matched the frozen detector reference. Rather than a universal threshold, we contribute an executable contract linking observable support, statistical calibration, and justified claims.

cs.AI↗

arXiv2026

Four Ledgers, Not One Score: Responsible Communication of LLM-Judge Calibration in Biomedical ML

Synthetic perturbations appear to offer inexpensive calibration data for LLM evaluators in biomedical ML, where expert review is scarce. Yet a planted mutation key is neither a detector output nor automatically human ground truth. We formalize four distinct ledgers: planted perturbations, independent detector outputs, source-linked human dispositions, and human-added discoveries. We then audit the evaluation design, scoring code, read paths, and current human records of a private synthetic Japanese care-handoff workflow. The factory stored 69 planted error cards across 47 targets. Final review covers 22 targets and contains 22 confirmed imported proposals, 9 rejected proposals, and 79 human-added cards; only 3 reviewed targets are double annotated. Passing imported plant keys to a generic detector scorer yields 22/(22+9)=0.710 and 22/(22+79)=0.218. A direct audit identity shows that these values are proposal-confirmation yield and submitted-ledger composition, not judge precision and recall, because no independent detector realization was preserved for the audited proposals in the available records. The audit also finds source-name collisions, row shadowing, forced severity, vacuous ratio defaults, and unsupported zero-support field weights. We contribute a provenance-aware claim audit, a storage contract, and a minimum calibration gate for responsibly communicating biomedical ML capability claims. This single-workflow forensic case is an existence proof of a failure mode, not an estimate of its prevalence: existing human work supports an exploratory audit of synthetic proposals, but not LLM-judge operating characteristics, clinical validity, corpus prevalence, or robust inter-annotator agreement.

cs.AI↗

arXiv2026

ClaimReceipt: Verifying Evidence Sufficiency and Coverage in Agent Evaluations

Agent evaluations face two distinct evidentiary questions: whether a reported claim is recomputable from retained evidence (sufficiency), and whether the retained records cover the committed experiment set (coverage). Generic logs and hash-linked transcripts answer neither reliably. We introduce ClaimReceipt, a claim-relative receipt specification and selective verifier that binds typed transaction evidence to a signed experiment manifest and returns PASS, INVALID, or INCONCLUSIVE per claim. We freeze the specification before implementation (SHA-256 18d109...b81). On 1,392 historical buyer--seller records, a CR-2 verifier reproduces all five manually labeled audit verdicts, exactly replays 600 deterministic and 792 post-generation records, makes every one of 13 declared field groups non-redundant under tested ablations, and returns the expected result on 11/11 semantic faults with 0/8 false positives. We then run a separate prospective CR-3 epoch: 30 assignments are committed before inference, terminal receipts are signed and chained, and private evidence is encrypted for an auditor. Complete evidence yields coverage and accounting PASS; withholding one terminal receipt returns INCONCLUSIVE_COVERAGE, while withholding all private openings preserves coverage and protocol verification but makes economic claims inconclusive, exactly matching a preregistered prediction. Receipt instrumentation adds 0.021% of model-inference time and 9.9 KB per transaction. A specification-legibility probe indicates that our own frozen specification is not yet unambiguous to an independent reader. Claim verification therefore requires both claim-sufficient evidence and a committed universe against which omissions become visible.

cs.AI↗

arXiv2026

When Guardrails Look Effective: Construct Validity Failures in LLM Agent Commerce Evaluation

Interactive simulations increasingly evaluate policies in markets populated by language-model agents. Their outputs can look economic---prices, profits, consumer surplus, and welfare---without instantiating the behavior named in the claim. We audit this risk in a multi-turn buyer--seller testbed for configurable hotel transactions. An initial implementation reported welfare gains from two marketplace guardrails of +87.4, +35.0, and +28.8 across a Qwen2.5 1.5B--14B ladder. It also gave guarded and unguarded agents different offer schemas and choice procedures. Holding the schema and buyer chooser fixed changes the paired contrasts to +7.2, -13.9, and +23.8. The four largest 14B single-generation effects averaged +229; after three generations per profile-condition, they averaged +37.6 (95% bootstrap interval [-34.2, 109.3]), while generation residuals account for 49.9% of variation in this post-hoc probe. A seller-incentive check is non-monotone: increasing profit pressure produces less profit than the default seller prompt. Scripted positive controls show why this matters. A profit-maximizing seller already attains first-best welfare, so guardrails mostly redistribute and reduce welfare; they create welfare only when the seller is explicitly programmed to force inefficient bundles. We contribute a construct-validity contract separating incentive validity, protocol isolation, stochastic stability, and welfare accounting, and returning INVALID or INCONCLUSIVE before substantive policy claims. In our case, the original estimate is INVALID under protocol isolation, while the controlled study remains INCONCLUSIVE under incentive validity and stochastic stability. The case does not show that guardrails are ineffective; it shows their apparent value is unidentified until the simulated agents and protocol pass these checks.

cs.AI↗

arXiv2026

Measurement Risk in Supervised Financial NLP: Rubric and Metric Sensitivity on JF-ICR

As LLMs become credible readers of earnings calls, investor-relations Q\&A, guidance, and disclosure language, supervised financial NLP benchmarks increasingly function as decision evidence for model selection and deployment. A hidden assumption is that gold labels make such evidence objective. This assumption breaks down when the benchmark ruler itself is sensitive to rubric wording, metric choice, or aggregation policy. We study this measurement risk on Japanese Financial Implicit-Commitment Recognition (JF-ICR; a pinned 253-item test split x 4 frontier LLMs x 5 rubrics x 3 temperatures x 5 ordinal metrics). Three findings follow. First, rubric wording materially changes model-assigned labels: R2--R3 agreement ranges from 70.0% to 83.4%, with the dominant movement near the +1 / 0 implicit-commitment boundary. This pattern is consistent with a pragmatic-boundary interpretation, but is not a validated linguistic-causality claim because the present rubric variants confound semantics, examples, and verbosity. Second, not every metric remains informative under the JF-ICR class distribution. Within-one accuracy is too easy because near misses receive credit and the majority class dominates; worst-class accuracy is too noisy because the rarest class has only two examples. Exact accuracy, macro-F1, and weighted \k{appa} are therefore the identifiable metrics under our operational rule. Third, ranking claims become more defensible only after this metric-identifiability audit: Bradley--Terry, Borda, and Ranked Pairs agree on the identifiable metric subset, while the full five-metric sweep produces disagreement on the closest pair. The contribution is not a new leaderboard, but a reporting discipline for supervised financial benchmarks whose gold labels exist and whose evaluation ruler still requires governance.

cs.AI↗

arXiv2026

When Outcome Looks Right But Discipline Fails: Trace-Based Evaluation Under Hidden Competitor State

Outcome-only evaluation can certify economically unsafe agents: a policy can hit a business KPI while violating deployable behavioral discipline. In hotel pricing with hidden competitor state, a learner can achieve plausible revenue per available room while failing to preserve the rate discipline of a rule-based revenue-management competitor. We introduce discipline stability, a trace-based evaluation paradigm: define the benchmark behavior, restrict observations to the deployment regime, induce trace diagnostics from failure, separate mechanisms with ablations, and test transfer and deployment. Across a two-hotel benchmark and a compact hidden-budget bidding task, reward-only PPO variants miss trace alignment; revealing hidden state reduces label uncertainty; deterministic copy collapses uncertainty; and trace-prior or corrected history policies better preserve price or bid distributions. Pure behavior cloning is nearly enough for symmetric imitation, while Trace-Prior RL adds bounded adaptation under capacity asymmetry. The contribution is an evaluation and benchmark paradigm, not a new optimizer or a universal claim about MARL

cs.AI↗

arXiv2026

When Aggregate Alignment Misleads: Auditing Policy Repair Without Per-State Expert Actions

Agentic AI systems are increasingly used to edit, refine, and repair decision policies, but evaluating these edits is difficult when per-state expert action labels are unavailable. We study this problem in a hotel-pricing simulator where an agentic policy editor receives only region-level diagnostic feedback: summaries of how its price distribution differs from a benchmark policy across time, inventory, and market regions. The editor cannot observe benchmark actions, benchmark source code, reward numbers, or held-out outcomes, and may only propose constrained edits to a target-action table. On 5,000 held-out episodes, a multi-restart LLM editor reaches RevPAR 108.47 (95% CI 107.61 - 109.34), close to the benchmark policy's 108.75 (107.81 - 109.68), with paired gap (LLM minus benchmark) -0.276 and 95% CI [-0.692, 0.146]. A cheap diagnostic projection already recovers much of the revenue (107.90), so the LLM editor's distinctive gain is not raw revenue lift alone: it also reduces episode composition distance from 1.153 to 0.609. This is the strongest non-benchmark repair result. This profile is not explained by restart search alone: non-semantic proposers with up to 2,500 evaluations fall 8.77 - 14.57 RevPAR points short. Nor is it explained by plausible prompt format: a shuffled-diagnostic control breaks region-error correspondence and falls to RevPAR 94.30. The match is genuine but partial. A tree editor achieves stronger pooled alignment, 0.214 versus 0.266, and stronger reference-state D1, 0.328 versus 1.197, yet revenue falls to 98.91. These results show that agentic policy repair should be evaluated by whether diagnostic feedback becomes reliable closed-loop outcome, not by a single behavioral distance.

cs.AI↗

arXiv2026

Inference-Driven Uplink for 6G: Architecture, Principles, and Challenges

Next-generation wireless networks (6G) face a critical uplink challenge arising from stringent device-side resource constraints and the growing demand for intelligent services. This article introduces InferCom, an inference-driven uplink architecture designed to enable robust communication under low signal-to-noise (SNR) conditions. It adopts a compute-asymmetric design with a lightweight transmitter and an inference-capable receiver empowered by generative artificial intelligence models. Grounded in the information bottleneck principle, InferCom redefines communications through task-agnostic compression, inference-driven reconstruction, error distribution channel code, and quality of experience-aware retransmission. A case study demonstrates that InferCom outperforms conventional 5G NR and Deep-JSCC in terms of transmitter-side computational complexity, uplink coverage and retransmission efficiency. Finally, we outline key challenges and research directions for inference driven uplink design in future intelligent 6G networks.

eess.SP↗

arXiv2026

Market-Alignment Risk in Pricing Agents: Trace Diagnostics and Trace-Prior RL under Hidden Competitor State

Outcome metrics can certify the wrong behavior. We study this failure in a two-hotel revenue-management simulator where Hotel A trains an agent against a fixed rule-based revenue-management competitor, Hotel B. A standard learning agent can obtain near-reference revenue per available room (RevPAR) while failing to learn market-like yield management: it sells too aggressively, undercuts, or collapses to modal price buckets. We diagnose this as a Goodhart-style failure under partial observability. Hotel A cannot observe the competitor's remaining inventory, booking curve, or pricing rule, so the same Hotel A-visible state maps to multiple plausible Hotel B prices. Deterministic value-based RL and deterministic copying collapse this unresolved uncertainty into shortcut behavior. We introduce a trace-level diagnostic protocol using RevPAR, occupancy, ADR, full price-bucket distributions, L1/JS distances, and seed-level confidence intervals. The verified repair is Trace-Prior RL: learn a distributional market prior from lagged market traces, then train a stochastic pricing policy with a RevPAR reward and a KL penalty to the learned prior. The final policy matches Hotel B's RevPAR, occupancy, ADR, and price distribution within seed-level uncertainty, while still optimizing Hotel A's own reward. We argue that the contribution is not a new optimizer and not a hotel-pricing leaderboard, but a reproducible failure-and-repair recipe for agentic systems where scalar rewards are easy to game and the intended behavior is only visible in traces. A key finding is that higher exact action accuracy can worsen aggregate trace alignment when the target is distributional.

cs.AI↗

arXiv2026

ValueBlindBench: Agreement-Gated Stress Testing of LLM-Judged Investment Rationales Before Returns Are Observable

LLM-based financial agents increasingly produce investment rationales before the outcomes needed to evaluate them are observable. This creates a delayed-ground-truth evaluation problem: realized returns remain the eventual arbiter of investment quality, but they arrive too late and are too noisy to guide many model-development and governance decisions. LLM judges offer a tempting shortcut for pre-deployment evaluation of AI-finance systems, but unvalidated judges may reward verbosity, confidence, or rubric mimicry rather than financial judgment. This paper introduces ValueBlindBench, a preregistered agreement-gated stress-test protocol for deciding when LLM-judged investment-rationale claims are publishable, qualified, or invalid. In a controlled market-state capital-allocation prototype with 1,000 honest decision cycles and 100 preregistered adversarial controls (1,100 trajectories, 5,500 judge calls), ValueBlindBench clears the aggregate agreement gate at $\barκ_w = 0.7168$ but prevents several overclaims. Lower-rank systems collapse into a tie-class, one rubric dimension fails the per-dimension gate (\texttt{constraint\_awareness}, $\barκ_w = 0.2022$), single-judge rankings are family-dependent, and terse-correct rationales receive a $Δ= -2.81$ rubric-point penalty relative to honest rationales. A targeted anchor-specificity probe further shows that financial constructs such as constraint awareness are operationally load-bearing. The scientific object is therefore not a leaderboard and not a claim to measure true investment skill. ValueBlindBench is a pre-calibration metrology layer for AI-finance evaluation: it governs whether a proposed LLM-judge-based investment-rationale claim is stable enough, agreed enough, and uncontaminated enough to be reported at all.

cs.AI↗

arXiv2026

The End of Rented Discovery: How AI Search Redistributes Power Between Hotels and Intermediaries

When a traveler asks an AI search engine to recommend a hotel, which sources get cited -- and does query framing matter? We audit 1,357 grounding citations from Google Gemini across 156 hotel queries in Tokyo and document a systematic pattern we call the Intent-Source Divide. Experiential queries draw 55.9% of their citations from non-OTA sources, compared to 30.8% for transactional queries -- a 25.1 percentage-point gap ($p < 5 \times 10^{-20}$). The effect is amplified in Japanese, where experiential queries draw 62.1% non-OTA citations compared to 50.0% in English -- consistent with a more diverse Japanese non-OTA content ecosystem. For an industry in which hotels have long paid OTAs for demand acquisition, this pattern matters because it suggests that AI search may make hotel discovery less exclusively controlled by commission-based intermediaries.

cs.IR↗

arXiv2024

Signal Processing and Learning for Next Generation Multiple Access in 6G

Wireless communication systems to date primarily rely on the orthogonality of resources to facilitate the design and implementation, from user access to data transmission. Emerging applications and scenarios in the sixth generation (6G) wireless systems will require massive connectivity and transmission of a deluge of data, which calls for more flexibility in the design concept that goes beyond orthogonality. Furthermore, recent advances in signal processing and learning, e.g., deep learning, provide promising approaches to deal with complex and previously intractable problems. This article provides an overview of research efforts to date in the field of signal processing and learning for next-generation multiple access, with an emphasis on massive random access and non-orthogonal multiple access. The promising interplay with new technologies and the challenges in learning-based NGMA are discussed.

eess.SP↗

arXiv2024

A Primer on Near-Field Communications for Next-Generation Multiple Access

Multiple-antenna technologies are advancing toward the development of extremely large aperture arrays and the utilization of extremely high frequencies, driving the progress of next-generation multiple access (NGMA). This evolution is accompanied by the emergence of near-field communications (NFC), characterized by spherical-wave propagation, which introduces additional range dimensions to the channel and enhances system throughput. In this context, a tutorial-based primer on NFC is presented, emphasizing its applications in multiuser communications and multiple access (MA). The following areas are investigated: \romannumeral1) the commonly used near-field channel models are reviewed along with their simplifications under various near-field conditions. \romannumeral2) Building upon these models, the information-theoretic capacity limits of NFC-MA are analyzed, including the derivation of sum-rate capacity and capacity region, and their upper limits for both downlink and uplink scenarios. \romannumeral3) A detailed investigation of near-field multiuser beamforming design is presented, offering low-complexity and effective NFC-MA design methodologies in both the spatial and wavenumber (angular) domains. Throughout these investigations, near-field MA is compared with its far-field counterpart to highlight its superiority and flexibility in terms of interference management, thereby laying the groundwork for achieving NGMA.

cs.IT↗

arXiv2023

Sensiverse: A dataset for ISAC study

In order to address the lack of applicable channel models for ISAC research and evaluation, we release Sensiverse, a dataset that can be used for ISAC research. In this paper, we present the method of generating Sensiverse, including the acquisition and formatting of the 3D scene models, the generation of the channel data and associations with Tx/Rx deployment. The file structure and usage of the dataset are also described, and finally the use of the dataset is illustrated with examples through the evaluation of use cases such as 3D environment reconstruction and moving targets.

eess.SP↗

arXiv2023

HAPS for 6G Networks: Potential Use Cases, Open Challenges, and Possible Solutions

High altitude platform station (HAPS), which is deployed in the stratosphere at an altitude of 20-50 kilometres, has attracted much attention in recent years due to their large footprint, line-of-sight links, and fixed position relative to the Earth. Compared with existing network infrastructure, HAPS has a much larger coverage area than terrestrial base stations and is much closer than satellites to the ground users. Besides small-cells and macro-cells, a HAPS can offer one mega-cell, which can complement legacy networks in 6G and beyond wireless systems. This paper explores potential use cases and discusses relevant open challenges of integrating HAPS into legacy networks, while also suggesting some solutions to these challenges. The cumulative density functions of spectral efficiency of the integrated network and cell-edge users are studied and compared with terrestrial network. The results show the capacity gains achieved by the integrated network are beneficial to cell-edge users. Furthermore, the advantages of a HAPS for backhauling aerial base stations are demonstrated by the simulation results.

cs.NI↗

arXiv2023

On the Road to 6G: Visions, Requirements, Key Technologies and Testbeds

Fifth generation (5G) mobile communication systems have entered the stage of commercial development, providing users with new services and improved user experiences as well as offering a host of novel opportunities to various industries. However, 5G still faces many challenges. To address these challenges, international industrial, academic, and standards organizations have commenced research on sixth generation (6G) wireless communication systems. A series of white papers and survey papers have been published, which aim to define 6G in terms of requirements, application scenarios, key technologies, etc. Although ITU-R has been working on the 6G vision and it is expected to reach a consensus on what 6G will be by mid-2023, the related global discussions are still wide open and the existing literature has identified numerous open issues. This paper first provides a comprehensive portrayal of the 6G vision, technical requirements, and application scenarios, covering the current common understanding of 6G. Then, a critical appraisal of the 6G network architecture and key technologies is presented. Furthermore, existing testbeds and advanced 6G verification platforms are detailed for the first time. In addition, future research directions and open challenges are identified for stimulating the on-going global debate. Finally, lessons learned to date concerning 6G networks are discussed.

eess.SP↗

arXiv2023

Sustaining Dynamic Traffic in Dense Urban Areas with High Altitude Platform Stations (HAPS)

The impact of information and communication technologies on global energy consumption is increasing every year, and mobile networks account for a significant portion of it. More than 50% of the total energy consumption of mobile networks is issued from radio access networks (RANs), due mainly to the rapid penetration of data-intensive applications and the increasing heterogeneity, dynamicity, and unpredictability of traffic. To tackle these high-demanding problems, RAN densification through the installation of additional base stations in high-demand areas is conventionally used. However, this leads to inefficient energy use and over-provisioning issues. In this context, high altitude platform stations (HAPS) may be used to complement RANs and sustain their services in densely populated areas, where traffic can peak unpredictably. Due to their wide coverage areas, substantial communication payloads, and green energy model, HAPS super macro base stations (SMBSs) are capable of handling the massive and dynamic mobile data traffic of ground users. In this paper, we show how HAPS-SMBSs can complement RANs and serve the dynamic and unpredictable traffic demands of users in an energy-efficient manner. Through the simulation of a case study, we demonstrate the performance of a HAPS-SMBS compared to the conventional RAN densification method and analyze the two approaches in terms of sustainability.

cs.NI↗