Searcharxiv⌕ Search

arXiv subjects

Qi He

Publications and source records attributed to Qi He.

At least 37 records · Page 2Linked to original sources
arXiv2026

Can Large Language Models Adequately Perform Symbolic Reasoning Over Time Series?

Uncovering hidden symbolic laws from time series data, as an aspiration dating back to Kepler's discovery of planetary motion, remains a core challenge in scientific discovery and artificial intelligence. While Large Language Models show promise in structured reasoning tasks, their ability to infer interpretable, context-aligned symbolic structures from time series data is still underexplored. To systematically evaluate this capability, we introduce SymbolBench, a comprehensive benchmark designed to assess symbolic reasoning over real-world time series across three tasks: multivariate symbolic regression, Boolean network inference, and causal discovery. Unlike prior efforts limited to simple algebraic equations, SymbolBench spans a diverse set of symbolic forms with varying complexity. We further propose a unified framework that integrates LLMs with genetic programming to form a closed-loop symbolic reasoning system, where LLMs act both as predictors and evaluators. Our empirical results reveal key strengths and limitations of current models, highlighting the importance of combining domain knowledge, context alignment, and reasoning structure to improve LLMs in automated scientific discovery. https://github.com/nuuuh/SymbolBench.

cs.AI↗
arXiv2026

STReasoner: Empowering LLMs for Spatio-Temporal Reasoning in Time Series via Spatial-Aware Reinforcement Learning

Spatio-temporal reasoning in time series involves the explicit synthesis of temporal dynamics, spatial dependencies, and textual context. This capability is vital for high-stakes decision-making in systems such as traffic networks, power grids, and disease propagation. However, the field remains underdeveloped because most existing works prioritize predictive accuracy over reasoning. To address the gap, we introduce ST-Bench, a benchmark consisting of four core tasks, including etiological reasoning, entity identification, correlation reasoning, and in-context forecasting, developed via a network SDE-based multi-agent data synthesis pipeline. We then propose STReasoner, which empowers LLM to integrate time series, graph structure, and text for explicit reasoning. To promote spatially grounded logic, we introduce S-GRPO, a reinforcement learning algorithm that rewards performance gains specifically attributable to spatial information. Experiments show that STReasoner achieves average accuracy gains between 17% and 135% at only 0.004X the cost of proprietary models and generalizes robustly to real-world data.

cs.CL↗
arXiv2026

Adaptive Multi-Scale Channel-Spatial Attention Aggregation Framework for 3D Indoor Semantic Scene Completion Toward Assisting Visually Impaired

Independent indoor mobility remains a critical challenge for individuals with visual impairments, largely due to the limited capability of existing assistive systems in detecting fine-grained hazardous objects such as chairs, tables, and small obstacles. These perceptual blind zones substantially increase the risk of collision in unfamiliar environments. To bridge the gap between monocular 3D vision research and practical assistive deployment, this paper proposes an Adaptive Multi-scale Attention Aggregation (AMAA) framework for monocular 3D semantic scene completion using only a wearable RGB camera. The proposed framework addresses two major limitations in 2D-to-3D feature lifting: noise diffusion during back-projection and structural instability in multi-scale fusion. A parallel channel--spatial attention mechanism is introduced to recalibrate lifted features along semantic and geometric dimensions, while a hierarchical adaptive gating strategy regulates cross-scale information flow to preserve fine-grained structural details. Experiments on the NYUv2 benchmark demonstrate that AMAA achieves an overall mIoU of 27.88%. Crucially, it yields significant relative improvements of 16.9% for small objects and 10.4% for tables over the MonoScene baseline. Furthermore, a wearable prototype based on an NVIDIA Jetson Orin NX and a ZED~2i camera validates stable real-time performance in indoor environments, demonstrating the feasibility of deploying monocular 3D scene completion for assistive navigation.

cs.CV↗
arXiv2026

To trust or not to trust: Attention-based Trust Management for LLM Multi-Agent Systems

Large Language Model-based Multi-Agent Systems (LLM-MAS) have demonstrated strong capabilities in solving complex tasks but remain vulnerable when agents receive unreliable messages. This vulnerability stems from a fundamental gap: LLM agents treat all incoming messages equally without evaluating their trustworthiness. While some existing studies approach trustworthiness, they focus on a single type of harmfulness rather than analyze it in a holistic approach from multiple trustworthiness perspectives. We address this gap by proposing a comprehensive definition of trustworthiness inspired by human communication theory (Grice, 1975). Our definition identifies six orthogonal trust dimensions that provide interpretable measures of trustworthiness. Building on this definition, we introduce the Attention Trust Score (A -Trust), a lightweight, attention-based method for evaluating the trustworthiness of messages. We then develop a principled trust management system (TMS) for LLM -MAS that supports both message-level and agent-level trust assessments. Experiments across diverse multi-agent settings and tasks demonstrate that our TMS significantly improves robustness against malicious inputs.

cs.CR↗
arXiv2026

Adaptive Serverless Resource Management via Slot-Survival Prediction and Event-Driven Lifecycle Control

Serverless computing eliminates infrastructure management overhead but introduces significant challenges regarding cold start latency and resource utilization. Traditional static resource allocation often leads to inefficiencies under variable workloads, resulting in performance degradation or excessive costs. This paper presents an adaptive engineering framework that optimizes serverless performance through event-driven architecture and probabilistic modeling. We propose a dual-strategy mechanism that dynamically adjusts idle durations and employs an intelligent request waiting strategy based on slot survival predictions. By leveraging sliding window aggregation and asynchronous processing, our system proactively manages resource lifecycles. Experimental results show that our approach reduces cold starts by up to 51.2% and improves cost-efficiency by nearly 2x compared to baseline methods in multi-cloud environments.

cs.AI↗
arXiv2026

Secure Forgetting: A Framework for Privacy-Driven Unlearning in Large Language Model (LLM)-Based Agents

Large language model (LLM)-based agents have recently gained considerable attention due to the powerful reasoning capabilities of LLMs. Existing research predominantly focuses on enhancing the task performance of these agents in diverse scenarios. However, as LLM-based agents become increasingly integrated into real-world applications, significant concerns emerge regarding their accumulation of sensitive or outdated knowledge. Addressing these concerns requires the development of mechanisms that allow agents to selectively forget previously learned knowledge, giving rise to a new term LLM-based agent unlearning. This paper initiates research on unlearning in LLM-based agents. Specifically, we propose a novel and comprehensive framework that categorizes unlearning scenarios into three contexts: state unlearning (forgetting specific states or items), trajectory unlearning (forgetting sequences of actions) and environment unlearning (forgetting entire environments or categories of tasks). Within this framework, we introduce a natural language-based unlearning method that trains a conversion model to transform high-level unlearning requests into actionable unlearning prompts, guiding agents through a controlled forgetting process. Moreover, to evaluate the robustness of the proposed framework, we introduce an unlearning inference adversary capable of crafting prompts, querying agents, and observing their behaviors in an attempt to infer the forgotten knowledge. Experimental results show that our approach effectively enables agents to forget targeted knowledge while preserving performance on untargeted tasks, and prevents the adversary from inferring the forgotten knowledge.

cs.MA↗
arXiv2026

Multi-Crit: Benchmarking Multimodal Judges on Pluralistic Criteria-Following

Large multimodal models (LMMs) are increasingly adopted as judges in multimodal evaluation systems due to their strong instruction following and consistency with human preferences. However, their ability to follow diverse, fine-grained evaluation criteria remains underexplored. We develop Multi-Crit, a benchmark for evaluating multimodal judges on their capacity to follow pluralistic criteria and produce reliable criterion-level judgments. Covering both open-ended generation and verifiable reasoning tasks, Multi-Crit is built through a rigorous data curation pipeline that gathers challenging response pairs with multi-criterion human annotations. It further introduces three novel metrics for systematically assessing pluralistic adherence, criterion-switching flexibility, and the ability to recognize criterion-level preference conflicts. Comprehensive analysis of 25 LMMs reveals that 1) proprietary models still struggle to maintain consistent adherence to pluralistic criteria--especially in open-ended evaluation; 2) open-source models lag further behind in flexibly following diverse criteria; and 3) critic fine-tuning with holistic judgment signals enhances visual grounding but fails to generalize to pluralistic criterion-level judgment. Additional analyses on reasoning fine-tuning, test-time scaling, and boundary consistency between open-source and proprietary models further probe the limits of current multimodal judges. As a pioneering study, Multi-Crit lays the foundation for building reliable and steerable multimodal AI evaluation.

cs.CV↗
arXiv2026

ViTex: Visual Texture Control for Multi-Track Symbolic Music Generation via Discrete Diffusion Models

In automatic music generation, a central challenge is to design controls that enable meaningful human-machine interaction. Existing systems often rely on extrinsic inputs such as text prompts or metadata, which do not allow humans to directly shape the composition. While prior work has explored intrinsic controls such as chords or hierarchical structure, these approaches mainly address piano or vocal-accompaniment settings, leaving multitrack symbolic music largely underexplored. We identify instrumentation, the choice of instruments and their roles, as a natural dimension of control in multi-track composition, and propose ViTex, a visual representation of instrumental texture. In ViTex, color encodes instrument choice, spatial position represents pitch and time, and stroke properties capture local textures. Building on this representation, we develop a discrete diffusion model conditioned on ViTex and chord progressions to generate 8-measure multi-track symbolic music, enabling explicit texture-level control while maintaining strong unconditional generation quality. The demo page and code are avaliable at https://vitex2025.github.io/.

cs.SD↗
arXiv2026

Harnessing the Unseen: The Hidden Influence of Intrinsic Knowledge in Long-Context Language Models

Recent advances in long-context language models (LCLMs), designed to handle extremely long contexts, primarily focus on utilizing external contextual information, often leaving the influence of language models' parametric knowledge underexplored. In this work, we firstly investigate how this parametric knowledge affects content generation and demonstrate that its impact becomes increasingly pronounced as context length extends. Furthermore, we show that the model's ability to utilize parametric knowledge, which we call parametric recall ability, does not improve simultaneously with its ability to leverage contextual knowledge through extrinsic retrieval ability. Moreover, better extrinsic retrieval ability can interfere with the model's parametric recall ability, limiting its full potential. To bridge this gap, we design a simple yet effective Hybrid Needle-in-a-Haystack test that evaluates models based on their capabilities across both abilities, rather than solely emphasizing extrinsic retrieval ability. Our experimental results reveal that Qwen-2.5 models significantly outperform Llama-3.1 models, demonstrating superior potential to combine various abilities. Moreover, even the more powerful Llama-3.1-70B-Instruct model fails to exhibit better performance, highlighting the importance of evaluating models from a dual-ability perspective.

cs.CL↗
arXiv2026

How Far Are LLMs from Professional Poker Players? Revisiting Game-Theoretic Reasoning with Agentic Tool Use

As Large Language Models (LLMs) are increasingly applied in high-stakes domains, their ability to reason strategically under uncertainty becomes critical. Poker provides a rigorous testbed, requiring not only strong actions but also principled, game-theoretic reasoning. In this paper, we conduct a systematic study of LLMs in multiple realistic poker tasks, evaluating both gameplay outcomes and reasoning traces. Our analysis reveals LLMs fail to compete against traditional algorithms and identifies three recurring flaws: reliance on heuristics, factual misunderstandings, and a "knowing-doing" gap where actions diverge from reasoning. An initial attempt with behavior cloning and step-level reinforcement learning improves reasoning style but remains insufficient for accurate game-theoretic play. Motivated by these limitations, we propose ToolPoker, a tool-integrated reasoning framework that combines external solvers for GTO-consistent actions with more precise professional-style explanations. Experiments demonstrate that ToolPoker achieves state-of-the-art gameplay while producing reasoning traces that closely reflect game-theoretic principles.

cs.AI↗
arXiv2026

Location-Robust Cost-Preserving Blended Pricing for Multi-Campus AI Data Centers

Large-scale AI data center portfolios procure identical SKUs across geographically heterogeneous campuses, yet finance and operations require a single system-level 'world price' per SKU for budgeting and planning. A common practice is deployment-weighted blending of campus prices, which preserves total cost but can trigger Simpson-type aggregation failures: heterogeneous location mixes can reverse SKU rankings and distort decision signals. I formalize cost-preserving blended pricing under location heterogeneity and propose two practical operators that reconcile accounting identity with ranking robustness and production implementability. A two-way fixed-effects operator separates global SKU effects from campus effects and restores exact cost preservation via scalar normalization, providing interpretable decomposition and smoothing under mild missingness. A convex common-weight operator computes a single set of campus weights under accounting constraints to enforce a location-robust benchmark and prevent dominance reversals; I also provide feasibility diagnostics and a slack-based fallback for extreme mix conditions. Simulations and an AI data center OPEX illustration show substantial reductions in ranking violations relative to naive blending while maintaining cost accuracy, with scalable distributed implementation.

econ.GN↗
arXiv2026

A Unified Metric Architecture for AI Infrastructure: A Cross-Layer Taxonomy Integrating Performance, Efficiency, and Cost

The growth of large-scale AI systems is increasingly constrained by infrastructure limits: power availability, thermal and water constraints, interconnect scaling, memory pressure, data-pipeline throughput, and rapidly escalating lifecycle cost. Across hyperscale clusters, these constraints interact, yet the main metrics remain fragmented. Existing metrics, ranging from facility measures (PUE) and rack power density to network metrics (all-reduce latency), data-pipeline measures, and financial metrics (TCO series), each capture only their own domain and provide no integrated view of how physical, computational, and economic constraints interact. This fragmentation obscures the structural relationships among energy, computation, and cost, preventing a coherent optimization across sector and how bottlenecks emerge, propagate, and jointly determine the efficiency frontier of AI infrastructure. This paper develops an integrated framework that unifies these disparate metrics through a three-domain semantic classification and a six-layer architectural decomposition, producing a 6x3 taxonomy that maps how various sectors propagate across the AI infrastructure stack. The taxonomy is grounded in a systematic review and meta-analysis of all metrics with economic and financial relevance, identifying the most widely used measures, their research intensity, and their cross-domain interdependencies. Building on this evidence base, the Metric Propagation Graph (MPG) formalizes cross-layer dependencies, enabling systemwide interpretation, composite-metric construction, and multi-objective optimization of energy, carbon, and cost. The framework offers a coherent foundation for benchmarking, cluster design, capacity planning, and lifecycle economic analysis by linking physical operations, computational efficiency, and cost outcomes within a unified analytic structure.

econ.GN↗
arXiv2026

Higher Satisfaction, Lower Cost: A Technical Report on How LLMs Revolutionize Meituan's Intelligent Interaction Systems

Enhancing customer experience is essential for business success, particularly as service demands grow in scale and complexity. Generative artificial intelligence and Large Language Models (LLMs) have empowered intelligent interaction systems to deliver efficient, personalized, and 24/7 support. In practice, intelligent interaction systems encounter several challenges: (1) Constructing high-quality data for cold-start training is difficult, hindering self-evolution and raising labor costs. (2) Multi-turn dialogue performance remains suboptimal due to inadequate intent understanding, rule compliance, and solution extraction. (3) Frequent evolution of business rules affects system operability and transferability, constraining low-cost expansion and adaptability. (4) Reliance on a single LLM is insufficient in complex scenarios, where the absence of multi-agent frameworks and effective collaboration undermines process completeness and service quality. (5) The open-domain nature of multi-turn dialogues, lacking unified golden answers, hampers quantitative evaluation and continuous optimization. To address these challenges, we introduce WOWService, an intelligent interaction system tailored for industrial applications. With the integration of LLMs and multi-agent architectures, WOWService enables autonomous task management and collaborative problem-solving. Specifically, WOWService focuses on core modules including data construction, general capability enhancement, business scenario adaptation, multi-agent coordination, and automated evaluation. Currently, WOWService is deployed on the Meituan App, achieving significant gains in key metrics, e.g., User Satisfaction Metric 1 (USM 1) -27.53% and User Satisfaction Metric 2 (USM 2) +25.51%, demonstrating its effectiveness in capturing user needs and advancing personalized service.

cs.CL↗
arXiv2025

Modular Landfill Remediation for AI Grid Resilience

Rising AI electricity demand and persistent landfill methane emissions constitute coupled constraints on U.S. digital infrastructure and decarbonization. While China has achieved a rapid 'de-landfilling' transition through centralized coordination, the U.S. remains structurally 'locked in' to landfilling due to fragmented governance and carbon accounting incentives. This paper proposes a modular legacy landfill remediation framework to address these dual challenges within U.S. institutional constraints. By treating legacy sites as stock resources, the proposed system integrates excavation, screening, and behind-the-meter combined heat and power (CHP) to transform environmental liabilities into resilience assets. A system analysis of a representative AI corridor demonstrates that such modules can mitigate site-level methane by 60-70% and recover urban land, while supplying approximately 20 MW of firm, islandable power. Although contributing only approximately 5% of a hyperscale data center's bulk load, it provides critical microgrid resilience and black-start capability. We conclude that remediation-oriented waste-to-energy should be valued not as a substitute for bulk renewables, but as a strategic control volume for buffering critical loads against grid volatility while resolving long-term environmental liabilities.

eess.SY↗
arXiv2025

Bilevel ZOFO: Efficient LLM Fine-Tuning and Meta-Training

Fine-tuning pre-trained Large Language Models (LLMs) for downstream tasks using First-Order (FO) optimizers presents significant computational challenges. Parameter-Efficient Fine-Tuning (PEFT) methods address these by freezing most model parameters and training only a small subset. However, PEFT often underperforms compared to full fine-tuning when high task-specific accuracy is required. Zeroth-Order (ZO) methods fine-tune the entire pre-trained model without back-propagation, estimating gradients through forward passes only. While memory-efficient, ZO methods suffer from slow convergence and high sensitivity to prompt selection. We bridge these two worlds with Bilevel-ZOFO, a bilevel optimization method that couples fast, local FO-PEFT adaptation at the inner level with stable, memory-efficient ZO updates of the full backbone at the outer level. The FO-PEFT inner loop performs fast, low-memory local adaptation that reduces the variance of ZO estimates and stabilizes the search, guiding the outer ZO updates of the full backbone and reducing prompt sensitivity. In the mean time, the outer ZO provides better generalization ability for PEFT. We provide theoretical convergence guarantees and empirically demonstrate that Bilevel-ZOFO significantly outperforms existing ZO and FO-PEFT methods, achieving 2-4 times faster training while maintaining similar memory efficiency. Additionally, we show by updating the backbone with ZO and adapting only a tiny FO-PEFT block per task, Bilevel-ZOFO combines full-model capacity with few-shot efficiency, making it a very efficient meta-learning algorithm that quickly adapts to new tasks.

cs.LG↗
arXiv2025

Bounds of Block Rewards in Honest PinFi Systems

PinFi is a class of novel protocols for decentralized pricing of dissipative assets, whose value naturally declines over time. Central to the protocol's functionality and its market efficiency is the role of liquidity providers (LPs). This study addresses critical stability and sustainability challenges within the protocol, namely: the propensity of LPs to prefer selling in external markets over participation in the protocol; a similar inclination towards selling within the PinFi system rather than contributing as LPs; and a scenario where LPs are disinclined to sell within the protocol. Employing a game-theoretic approach, we explore PinFi's mechanisms and its broader ramifications. Our findings reveal that, under a variety of common conditions and with an assumption of participant integrity, PinFi is capable of fostering a dynamic equilibrium among LPs, sellers, and buyers. This balance is maintained through a carefully calibrated range of block rewards for LPs, ensuring the protocol's long-term stability and utility.

cs.GT↗
arXiv2025

LooPIN: A PinFi protocol for decentralized computing

Networked computing power is a critical utility in the era of artificial intelligence. This paper presents a novel Physical Infrastructure Finance (PinFi) protocol designed to facilitate the distribution of computing power within networks in a decentralized manner. Addressing the core challenges of coordination, pricing, and liquidity in decentralized physical infrastructure networks (DePIN), the PinFi protocol introduces a distinctive dynamic pricing mechanism. It enables providers to allocate excess computing resources to a "dissipative" PinFi liquidity pool, distinct from traditional DeFi liquidity pools, ensuring seamless access for clients at equitable, market-based prices. This approach significantly reduces the costs of accessing computing power, potentially to as low as 1% compared to existing services, while simultaneously enhancing security and dependability. The PinFi protocol is poised to transform the dynamics of supply and demand in computing power networks, setting a new standard for efficiency and accessibility.

cs.DC↗
arXiv2025

Unlocking the Power of Multi-Agent LLM for Reasoning: From Lazy Agents to Deliberation

Large Language Models (LLMs) trained with reinforcement learning and verifiable rewards have achieved strong results on complex reasoning tasks. Recent work extends this paradigm to a multi-agent setting, where a meta-thinking agent proposes plans and monitors progress while a reasoning agent executes subtasks through sequential conversational turns. Despite promising performance, we identify a critical limitation: lazy agent behavior, in which one agent dominates while the other contributes little, undermining collaboration and collapsing the setup to an ineffective single agent. In this paper, we first provide a theoretical analysis showing why lazy behavior naturally arises in multi-agent reasoning. We then introduce a stable and efficient method for measuring causal influence, helping mitigate this issue. Finally, as collaboration intensifies, the reasoning agent risks getting lost in multi-turn interactions and trapped by previous noisy responses. To counter this, we propose a verifiable reward mechanism that encourages deliberation by allowing the reasoning agent to discard noisy outputs, consolidate instructions, and restart its reasoning process when necessary. Extensive experiments demonstrate that our framework alleviates lazy agent behavior and unlocks the full potential of multi-agent framework for complex reasoning tasks.

cs.AI↗