arXiv subjects

Yong Li

Publications and source records attributed to Yong Li.

At least 163 records · Page 9Linked to original sources

arXiv2026

Efficient Reasoning with Balanced Thinking

Large Reasoning Models (LRMs) have shown remarkable reasoning capabilities, yet they often suffer from overthinking, expending redundant computational steps on simple problems, or underthinking, failing to explore sufficient reasoning paths despite inherent capabilities. These issues lead to inefficiencies and potential inaccuracies, limiting practical deployment in resource-constrained settings. Existing methods to mitigate overthinking, such as suppressing reflective keywords or adjusting reasoning length, may inadvertently induce underthinking, compromising accuracy. Therefore, we propose ReBalance, a training-free framework that achieves efficient reasoning with balanced thinking. ReBalance leverages confidence as a continuous indicator of reasoning dynamics, identifying overthinking through high confidence variance and underthinking via consistent overconfidence. By aggregating hidden states from a small-scale dataset into reasoning mode prototypes, we compute a steering vector to guide LRMs' reasoning trajectories. A dynamic control function modulates this vector's strength and direction based on real-time confidence, pruning redundancy during overthinking, and promoting exploration during underthinking. Extensive experiments conducted on four models ranging from 0.5B to 32B, and across nine benchmarks in math reasoning, general question answering, and coding tasks demonstrate that ReBalance effectively reduces output redundancy while improving accuracy, offering a general, training-free, and plug-and-play strategy for efficient and robust LRM deployment. Project page and code are available at https://rebalance-ai.github.io .

cs.AI↗

arXiv2026

AgentExpt: Automating AI Experiment Design with LLM-based Resource Retrieval Agent

Large language model agents are becoming increasingly capable at web-centric tasks such as information retrieval, complex reasoning. These emerging capabilities have given rise to surge research interests in developing LLM agent for facilitating scientific quest. One key application in AI research is to automate experiment design through agentic dataset and baseline retrieval. However, prior efforts suffer from limited data coverage, as recommendation datasets primarily harvest candidates from public portals and omit many datasets actually used in published papers, and from an overreliance on content similarity that biases model toward superficial similarity and overlooks experimental suitability. Harnessing collective perception embedded in the baseline and dataset citation network, we present a comprehensive framework for baseline and dataset recommendation. First, we design an automated data-collection pipeline that links roughly one hundred thousand accepted papers to the baselines and datasets they actually used. Second, we propose a collective perception enhanced retriever. To represent the position of each dataset or baseline within the scholarly network, it concatenates self-descriptions with aggregated citation contexts. To achieve efficient candidate recall, we finetune an embedding model on these representations. Finally, we develop a reasoning-augmented reranker that exact interaction chains to construct explicit reasoning chains and finetunes a large language model to produce interpretable justifications and refined rankings. The dataset we curated covers 85\% of the datasets and baselines used at top AI conferences over the past five years. On our dataset, the proposed method outperforms the strongest prior baseline with average gains of +5.85\% in Recall@20, +8.30\% in HitRate@5. Taken together, our results advance reliable, interpretable automation of experimental design.

cs.CL↗

arXiv2026

NES: An Instruction-Free, Low-Latency Next Edit Suggestion Framework Powered by Learned Historical Editing Trajectories

Code editing is a frequent yet cognitively demanding task in software development. Existing AI-powered tools often disrupt developer flow by requiring explicit natural language instructions and suffer from high latency, limiting real-world usability. We present NES (Next Edit Suggestion), an instruction-free, low-latency code editing framework that leverages learned historical editing trajectories to implicitly capture developers' goals and coding habits. NES features a dual-model architecture: one model predicts the next edit location and the other generates the precise code change, both without any user instruction. Trained on our open-sourced SFT and DAPO datasets, NES achieves state-of-the-art performance (75.6% location accuracy, 27.7% exact match rate) while delivering suggestions in under 250ms. Deployed at Ant Group, NES serves over 20,000 developers through a seamless Tab-key interaction, achieving effective acceptance rates of 51.55% for location predictions and 43.44% for edits, demonstrating its practical impact in real-world development workflows.

cs.SE↗

arXiv2026

An infinite-dimensional Kolmogorov theorem and the construction of almost periodic breathers

In this paper, we present two infinite-dimensional Kolmogorov theorems based on non-resonant frequencies of Bourgain's Diophantine type or even weaker conditions. To be more precise, under a Legendre-type nondegeneracy condition for an infinite-dimensional Hamiltonian system, we prove the persistence of a full-dimensional KAM torus with a universally prescribed frequency independent of any spectral asymptotics. As an application, we prove that for a class of perturbed networks with weakly coupled oscillators described by \[\frac{{{{\rm d}^2}{x_n}}}{{{\rm d}{t^2}}} + V'\left( {x_n} \right) = \varepsilon_n {W'\left( {x_{n + 1} - {x_n}} \right) - \varepsilon_{n-1}W'\left( {{x_n} - {x_{n - 1}}} \right)} ,\quad n \in \mathbb{Z},\] or even for more general perturbed networks, frequency-preserving almost periodic breathers do persist, provided that the local potential $ V $ and the coupling potential $ W $ satisfy certain assumptions. In particular, this yields the first frequency-preserving result for the Aubry--MacKay conjecture [MA94,Aub95].

math.DS↗

arXiv2026

VecAttention: Vector-wise Sparse Attention for Accelerating Long Context Inference

Long-context video understanding and generation pose a significant computational challenge for Transformer-based video models due to the quadratic complexity of self-attention. While existing sparse attention methods employ coarse-grained patterns to improve efficiency, they typically incur redundant computation and suboptimal performance. To address this issue, in this paper, we propose \textbf{VecAttention}, a novel framework of vector-wise sparse attention that achieves superior accuracy-efficiency trade-offs for video models. We observe that video attention maps exhibit a strong vertical-vector sparse pattern, and further demonstrate that this vertical-vector pattern offers consistently better accuracy-sparsity trade-offs compared with existing coarse-grained sparse patterns. Based on this observation, VecAttention dynamically selects and processes only informative vertical vectors through a lightweight important-vector selection that minimizes memory access overhead and an optimized kernel of vector sparse attention. Comprehensive evaluations on video understanding (VideoMME, LongVideoBench, and VCRBench) and generation (VBench) tasks show that VecAttention delivers a 2.65$\times$ speedup over full attention and a 1.83$\times$ speedup over state-of-the-art sparse attention methods, with comparable accuracy to full attention. Our code is available at https://github.com/anminliu/VecAttention.

cs.CV↗

arXiv2026

Predicting Human Mobility during Extreme Events via LLM-Enhanced Cross-City Learning

The vulnerability of cities has increased with urbanization and climate change, making it more important to predict human mobility during extreme events (e.g., extreme weather) for downstream tasks including location-based early disaster warning and pre-allocating rescue resources, etc. However, existing human mobility prediction models are mainly designed for normal scenarios, and fail to adapt to extreme scenarios due to the shift of human mobility patterns under extreme scenarios. To address this issue, we introduce \textbf{X-MLM}, a cross-e\textbf{X}treme-event \textbf{M}obility \textbf{L}anguge \textbf{M}odel framework for extreme scenarios that can be integrated into existing deep mobility prediction methods by leveraging LLMs to model the mobility intention and transferring the common knowledge of how different extreme events affect mobility intentions between cities. This framework utilizes a RAG-Enhanced Intention Predictor to forecast the next intention, refines it with an LLM-based Intention Refiner, and then maps the intention to an exact location using an Intention-Modulated Location Predictor. Extensive experiments illustrate that X-MLM can achieve a 32.8\% improvement in terms of Acc@1 and a 35.0\% improvement in terms of the F1-score of predicting immobility compared to the baselines. The code is available at https://github.com/tsinghua-fib-lab/XMLM.

cs.LG↗

arXiv2026

LEAF: Language-EEG Aligned Foundation Model for Brain-Computer Interfaces

Recent advances in electroencephalography (EEG) foundation models, which capture transferable EEG representations, have greatly accelerated the development of brain-computer interfaces (BCIs). However, existing approaches still struggle to incorporate language instructions as prior constraints for EEG representation learning, limiting their ability to leverage the semantic knowledge inherent in language to unify different labels and tasks. To address this challenge, we present LEAF, a foundation model for EEG--Language Alignment with Semantic Task Instruction and Querying. LEAF integrates task-aware semantic guidance to produce structured and linguistically aligned EEG embeddings, thereby enhancing decoding robustness and transferability. In the pretraining stage, we introduce a joint Spectral--Temporal Reconstruction (STR) framework that captures the coupled spectral rhythms and temporal dynamics of EEG signals. STR applies randomized spectral perturbation to enhance frequency robustness and uses two complementary temporal objectives to learn both contextual and sequential structure. In the EEG-Language alignment stage, we propose the Instruction-conditioned Q-Former (IQF). This query-based cross-attention transformer injects instruction embeddings into EEG tokens and achieves semantic alignment with textual label embeddings through learnable queries. We evaluate LEAF on 16 downstream datasets spanning motor imagery, emotion recognition, steady-state visual evoked potentials, covert speech, and healthcare tasks. LEAF achieves state-of-the-art performance on 12 of the 16 datasets and obtains the best average results across all five task categories. Importantly, our analyses reveal for the first time that explicit task instructions serve as semantic priors guiding EEG embeddings into coherent and linguistically grounded spaces. The code and pre-trained weights will be released.

cs.LG↗

arXiv2026

FingerTip 20K: A Benchmark for Proactive and Personalized Mobile LLM Agents

Mobile GUI agents are becoming critical tools to improve user experience on smart devices, with multimodal large language models (MLLMs) emerging as the dominant paradigms in this domain. Current agents, however, rely on explicit human instructions, overlooking the potential to leverage the contextual information (like location, time, user profile) and historical data for proactive task suggestions. Besides, previous works focus on optimizing the success rate during task execution, but pay less attention to the personalized execution trajectory, thereby neglecting potentially vast differences in user preferences. To address these challenges, we introduce the FingerTip 20K benchmark. We collected 20K unique human demonstrations of multi-step Android device interactions across a variety of everyday apps. These demonstrations are not isolated but are continuously acquired from the users' long-term usage in their real lives, and encompass essential user-related contextual information. The benchmark contains two new tracks: proactive task suggestions by analyzing environment observation and users' previous intents, and personalized task execution by catering to users' action preferences. Our experiments reveal that the tracks we propose pose significant challenges for leveraging user-related information in GUI tasks. We also performed a human study to show that there exists a huge gap between existing agents and humans. The model fine-tuned with the data we collected effectively utilized user information and achieved good results, highlighting the potential of our approach in building more user-oriented mobile LLM agents. Our code is open-source at https://github.com/tsinghua-fib-lab/FingerTip-20K for reproducibility.

cs.HC↗

arXiv2026

UrbanAlign: Post-hoc Semantic Calibration for VLM-Human Preference Alignment

Vision-language models (VLMs) can describe urban scenes in rich detail, yet consistently fail to produce reliable human preference labels in domain-specific tasks such as safety assessment and aesthetic evaluation. The standard fix, fine-tuning or RLHF, requires large-scale annotations and model retraining. We ask a different question: can a frozen VLM be aligned with human preferences without modifying any weights? Our key insight is that VLMs are strong concept extractors but poor decision calibrators. We propose a three-stage post-hoc pipeline that exploits this asymmetry: (i) interpretable evaluation dimensions are automatically mined from consensus exemplars; (ii) an Observer-Debater-Judge chain extracts robust concept scores from the frozen VLM; and (iii) locally-weighted ridge regression on a hybrid manifold calibrates these scores to human ratings. Applied as UrbanAlign on Place Pulse 2.0, the framework reaches 72.2% accuracy (kappa=0.45) across six perception categories, outperforming all baselines by +11.0 pp and zero-shot VLM by +15.5 pp, with full interpretability and zero weight modification.

cs.CV↗

arXiv2026

Debiasing International Attitudes: LLM Agents for Simulating US-China Perception Changes

Large Language Models (LLMs) offer transformative opportunities to address the longstanding challenge of modeling opinion evolution in computational social science. This study investigates how media influences cross-border attitudes - a key driver of global polarization - by developing an LLM-agent framework to disentangle sources of bias and assess LLMs' capacity for human-like opinion formation in response to external information. We introduce an LLM-agent-based framework that models U.S. citizens' attitudes toward China from 2005 to 2025. Our approach integrates large-scale news data with social media profiles to initialize agent populations, which then undergo cognitive-aware reflection and opinion updating. We propose three debiasing mechanisms: (1) fact elicitation, extracting neutral events from subjectively framed news; (2) a devil's advocate agent that simulates critical contextualization; and (3) counterfactual exposure to surface inherent model biases. Simulations with two state-of-the-art LLMs (Qwen3-14b and GPT4o) reveal the expected negative attitudinal trend following media exposure. While all three mechanisms mitigate this trend to varying degrees, results indicate that subjective news framing contributes only modestly to negative attitudes, whereas the devil's advocate agent proves most effective overall, suggesting that intermediate analytical steps can produce more human-like agent opinions. Notably, the counterfactual study reveals contradictory findings across models, suggesting region-specific inherent biases tied to models' geographic origins. By advancing understanding of LLM-based opinion formation and debiasing methods, this study contributes to developing more objective models that better align with human cognitive tendencies.

cs.SI↗

arXiv2026

Generative Diffusion Models for High Dimensional Channel Estimation

Along with the prosperity of generative artificial intelligence (AI), its potential for solving conventional challenges in wireless communications has also surfaced. Inspired by this trend, we investigate the application of the advanced diffusion models (DMs), a representative class of generative AI models, to high dimensional wireless channel estimation. By capturing the structure of multiple-input multiple-output (MIMO) wireless channels via a deep generative prior encoded by DMs, we develop a novel posterior inference method for channel reconstruction. We further adapt the proposed method to recover channel information from low-resolution quantized measurements. Additionally, to enhance the over-the-air viability, we integrate the DM with the unsupervised Stein's unbiased risk estimator to enable learning from noisy observations and circumvent the requirements for ground truth channel data that is hardly available in practice. Results reveal that the proposed estimator achieves high-fidelity channel recovery while reducing estimation latency by a factor of 10 compared to state-of-the-art schemes, facilitating real-time implementation. Moreover, our method outperforms existing estimators while reducing the pilot overhead by half, showcasing its scalability to ultra-massive antenna arrays.

cs.IT↗

arXiv2026

EROICA: Online Performance Troubleshooting for Large-scale Model Training

Troubleshooting performance problems of large model training (LMT) is immensely challenging, due to unprecedented scales of modern GPU clusters, the complexity of software-hardware interactions, and the data intensity of the training process. Existing troubleshooting approaches designed for traditional distributed systems or datacenter networks fall short and can hardly apply to real-world training systems. In this paper, we present EROICA, the first online troubleshooting system that provides both fine-grained observation based on profiling, and coverage of all machines in GPU clusters, to diagnose performance issues in production, including both hardware and software problems (or the mixture of both). EROICA effectively summarizes runtime behavior patterns of LMT function executions via online profiling, and leverages differential observability to localize the root cause with minimal production impact. EROICA has been deployed as a production service for large-scale GPU clusters of ~100,000 GPUs for 1.5 years. It has diagnosed a variety of difficult performance issues with 97.5% success.

cs.DC↗

arXiv2026

GeoNav: Empowering MLLMs with dual-scale geospatial reasoning for language-goal aerial navigation

Language-goal aerial navigation requires UAVs to localize targets in the complex outdoors, such as urban blocks based on textual instructions. The indoor methods are often hard to scale to urban scenes due to ambiguous objects, limited visual field, and spatial reasoning. In this work, we propose GeoNav, a multi-modal agent for long-range aerial navigation with geospatial awareness. GeoNav operates in three phases-landmark navigation, target search, and precise localization-mimicking human coarse-to-fine spatial reasoning patterns. To support such reasoning, it dynamically builds dual-scale spatial representations. The first is a global but schematic cognitive map, which fuses prior geographic knowledge and embodied visual cues into a top-down and explicit annotated form. It enables fast navigation to the landmark region via intuitive map-based reasoning. The second is a local but delicate scene graph representing hierarchical spatial relationships between landmarks and objects, utilized for accurate target localization. On top of the structured memory, GeoNav employs a spatial chain-of-thought mechanism to enable MLLMs with efficient and interpretable action-making across stages. On the CityNav benchmark, GeoNav surpasses the current SOTA up to 18.4% in success rate and significantly eliminates navigation error. The ablation studies highlight the importance of each module, positioning structured spatial perception as the key to advanced UAV navigation. Published in Pattern Recognition, 2026.

cs.RO↗

arXiv2026

An Efficient Self-supervised Seismic Data Reconstruction Method Based on Self-Consistency Learning

Seismic exploration remains the most critical method for characterizing subsurface structures in geophysics. However, complex surface conditions often cause a non-uniform distribution of seismic receivers along survey lines, leading to irregularly acquired seismic data, which affects subsequent processing and inversion. Prior deep learning-based seismic data reconstruction methods typically rely on datasets for supervised training. While some existing methods avoid extra data, they lack effective constraints on reconstructed data, leading to unstable performance. In this study, we propose a self-supervised self-consistency learning strategy with a lightweight network for seismic data reconstruction. Our method requires no extra datasets, and it leverages inter-component correlations in seismic data to design a loss function, optimizing a network with only 188,849 learnable parameters. Validated on two public seismic datasets, results demonstrate our approach yields high-quality reconstruction, providing significant value for large-scale and complex seismic exploration tasks.

eess.IV↗

arXiv2026

HDINO: A Concise and Efficient Open-Vocabulary Detector

Despite the growing interest in open-vocabulary object detection in recent years, most existing methods rely heavily on manually curated fine-grained training datasets as well as resource-intensive layer-wise cross-modal feature extraction. In this paper, we propose HDINO, a concise yet efficient open-vocabulary object detector that eliminates the dependence on these components. Specifically, we propose a two-stage training strategy built upon the transformer-based DINO model. In the first stage, noisy samples are treated as additional positive object instances to construct a One-to-Many Semantic Alignment Mechanism(O2M) between the visual and textual modalities, thereby facilitating semantic alignment. A Difficulty Weighted Classification Loss (DWCL) is also designed based on initial detection difficulty to mine hard examples and further improve model performance. In the second stage, a lightweight feature fusion module is applied to the aligned representations to enhance sensitivity to linguistic semantics. Under the Swin Transformer-T setting, HDINO-T achieves \textbf{49.2} mAP on COCO using 2.2M training images from two publicly available detection datasets, without any manual data curation and the use of grounding data, surpassing Grounding DINO-T and T-Rex2 by \textbf{0.8} mAP and \textbf{2.8} mAP, respectively, which are trained on 5.4M and 6.5M images. After fine-tuning on COCO, HDINO-T and HDINO-L further achieve \textbf{56.4} mAP and \textbf{59.2} mAP, highlighting the effectiveness and scalability of our approach. Code and models are available at https://github.com/HaoZ416/HDINO.

cs.CV↗

arXiv2026

Securing the Floor and Raising the Ceiling: A Merging-based Paradigm for Multi-modal Search Agents

Recent advances in Vision-Language Models (VLMs) have motivated the development of multi-modal search agents that can actively invoke external search tools and integrate retrieved evidence through multi-step reasoning. While promising, existing approaches typically rely on large-scale supervised trajectories or expensive reinforcement learning (RL), leading to high training cost, instability, and a severe cold-start problem for standard VLMs. We propose a training-free paradigm to empower VLMs with autonomous search capabilities via cross-modal model merging. By fusing a text-based search agent with a base VLM, we show that multi-modal search capabilities can be effectively composed without any additional multi-modal training data. To mitigate parameter interference during cross-modal integration, we introduce Optimal Brain Merging (OBM), a saliency-aware merging algorithm that identifies task-critical parameters based on their impact on model loss using only a small set of calibration samples. Extensive experiments on search-intensive benchmarks (e.g., InfoSeek, MMSearch) reveal that: (1) Model merging secures a reasonable performance floor as a zero-shot agent, with OBM achieving superior search rates; (2) OBM significantly raises the performance ceiling as a warm-start strategy, achieving faster convergence and higher peak accuracy than standard VLM initialization.

cs.AI↗

arXiv2026

CityLens: Evaluating Large Vision-Language Models for Urban Socioeconomic Sensing

Understanding urban socioeconomic conditions through visual data is a challenging yet essential task for sustainable urban development and policy planning. In this work, we introduce \textit{CityLens}, a comprehensive benchmark designed to evaluate the capabilities of Large Vision-Language Models (LVLMs) in predicting socioeconomic indicators from satellite and street view imagery. We construct a multi-modal dataset covering a total of 17 globally distributed cities, spanning 6 key domains: economy, education, crime, transport, health, and environment, reflecting the multifaceted nature of urban life. Based on this dataset, we define 11 prediction tasks and utilize 3 evaluation paradigms: Direct Metric Prediction, Normalized Metric Estimation, and Feature-Based Regression. We benchmark 17 state-of-the-art LVLMs across these tasks. These make CityLens the most extensive socioeconomic benchmark to date in terms of geographic coverage, indicator diversity, and model scale. Our results reveal that while LVLMs demonstrate promising perceptual and reasoning capabilities, they still exhibit limitations in predicting urban socioeconomic indicators. CityLens provides a unified framework for diagnosing these limitations and guiding future efforts in using LVLMs to understand and predict urban socioeconomic patterns. The code and data are available at https://github.com/tsinghua-fib-lab/CityLens.

cs.AI↗

arXiv2026

Pinching Antennas for Multiple Access in Multigroup Multicast Communications

This paper aims to design multiple access (MA) schemes to improve the max-min fairness (MMF) for pinching antennas (PAs)-based multigroup multicast communications, where PA placement and resource allocation are jointly optimized. Specifically, three MA schemes are considered to facilitate the multicast transmission: i) treating interference as noise (TIN), ii) non-orthogonal multiple access (NOMA), and iii) time-division multiple access (TDMA) with two PA reconfiguration protocols, namely pinching switching (PS) and pinching multiplexing (PM). i) For TIN, a closed-form solution is derived for optimal power allocation, while a sequential element-wise optimization (SEO) is developed for the PA placement. ii) For NOMA, a recursive power allocation framework incorporating a bisection search is developed, and a hierarchical objective evaluation (HOE) mechanism is incorporated to simplify the SEO process for PA location update. iii) For TDMA, the PS protocol allows the PA locations to be optimized separately using the SEO method, after which the time-power allocation is solved as a convex problem with a global optimum. Under the PM protocol, the PA locations are jointly optimized with the time-power resources through a Karush-Kuhn-Tucker (KKT)-based analytical solution. Numerical results demonstrate that: i) the pinching-antenna system (PASS) architecture significantly outperforms traditional fixed-antenna systems. ii) TDMA-PS achieves superior performance by fully leveraging the flexible PA reconfiguration and benefiting from interference-free transmission, whereas TIN serves as a practical lower-bound solution due to its simplicity despite its limited performance. iii) NOMA consistently outperforms TDMA-PM and, in high transmit power regimes with heterogeneous multicast group distributions, can even surpass the performance achieved by TDMA-PS.

eess.SP↗