Searcharxiv⌕ Search

arXiv subjects

Yi Li

Publications and source records attributed to Yi Li.

At least 109 records · Page 6Linked to original sources
arXiv2026

LEAD: Length-Efficient Adaptive and Dynamic Reasoning for Large Language Models

Large reasoning models, such as OpenAI o1 and DeepSeek-R1, tend to become increasingly verbose as their reasoning capabilities improve. These inflated Chain-of-Thought (CoT) trajectories often exceed what the underlying problems require, wasting compute, latency, and context budgets. While introducing length-based efficiency rewards during reinforcement learning offers a natural remedy, existing methods struggle with two fundamental challenges: the optimal balance between correctness and efficiency is non-stationary throughout training, and intrinsic reasoning budgets vary drastically across problems. Relying on static reward weights and global length constraints inevitably forces a compromise between degraded accuracy and unrealized compression. To overcome these limitations, we propose LEAD (Length-Efficient Adaptive and Dynamic reasoning), a method that replaces static heuristics with online, self-adaptive mechanisms. LEAD dynamically calibrates the correctness-efficiency trade-off at each step using a Potential-Scaled Instability, directing optimization capacity to the most informative learning signal. Furthermore, it estimates an adaptive per-problem target length online based on the model's own correct rollouts, applying a symmetric efficiency reward that penalizes both overthinking and over-compression. Evaluated on five mathematical reasoning benchmarks, LEAD achieves the highest accuracy and Accuracy-Efficiency Score among RL-trained efficient-reasoning methods while producing substantially shorter outputs than the base model.

cs.LG↗
arXiv2026

Winding feature and thermal evolution of the Dirac magnons in CrI$_3$

Two-dimensional honeycomb lattice ferromagnet chromium tri-iodide (CrI$_3$) has attracted tremendous interest because it retains ferromagnetism down to the monolayer limit and hosts intriguing topological magnons. As a prototypical van der Waals magnet, CrI$_3$ provides an ideal platform for exploring the interplay between reduced dimensionality, magnetic order, and nontrivial spin excitations. Here, using inelastic neutron scattering together with improved sample quality, we uncover the magnon winding feature around the $K$-point of the hexagonal Brillouin zone, a key signature of Dirac magnons. In addition, we find that the magnon energy follows a $T^2$-renormalization behavior at elevated temperatures, consistent with magnon-magnon interactions. These results provide previously missing information on the magnon spectrum of CrI$_3$ and further consolidate the topological nature of its spin excitations.

cond-mat.mtrl-sci↗
arXiv2026

Anisotropic Josephson coupling of $d$ vectors in triplet superconductors arising from frustrated spin textures

We demonstrate that coupling itinerant electrons to a noncollinear classical exchange field can induce anisotropic Josephson coupling between superconducting $d$ vectors, analogous to the Dzyaloshinskii-Moriya and $Γ$-type interactions in magnetism. Using perturbative methods, we analyze an $s$-$d$ model on a geometrically frustrated lattice. Noncollinear local spin textures generate spin triplet pairing correlations and can favor spatially varying superconducting order due to anisotropic Josephson couplings between $d$ vectors, endowing a ``pliability'' to the pairing order that competes with the superfluid stiffness. For nonunitary pairing, this spatial texture of $d$ vectors can give rise to anomalous vortices in the absence of an external magnetic field. We further predict a Josephson diode effect with efficiency proportional to the spin chirality of the underlying magnetic texture. These results establish a link between frustrated magnetism and spatial textures of triplet superconducting pairing, with implications for a range of materials such as Mn$_3$Ge and $4H_b$-TaS$_2$, where superconductivity can be proximity-induced or intrinsic.

cond-mat.supr-con↗
arXiv2026

Spatially Inhomogeneous Triplet Pairing Order and Josephson Diode Effect Induced by Frustrated Spin Textures

We show that frustrated spin textures can generate anisotropic Josephson couplings between $d$ vectors that can stabilize spatially varying pairing orders in spin triplet superconductors. These couplings depend on the relative orientation of $d$ vectors, analogous to Dzyaloshinskii-Moriya and $Γ$-type interactions in magnetism, leading to an effective ``pliability'' of the pairing order that competes with superfluid stiffness. Such couplings cannot originate from spin-orbit coupling; rather, they can arise, for example, when itinerant electrons are coupled to a local exchange field composed of frustrated spin moments. Using a $T$-matrix expansion, we show that coupling to a local exchange field leads to an effective tunneling of itinerant electrons that is dependent on the underlying spin configurations at the barrier between superconducting grains. Furthermore, Josephson tunneling through frustrated spin textures can produce a Josephson diode effect. The diode effect originates either from nonvanishing spin chirality in the barrier, or from antisymmetric Josephson coupling between noncollinear $d$ vectors, both of which break inversion and time-reversal symmetries.

cond-mat.supr-con↗
arXiv2026

Berry Phase Enforced Spinor Pairing Order

We introduce a class of topological pairing orders characterized by a half-integer pair monopole charge, leading to Berry phase enforced half-integer partial wave symmetry. This exotic spinor order emerges from pairing between Fermi surfaces with Chern numbers differing by an odd integer. Using tight-binding models, we demonstrate spinor superconducting orders with monopole charges $\pm 1/2$, featuring a single gap node and nontrivial surface states. Additionally, the superfluid velocity follows a fractionalized Mermin-Ho relation in spatially inhomogeneous pairing orders. The concept extends to spinor density waves and excitons.

cond-mat.supr-con↗
arXiv2026

Latent Bridge: Feature Delta Prediction for Efficient Dual-System Vision-Language-Action Model Inference

Dual-system Vision-Language-Action (VLA) models achieve state-of-the-art robotic manipulation but are bottlenecked by the VLM backbone, which must execute at every control step while producing temporally redundant features. We propose Latent Bridge, a lightweight model that predicts VLM output deltas between timesteps, enabling the action head to operate on predicted outputs while the expensive VLM backbone is called only periodically. We instantiate Latent Bridge on two architecturally distinct VLAs: GR00T-N1.6 (feature-space bridge) and π0.5 (KV-cache bridge), demonstrating that the approach generalizes across VLA designs. Our task-agnostic DAgger training pipeline transfers across benchmarks without modification. Across four LIBERO suites, 24 RoboCasa kitchen tasks, and the ALOHA sim transfer-cube task, Latent Bridge achieves 95-100% performance retention while reducing VLM calls by 50-75%, yielding 1.65-1.73x net per-episode speedup.

cs.RO↗
arXiv2026

From Test-taking to Cognitive Scaffolding: A Pedagogical Diagnostic Benchmark for LLMs on English Standardized Tests

As large language models (LLMs) are increasingly integrated into educational tools, current evaluations on standardized tests predominantly focus on binary outcome accuracy. Instead, an effective AI tutor must exhibit faithful reasoning, elucidate solution strategies, and diagnose specific human misconceptions. To bridge this gap, we introduce a pedagogical diagnostic framework that models English Standardized Test (EST) problem-solving as a traversal through a cognitive framework. Based on this framework, we present ESTBook, a multimodal benchmark encompassing 10,576 questions and 29 task types across five major exams. Unlike traditional datasets, ESTBook goes beyond data aggregation by enriching questions with formalized reasoning trajectories and distractor rationales that capture specific cognitive traps. Through extensive evaluations, we empirically demonstrate the practical utility of our diagnostic framework, showing that identifying cognitive trajectories facilitates the mitigation of performance gap and improves pedagogical reasoning through guided elicitation.

cs.CL↗
arXiv2026

Nonintegral Flux Trapping in Frustrated Josephson Networks of Triplet Superconductors

In a Josephson junction network, anisotropic coupling between spin triplet pairing correlations can lead to frustrated $d$ vector textures that support spontaneous Josephson currents and nonintegral flux trapping. Such networks can appear in superconducting polycrystals, as well as single-crystal superconductors. In analogy to classical spin systems, in which the presence of geometric frustration and anisotropic superexchange can lead to nontrivial spin textures, Josephson networks with anisotropic Josephson couplings cannot simultaneously optimize their $\mathrm{U}(1)$ superconducting phase difference and relative $d$ vector orientations. The internal pairing structure of Cooper pairs twists as they tunnel across the Josephson junction, and the $d$ vector texture enters as an emergent geometric phase which can spontaneously trap fractional flux. For unitary triplet pairing order, this mechanism can support $π$-flux trapping above a critical value of antisymmetric Josephson coupling, and is distinct from usual half-quantum vortices. The results of this work reveal new routes to engineer frustrated Josephson networks from the interplay of magnetic textures and spin triplet superconducting pairing order.

cond-mat.supr-con↗
arXiv2026

Radiomics-Guided Vision Transformers for Survival Analysis

Vision Transformers (ViTs) have shown strong empirical performance on high-dimensional medical imaging data, yet their behavior under survival objectives and the interpretability of their attention mechanisms remain poorly understood. Under shallow ViTs, we design controlled experiments showing that token-level attention dynamics can recover outcome-relevant regions and that attention-based thresholding enables effective token pruning, improving both interpretability and predictive performance. We also study pretrained deep ViTs for survival analysis and propose a radiomics-guided hybrid model that integrates pixel-based embeddings with interpretable radiomic features through a multimodal Cox framework and contrastive alignment. Applied to a COVID-19 chest X-ray cohort with a composite ICU admission or mortality endpoint, the proposed approach achieves competitive discrimination while providing clinically meaningful attention maps and feature-group importance.

physics.med-ph↗
arXiv2026

Multi-modal Test-time Adaptation via Adaptive Probabilistic Gaussian Calibration

Multi-modal test-time adaptation (TTA) enhances the resilience of benchmark multi-modal models against distribution shifts by leveraging the unlabeled target data during inference. Despite the documented success, the advancement of multi-modal TTA methodologies has been impeded by a persistent limitation, i.e., the lack of explicit modeling of category-conditional distributions, which is crucial for yielding accurate predictions and reliable decision boundaries. Canonical Gaussian discriminant analysis (GDA) provides a vanilla modeling of category-conditional distributions and achieves moderate advancement in uni-modal contexts. However, in multi-modal TTA scenario, the inherent modality distribution asymmetry undermines the effectiveness of modeling the category-conditional distribution via the canonical GDA. To this end, we introduce a tailored probabilistic Gaussian model for multi-modal TTA to explicitly model the category-conditional distributions, and further propose an adaptive contrastive asymmetry rectification technique to counteract the adverse effects arising from modality asymmetry, thereby deriving calibrated predictions and reliable decision boundaries. Extensive experiments across diverse benchmarks demonstrate that our method achieves state-of-the-art performance under a wide range of distribution shifts. The code is available at https://github.com/XuJinglinn/AdaPGC.

cs.CV↗
arXiv2026

Structure Guided Retrieval-Augmented Generation for Factual Queries

Retrieval-Augmented Generation (RAG) has been proposed to mitigate hallucinations in large language models (LLMs), where generated outputs may be factually incorrect. However, existing RAG approaches predominantly rely on vector similarity for retrieval, which is prone to semantic noise and fails to ensure that generated responses fully satisfy the complex conditions specified by factual queries, often leading to incorrect answers. To address this challenge, we introduce a novel research problem, named Exact Retrieval Problem (ERP). To the best of our knowledge, this is the first problem formulation that explicitly incorporates structural information into RAG for factual questions to satisfy all query conditions. For this novel problem, we propose Structure Guided Retrieval-Augmented Generation (SG-RAG), which models the retrieval process as an embedding-based subgraph matching task, and uses the retrieved topological structures to guide the LLM to generate answers that meet all specified query conditions. To facilitate evaluation of ERP, we construct and publicly release Exact Retrieval Question Answering (ERQA), a large-scale dataset comprising 120000 fact-oriented QA pairs, each involving complex conditions, spanning 20 diverse domains. The experimental results demonstrate that SG-RAG significantly outperforms strong baselines on ERQA, delivering absolute improvements from 20.68 to 50.88 points across all evaluation metrics, while maintaining reasonable computational overhead.

cs.IR↗
arXiv2026

Neurosymbolic Repo-level Code Localization

Code localization is a cornerstone of autonomous software engineering. Recent advancements have achieved impressive performance on real-world issue benchmarks. However, we identify a critical yet overlooked bias: these benchmarks are saturated with keyword references (e.g. file paths, function names), encouraging models to rely on superficial lexical matching rather than genuine structural reasoning. We term this phenomenon the Keyword Shortcut. To address this, we formalize the challenge of Keyword-Agnostic Logical Code Localization (KA-LCL) and introduce KA-LogicQuery, a diagnostic benchmark requiring structural reasoning without any naming hints. Our evaluation reveals a catastrophic performance drop of state-of-the-art approaches on KA-LogicQuery, exposing their lack of deterministic reasoning capabilities. We propose LogicLoc, a novel agentic framework that combines large language models with the rigorous logical reasoning of Datalog for precise localization. LogicLoc extracts program facts from the codebase and leverages an LLM to synthesize Datalog programs, with parser-gated validation and mutation-based intermediate-rule diagnostic feedback to ensure correctness and efficiency. The validated programs are executed by a high-performance inference engine, enabling accurate and verifiable localization in a fully automated, closed-loop workflow. Experimental results demonstrate that LogicLoc significantly outperforms SOTA methods on KA-LogicQuery while maintaining competitive performance on popular issue-driven benchmarks. Notably, LogicLoc attains superior performance with significantly lower token consumption and faster execution by offloading structural traversal to a deterministic engine, reducing the overhead of iterative LLM inference.

cs.SE↗
arXiv2026

MAGMA: A Multi-Graph based Agentic Memory Architecture for AI Agents

Memory-Augmented Generation (MAG) extends Large Language Models with external memory to support long-context reasoning, but existing approaches largely rely on semantic similarity over monolithic memory stores, entangling temporal, causal, and entity information. This design limits interpretability and alignment between query intent and retrieved evidence, leading to suboptimal reasoning accuracy. In this paper, we propose MAGMA, a multi-graph agentic memory architecture that represents each memory item across orthogonal semantic, temporal, causal, and entity graphs. MAGMA formulates retrieval as policy-guided traversal over these relational views, enabling query-adaptive selection and structured context construction. By decoupling memory representation from retrieval logic, MAGMA provides transparent reasoning paths and fine-grained control over retrieval. Experiments on LoCoMo and LongMemEval demonstrate that MAGMA consistently outperforms state-of-the-art agentic memory systems in long-horizon reasoning tasks.

cs.AI↗
arXiv2026

Enhancing Inference for Small Cohorts via Transfer Learning and Weighted Integration of Multiple Datasets

Lung sepsis remains a significant concern in the Northeastern U.S., yet the national eICU Collaborative Database includes only a small number of patients from this region, highlighting underrepresentation. Understanding clinical variables such as FiO2, creatinine, platelets, and lactate, which reflect oxygenation, kidney function, coagulation, and metabolism, is crucial because these markers influence sepsis outcomes and may vary by sex. Transfer learning helps address small sample sizes by borrowing information from larger datasets, although differences in covariates and outcome-generating mechanisms between the target and external cohorts can complicate the process. We propose a novel weighting method, TRANSfer LeArning wiTh wEights (TRANSLATE), to integrate data from various sources by incorporating domain-specific characteristics through learned weights that align external data with the target cohort. These weights adjust for cohort differences, are proportional to each cohort's effective sample size, and downweight dissimilar cohorts. TRANSLATE offers theoretical guarantees for improved precision and applies to a wide range of estimands, including means, variances, and distribution functions. Simulations and a real-data application to sepsis outcomes in the Northeast cohort, using a much larger sample from other U.S. regions, show that the method enhances inference while accounting for regional heterogeneity.

stat.ME↗
arXiv2026

Intelligent bear deterrence system based on computer vision: Reducing human-bear conflicts in remote areas

Human-bear conflicts on the Tibetan Plateau threaten both local livelihoods and the conservation of Tibetan brown bears (Ursus arctos pruinosus). To address this challenge, we developed a low-power, network-independent deterrence system that combines computer vision with Internet of Things (IoT) hardware. The system integrates a YOLOv5-MobileNet detection model deployed on a low-power edge artificial intelligence (AI) board with a solar-powered bear spray device. We compiled a data set of 1,243 wildlife images (including 795 bears with 100 infrared captures for nighttime detection, plus other common objects and animals such as mastiffs, yaks, humans, and vehicles), from which 80% were used for training and 20% for validation. Validation showed robust performance (mean average precision = 91.4%, recall = 93.6%). In 100 controlled activation tests involving simulated approaches by bears, humans, and other animals, the spray deployed within 0.2 seconds of detection with 97.2% accuracy, confirming timely and reliable responses. A 30-day field trial in Zadoi County, Qinghai Province, China, recorded 3 successful deterrence events without false activations. By using energy-efficient components and ensuring continuous and stable system operation, this solution provides a practical, sustainable, and scalable approach to mitigating human-bear conflicts, effectively enhancing human safety and bear conservation in remote areas without network or grid coverage.

cs.CV↗
arXiv2026

Estimating heterogeneous treatment effects with survival outcomes via a deep survival learner

Estimating heterogeneous treatment effects in survival settings is complicated by right censoring as well as the time-varying nature of the estimand. While the conditional average treatment effect (CATE) provides a natural target, most existing approaches focus on a single prespecified time point and do not account for the temporal trajectory, leading to instability in estimation. We propose a deep survival learner (DSL) for estimating heterogeneous treatment effects with right-censored outcomes. The method is based on a doubly robust pseudo-outcome whose conditional expectation identifies time-specific CATEs under standard assumptions. This construction remains unbiased if either the outcome model or the treatment assignment model is correctly specified, when properly accounting for censoring. To estimate CATEs over a clinically relevant time spectrum, DSL employs a multi-output deep neural network with shared representations, enabling joint estimation of treatment effect trajectories. From a theoretical perspective, we derive error bounds for both pointwise and joint estimation over time. We show that joint estimation can leverage temporal structure to control estimation error without incurring much additional approximation cost under smoothness conditions, leading to improved stability relative to separate estimation. Cross-fitting is incorporated to reduce overfitting and mitigate bias arising from flexible nuisance estimation. Simulation studies demonstrate favorable finite-sample performance, particularly under nuisance model misspecification. Applied to the Boston Lung Cancer Study, DSL reveals heterogeneity in the effects of perioperative chemotherapy across patient characteristics and over time.

stat.ME↗
arXiv2026

ReadMOF: Structure-Free Semantic Embeddings from Systematic MOF Nomenclature for Machine Learning

Systematic chemical names, such as IUPAC-style nomenclature for metal-organic frameworks (MOFs), contain rich structural and compositional information in a standardized textual format. Here we introduce ReadMOF, which is, to our knowledge, the first nomenclature-free machine learning framework that leverages these names to model structure-property relationships without requiring atomic coordinates or connectivity graphs. By employing pretrained language models, ReadMOF converts systematic MOF names from the Cambridge Structural Database (CSD) into vector embeddings that closely represent traditional structure-based descriptors. These embeddings enable applications in materials informatics, including property prediction, similarity retrieval, and clustering, with performance comparable to geometry-dependent methods. When combined with large language models, ReadMOF also establishes chemically meaningful reasoning ability with textual input only. Our results show that structured chemical language, interpreted through modern natural language processing techniques, can provide a scalable, interpretable, and geometry-independent alternative to conventional molecular representations. This approach opens new opportunities for language-driven discovery in materials science.

cs.LG↗
arXiv2026

MedConclusion: A Benchmark for Biomedical Conclusion Generation from Structured Abstracts

Large language models (LLMs) are widely explored for reasoning-intensive research tasks, yet resources for testing whether they can infer scientific conclusions from structured biomedical evidence remain limited. We introduce $\textbf{MedConclusion}$, a large-scale dataset of $\textbf{5.7M}$ PubMed structured abstracts for biomedical conclusion generation. Each instance pairs the non-conclusion sections of an abstract with the original author-written conclusion, providing naturally occurring supervision for evidence-to-conclusion reasoning. MedConclusion also includes journal-level metadata such as biomedical category and SJR, enabling subgroup analysis across biomedical domains. As an initial study, we evaluate diverse LLMs under conclusion and summary prompting settings and score outputs with both reference-based metrics and LLM-as-a-judge. We find that conclusion writing is behaviorally distinct from summary writing, strong models remain closely clustered under current automatic metrics, and judge identity can substantially shift absolute scores. MedConclusion provides a reusable data resource for studying scientific evidence-to-conclusion reasoning. Our code and data are available at: https://github.com/Harvard-AI-and-Robotics-Lab/MedConclusion.

cs.CL↗