Search arXiv⌕ Search

arXiv subjects

Yi Hu

Publications and source records attributed to Yi Hu.

At least 19 recordsLinked to original sources
arXiv2026

SpBench: An Agent-Native Framework for Cross-Engine Lakehouse Benchmarking

SpBench is an agent-native framework for cross-engine lakehouse benchmarking, in production use for benchmarking Snowpark, Snowpark Connect for Apache Spark, and other execution environments across standardized suites -- such as TPCx-BB and TPC-DI -- and custom benchmarks distilled from production workload patterns. It runs continuously across a broad portfolio of workloads, guiding the performance roadmap and detecting regressions for production-scale data engines. Benchmarking has historically been an expensive investment requiring continuous cross-functional collaboration and domain expertise, making it difficult to obtain actionable performance insights. SpBench provides (i) a typed, code-generated benchmark substrate that compiles platform-agnostic benchmark definitions into fair, comparable platform-specific implementations; (ii) a structured-telemetry contract recording every run as queryable fact and dimension data, letting LLM agents query results deterministically; and (iii) an agent-native harness of three composable skills to expedite building benchmarks and analyzing results. This agent-native architecture enabled the rapid development of more than 400 benchmark queries that run on a regular cadence to detect performance regressions; catching these has directly led to improvements in production.

cs.DB↗
arXiv2026

Decoding Motion-Blurred QR Codes: Motion-Parameter Error Tolerance and Model-Based Recovery

Quick Response (QR) code deblurring must ultimately recover the encoded payload, yet how accurately the underlying camera motion must be estimated remains an open question. To address this question, we formulate motion-blurred QR recovery within a unified mathematical framework of identifiability and systematically measure the decoding tolerance to perturbations in motion length, direction, rotation or scaling centre, and speed profile. These controlled experiments reveal three structural laws: 1) parameter tolerances shrink rapidly as blur amplitude grows and are tightest under uniform speed, whereas accelerating motion is substantially more forgiving; 2) length tolerance can differ markedly between underestimation and overestimation; 3) errors that are individually benign interact strongly and prevent decoding when combined. Guided by these findings, we propose a grey-box recovery pipeline in which two compact networks predict an initial corner trajectory, followed by tolerance-guided candidate search, alternating image-and-path fitting, and regularised linear inversion. On a new benchmark of 955 handheld motion-blurred photographs captured across four phone models and four motion categories, our method recovers the exact payload on 93.1% of the images, outperforming the strongest fine-tuned deblurring network at 83.6% and reducing the decoding error rate by 58.0%.

eess.IV↗
arXiv2026

ContinualSkillBench: Can LLM Agents Truly Evolve Their Capabilities?

Modern agent frameworks equip large language models with external skill libraries to solve complex tasks. However, it remains unclear whether these systems can effectively evolve their skills and whether the resulting skills improve task-solving capabilities. To bridge this gap, we introduce ContinualSkillBench, a dynamic evaluation framework for in-context continual skill learning. It covers five domains, each with 100 tasks ordered by difficulty and opportunities for skill transfer. Our experiments show that sequential execution generally improves performance, but the gains vary substantially across models and domains. Moreover, across two models and three domains, retaining context and feedback without skill maintenance yields nearly the same mean performance as maintaining an explicit skill library. Explicit skills nevertheless provide selective benefits for tasks requiring reusable procedures or precise outputs. We further find that the weaker model (GPT-4o) accumulates more skills but reuses them less often, suggesting difficulty consolidating task-specific experience into reusable procedures. These findings show that current in-context skill evolution mechanisms can support continual adaptation, but still struggle to consistently consolidate experience into robust and transferable skills.

cs.AI↗
arXiv2026

Which Knowledge Should LLMs Trust? Uncovering a Generalizable Steering Direction Across Knowledge Conflicts

Large Language Models (LLMs) frequently encounter conflicts between internal parametric knowledge and external contextual knowledge, yet the mechanisms governing this selection remain unclear. We show that the choice between context and parametric knowledge is linearly represented and can be controlled along a single direction in the model's residual stream. We first demonstrate that this direction generalizes across factual-conflict settings, including context formats, knowledge categories, and datasets, while consistently shifting behavior between parametric and contextual answers. We then transfer a direction learned from simple factual conflicts to three more complex task families: RAG, memory, and rule following, where models must acquire information from retrieved documents, interaction histories, or explicit rules. Across four model families and multiple transfer settings, steering consistently improves the use of prompt-provided information over the unsteered model. These results suggest that the direction captures a transferable preference for external knowledge rather than a mechanism specific to factual conflicts.

cs.CL↗
arXiv2026

Knowledge Interaction in Hypernetwork-based Continual Learning: Conflict and Composition

Continual learning aims to adapt pretrained large language models (LLMs) to evolving data while preserving their pretrained capabilities. A central challenge is catastrophic forgetting: adaptation to new data can impair performance on previously learned knowledge. Since continual learning commonly changes model parameters, the model's predictions may draw on knowledge from the base model, the parameter update, and the input context. Understanding how these sources interact is essential: conflicts reveal whether new knowledge can override outdated knowledge, while multi-hop reasoning tests whether new knowledge can be composed with existing knowledge to support generalization. We study these interactions in a hypernetwork-based adaptation framework, where a shared network generates adapters from context, allowing a frozen base model to adapt without task-specific updates. We find that, under knowledge conflict, models preferentially follow the context, and their choices depend more on where knowledge is presented than on its content. We also show that context and adapter knowledge perturb the base model's representations along directions orthogonal to its clean hidden states, enabling us to derive steering vectors that control which source the model follows. In multi-hop reasoning, we find that models fail to reliably compose knowledge across the three sources, with the key breakdown arising from the interaction of the hypernetwork with both the base model and the context. Code and reproduction materials: https://github.com/missswiftie/Knowledge-Interaction-in-CL

cs.AI↗
arXiv2026

Diameter and radius of an additive-multiplicative graph

Let $G$ be the graph on the positive integers in which distinct vertices are adjacent if they differ by one or their quotient in one order is prime. We prove that $6\leq\operatorname{diam}(G)\leq7$ and $4\leq\operatorname{rad}(G)\leq5$. Under Dickson's conjecture for two linear forms, we prove that $\operatorname{diam}(G)=6$ and $\operatorname{rad}(G)=4$.

math.CO↗
arXiv2026

Shared Weights, Selected Computations: How Looped Transformers Route What Each Loop Does

Looped Transformers repeatedly apply the same set of Transformer layers, giving them a recurrent architecture for latent computation. Their strong performance on iterative reasoning and length-generalization tasks suggests an appealing explanation: recurrence may provide an inductive bias that lets the model reuse a learned algorithm across loops. However, weight sharing alone does not imply that every loop performs the same operation. This raises a basic question: is each loop actually repeating the same computation, and if not, what routes the shared parameters to different operations? We study this question using graph walks as a test case. In the model's native trajectories, decoded predictions can advance by different numbers of graph steps or remain at a reached target, showing that recurrent progress need not follow a fixed one-loop-one-step pattern. We then show that a frozen loop can be steered toward different transitions by modifying its entering hidden state: a learned linear layer $J$ selects the desired transition without changing the shared Transformer layers. To test how this steering works, we use activation patching and find that attention patterns can recover its effects and switch the selected transition. Across five matched pairs of graph models, changing intermediate supervision during backbone training changes which transitions $J$ can induce. This suggests that $J$ selects computations learned by the backbone rather than creating new algorithms. Together, these results show that the hidden state can control shared computation, with attention routing as a causal pathway.

cs.LG↗
arXiv2026

MolLangData: A Large-Scale Dataset for Molecular Structure-Language Description via a Rule-Regularized Method

Molecular function is largely determined by structure. Accurately aligning molecular structure with natural language is therefore essential for enabling large language models (LLMs) to both understand structure for chemical reasoning and generate molecules from natural-language design intent. However, the substantial cost of human annotation makes it infeasible to construct large-scale, high-quality datasets of structure-grounded descriptions. This work proposes a fully automated annotation framework for generating precise molecular descriptions at scale, such that the original molecule can be unambiguously reconstructed from the description alone. Our approach extends a rule-based chemical nomenclature parser to interpret IUPAC names and construct enriched, XML metadata that explicitly encodes molecular structure. This is then used to guide LLMs in producing accurate natural-language descriptions. Using this framework, we curate MolLangData, a dataset of approximately $163$k molecule--description pairs. A rigorous validation protocol combining expert human and LLM-based reconstruction on a subset of $2,000$ molecules demonstrates $98.6\%$ description precision. Using the curated dataset, we train a $4$B-parameter LLM via large-scale reinforcement learning for language-conditional molecule generation. The proposed framework and dataset provide a reliable foundation for molecule--language alignment, readily beneficial to broader chemical tasks.

cs.CL↗
arXiv2026

ProofVerifier: A Scalable, Diversity-Driven Framework for Natural-Language Proof Verification

While large language models (LLMs) have achieved strong performance on mathematical problems with verifiable answers, many advanced problems are proof-based and require evaluating full proofs. However, training such verifiers requires diverse and trustworthy question-proof-check (QPC) examples at scale, which are scarce. To address this challenge, we develop a human-audited, LLM-assisted data pipeline that produces large-scale QPC triplets with limited human effort. By systematically varying problem sources, generation strategies, and generator models, the pipeline creates diverse problem-proof pairs spanning multiple difficulty levels, linguistic styles, and error types. We combine multi-LLM agreement with hierarchical human auditing to obtain accurate proof-correctness labels. Using these data, we train generative proof verifiers and introduce an auxiliary fluency filter together with balanced token weighting to stabilize binary-reward long-form verification RL. Experiments show that our verifier improves proof-judgment accuracy across different proof styles and provides useful guidance for test-time selection. Overall, our results provide a practical data and training framework for natural-language proof verification.

cs.CL↗
arXiv2026

A Closed-Loop Thermal Dynamic Model for AI Data Center Cooling Load Simulation

Cooling demand constitutes a significant and flexible component of AI data center electricity consumption, but time-synchronized measurements are scarce and constant coefficient-of-performance models cannot represent thermal dynamics. This letter proposes a closed-loop simulation model which couples a linear thermal dynamic model with deadband-based control to capture the nonlinear cooling dynamics. The model is validated using operational telemetry from the Marconi100 supercomputer. Compared with the baseline, the proposed model reduces the mean absolute error from 95.80 to 20.88~kW and the root-mean-square error from 109.79 to 27.27~kW. Evaluation over approximately 520 daily profiles further shows improved reproduction of daily peak demand and intraday variability. The proposed model provides a computationally tractable means of generating physically interpretable cooling load profiles for power system studies.

eess.SY↗
arXiv2026

DREAM Technical Report

Industrial recommender systems commonly use cascaded retrieval, ranking, and re-ranking pipelines. Although efficient, these pipelines fragment information and objectives across modules, rely on rigid rules, and have limited awareness of real-time intent, leaving session-level shifts among browsing, comparison, and purchase insufficiently addressed. We present DREAM (Developing Recommender Engine with Agentic Methods), an autonomous optimization control architecture that adds a perception-aware, orchestrable, and auditable policy layer atop existing pipelines without replacing them. DREAM has two core components. First, a three-tier Intent Engine fuses on-device signals into structured L0/L1/L2 intent representations; its edge-cloud trigger chain reduces reporting volume to approximately 8.7%. Second, a Meta Engine uses a MetaModel for layered M1-to-M2-to-M3 reasoning: intent summarization, strategy planning informed by Strategy Memory, and parameter translation. It dispatches the resulting parameters through a unified outlet with safety guardrails. A Reward Dual Loop continuously optimizes both components by combining offline simulation for strategy-space exploration with online feedback for outcome calibration, forming a cycle of generation, execution, evaluation, and experience accumulation. Large-scale A/B tests on Taobao's homepage feed show that re-ranking control alone improves IPV by 2.06%, Core IPV by 2.39%, and GMV by 0.88%. Extending control to fine ranking raises these gains to 2.71%, 3.06%, and 1.31%, respectively, while consistently improving PV by more than 1%. These gains require neither replacement of pipeline models nor compromise of serving stability, supporting agentic meta-control as a viable paradigm for industrial recommendation.

cs.IR↗
arXiv2026

HVR-Met: A Hypothesis-Verification-Replanning Agentic System for Extreme Weather Diagnosis

While deep learning-based weather forecasting paradigms have made significant strides, addressing extreme weather diagnostics remains a formidable challenge. This gap exists primarily because the diagnostic process demands sophisticated multi-step logical reasoning, dynamic tool invocation, and expert-level prior judgment. Although agents possess inherent advantages in task decomposition and autonomous execution, current architectures are still hampered by critical bottlenecks: inadequate expert knowledge integration, a lack of professional-grade iterative reasoning loops, and the absence of fine-grained validation and evaluation systems for complex workflows under extreme conditions. To this end, we propose HVR-Met, a multi-agent meteorological diagnostic system characterized by the deep integration of expert knowledge. Its central innovation is the ``Hypothesis-Verification-Replanning'' closed-loop mechanism, which facilitates sophisticated iterative reasoning for anomalous meteorological signals during extreme weather events. To bridge gaps within existing evaluation frameworks, we further introduce a novel benchmark focused on atomic-level subtasks. Experimental evidence demonstrates that the system excels in complex diagnostic scenarios.

cs.AI↗
arXiv2026

$J$ and $H$ band sky brightness measurements from polar day to polar night at Dome A, Antarctica

The near-infrared (NIR) sky brightness is a fundamental parameter for evaluating the performance of ground-based infrared observatories. Dome~A on the Antarctic plateau offers exceptional atmospheric conditions, yet its NIR sky background has not been continuously monitored. We present the first continuous $J/H$-band measurements of the sky background at Dome~A from polar day to polar night, and characterize their median levels and temporal variability. The Antarctic Infrared Binocular Telescope (AIRBT), operating in the $J$ and $H$ bands, obtained continuous fixed-pointing observations from February to May 2024, which were used to measure the NIR sky background. The median sky brightness is $5.2/2.9$ and $15.3/13.4~\mathrm{mag~arcsec^{-2}}$ in $J/H$ bands during daytime and nighttime, respectively. The twilight--nighttime boundaries occur at solar elevations of $-9.3^\circ$ in $J$ and $-7.4^\circ$ in $H$. At the same solar elevation, the NIR sky background during the polar night is darker by about $0.1$ and $0.4~\mathrm{mag~arcsec^{-2}}$ in the $J$ and $H$ bands compared with the period of regular day--night alternation. During the polar-night period, the nighttime sky brightness in the $H$ band shows a more evident association with the sunspot number, while the corresponding trend in the $J$ band is weaker. These results reveal systematic differences in sky background between polar and non-polar environments and between polar night and regular day--night cycles. The measured sky brightness may be elevated, as the observations were conducted near solar maximum, highlighting the importance of long-term monitoring across the solar cycle.

astro-ph.IM↗
arXiv2026

Generalized moment maps, reduction and complex quotients

In this note, we introduce the concept of momentumly closed forms. A non-degenerate momentumly closed two-form and its generalized moment map are the generalization of two well-known notions, symplectic forms and moment maps, in the almost Hermitian setting. We then generalize the classical theory of moment maps to this broader framework. As a first step, we prove a variant of the Darboux-Weinstein theorem for non-degenerate momentumly closed two-forms. Based on this, we further establish the convexity property of the generalized moment map, construct the corresponding reduction space and investigate the properties of the Kirwan-Ness stratification.

math.DG↗
arXiv2026

Time-independent theoretical framework for stroboscopic nonlinear dynamics based on time-nonlocal response

Recent experiments have demonstrated the ability to manipulate nonlinear interactions via time modulation, giving rise to the so-called stroboscopic nonlinearity. To date, however, this phenomenon has not been subjected to a rigorous theoretical analysis. In this work, we clarify the physical mechanism underlying stroboscopic nonlinear dynamics based on time-nonlocal response and establish an effective time-independent model under suitable modulation conditions. The proposed model almost exactly reproduces the full time-dependent dynamics in the quasi-steady state and significantly outperforms empirical descriptions used previously. Our results provide a clear physical picture of stroboscopic nonlinear dynamics, and can be extended to other systems with time-nonlocal response, establishing a general framework for engineering nonlinear interactions through temporal modulation.

physics.optics↗
arXiv2026

On the Controllability-Fidelity Frontier in Diffusion Editing

Diffusion-based generative models enable powerful image editing capabilities, but achieving precise control while maintaining fidelity and safety remains challenging. We present a comprehensive theoretical and empirical study of controllable diffusion-based image editing, analyzing the trade-offs between adherence to user intent, preservation of non-target content, and output quality. Our work spans text- and mask-guided edits, point/drag manipulation, and inversion-based pipelines. We derive mathematical formulations of editing objectives and analyze dynamics of noise injection, score guidance, and inversion error. We provide theoretical bounds on reconstruction error, stability under repeated edits, and locality of changes. We propose algorithmic frameworks (with pseudocode) for mask-localized and instruction-guided editing, and present extensive experiments comparing state-of-the-art methods (e.g.\ TF-ICON \cite{lu2023tficone}, DragFlow \cite{zhou2025dragflow}, InstructPix2Pix \cite{brooks2023instructpix2pix}, UltraEdit \cite{zhao2024ultraedit}) on multiple tasks and metrics (FID, identity similarity, CLIP alignment, artifact scores, etc). Our results reveal key failure modes, such as identity drift, prompt sensitivity, and compositional errors. We also discuss ethical considerations in image editing, including misuse risks, bias, consent, and concept erasure techniques (e.g.\ MACE \cite{lu2024mace}, ANT \cite{li2025ant}, EraseAnything \cite{gao2024eraseanything}) as safeguards. We conclude with best practices and future directions for responsible, high-fidelity diffusion-based editing.

cs.GR↗