Searcharxiv⌕ Search

arXiv subjects

Dian Jin

Publications and source records attributed to Dian Jin.

At least 19 recordsLinked to original sources
arXiv2026

MDSKILL: Learning to Internalize Skill Memory through Multi-View Distillation for Efficient Dynamic Agent Self-Evolution

Skill memories encode past experiences as textual prompts to improve reinforcement learning (RL) for self-evolving agents. Complementing RL, On-Policy Self-Distillation (OPSD) provides dense, fine-grained supervision to internalize knowledge from skill-guided trajectories and accelerate policy learning. However, self-generated memories often contain noisy supervision and functionally entangled, redundant, or conflicting knowledge, making direct distillation difficult. Under limited training budgets, these issues can hinder knowledge internalization and even degrade performance. To address this problem, we propose MDSKILL, a multi-view distillation approach for skill memory internalization. MDSKILL organizes skill memories into three complementary functional views (state grounding, action planning, and execution control) and selects high-quality skills using dynamic metrics to construct view-specific and integrated reasoning branches. OPSD then transfers and consolidates complementary experiential knowledge across these branches into a shared policy, enabling efficient skill internalization and continuous agent self-evolution. We evaluate MDSKILL on three agent benchmarks against RL, skill-augmented, and self-distillation baselines that do not rely on stronger external teacher models. Without additional computational overhead, MDSKILL improves upon SOTA performance by up to 7.1% and 8.1% on ALFWorld and WebShop. Extensive ablations and evolution analyses further validate its effectiveness in knowledge internalization and policy improvement. The code is available at https://anonymous.4open.science/r/MDSKILL-CF53.

cs.CL↗
arXiv2026

RecRSI: Mechanism-Guided Discovery of Recommendation Algorithms

Improving recommendation algorithms for the Web requires coordinating interacting design choices and learning from experiments. We present RecRSI, an LLM-based research framework that uses mechanism hypotheses to organize algorithm discovery. Each hypothesis links a proposed mechanism change to a reference design, a rationale, and an expected observation. MechSpace derives implementable primitives and composition rules from the task source; Research Line uses this space and prior evidence to develop and test hypotheses; Mechanism Feedback carries outcomes and their design context into subsequent research. This organization makes each experiment inform which mechanisms to develop, revise, or combine next. Across four initial recommenders and four datasets, RecRSI yields average relative improvements of 8.48% in NDCG@5 and 6.88% in HR@5 over the initial algorithms, and 2.60% and 2.20%, respectively, over the strongest competing result for each setting and metric. Component and process studies link these gains to coordinated mechanism choices and evidence-guided follow-up. Our code is publicly available at https://github.com/DynamisG210/RecRSI.

cs.IR↗
arXiv2026

SkillScaffold: Scaffolding Agentic Reinforcement Learning with Policy-Adaptive Skill Credit Assignment and Selective Injection

Agentic reinforcement learning (RL) enables large language models (LLMs) to learn from multi-turn interactions with verifiable outcomes. Textual skills can improve exploration by distilling prior interactions into reusable guidance. However, existing skill-based RL approaches often either assume skill injection is beneficial or rely on indirect feedback or rollout-intensive evaluation to assess its utility. Our pilot studies show that rollout-based utility estimation for skill credit assignment is costly and unstable under limited sampling, while skill benefits vary across candidates and depend on the current policy's competence on each query. To address these challenges, we introduce SkillScaffold, a policy-adaptive training scaffold that combines low-cost skill-utility estimation with selective skill injection. By reusing outcome-verified training trajectories, SkillScaffold generates candidate skills and estimates their utility through outcome-aware likelihood comparisons, without candidate-specific environment rollouts. The resulting utility guides whether and which skill to inject, retaining skill-free execution when no candidate is estimated to help. Joint training enables the Generator to produce skills tailored to the evolving Executor. Scaffolded RL encourages skill internalization for skill-free inference. Experiments on ALFWorld, WebShop, ScienceWorld, and $τ^2$-Bench demonstrate gains over strong agentic-RL and skill-based RL baselines across both short- and long-horizon tasks. Ablations and further analyses support the importance of utility-based selection and abstention. Our code is available at https://github.com/Vaidurya00/SkillScaffold.

cs.LG↗
arXiv2026

olBlend: Language-Guided Molecular Blend Retrieval with Mixture Odor Prediction

Converting open-ended natural-language odor descriptions into molecular mixtures that can be physically prepared from an available inventory remains an important challenge in computational olfaction. We propose olBlend, a language-guided framework for molecular mixture retrieval. olBlend first uses MixPOM to predict odor probabilities for candidate molecular pairs and stores these predictions for retrieval. It then uses a language model to translate each odor description into three types of semantic requirements: lead attributes that define the main odor, ambient attributes that describe its background, and absent attributes that should be avoided. A structured scoring algorithm uses the precomputed odor probabilities to score how well each candidate mixture meets these requirements, then retrieves and ranks mixtures from the available inventory. Evaluation on established odor datasets shows that MixPOM improves mixture odor prediction over averaged single-molecule predictions. Retrieval comparisons across natural-language queries demonstrate how background and exclusion requirements guide mixture selection. We further prepare retrieved formulations as physical odor samples, demonstrating the workflow from natural-language descriptions to realizable molecular formulations.

cs.HC↗
arXiv2026

SLICE-Mem: Salience-Gated Lightweight Information Curation for Efficient LLM Agent Memory

Long-term memory enables large language model (LLM) agents to maintain personalized and temporally coherent interactions, but repeated LLM inference for memory admission, extraction, and organization introduces substantial cost and latency. We present SLICE-Mem, a lightweight memory framework that combines salience-guided admission and predicate-centered information extraction. A compact classifier identifies user turns worth retaining, and a semantic-dependency-based extractor converts admitted turns into concise memory entries. This online pipeline requires no generative LLM calls, while cross-entry reconciliation is deferred to an offline consolidation stage. We evaluate SLICE-Mem on the weekly subset of Memora and LongMemEval-S using 8B and 30B Qwen backbones for memory construction and consolidation on both benchmarks. Compared with the generative-memory baseline MemOS under the Qwen3-30B configuration, SLICE-Mem's online pipeline achieves 46.60 versus 46.68 FAMA (Forgetting-Aware Memory Accuracy) on Memora Weekly and 56.8% versus 59.3% accuracy on LongMemEval-S, while reducing total LLM token consumption by 92.4% and 99.5%, respectively. These results demonstrate that SLICE-Mem can preserve useful evidence for long-term memory with substantially less reliance on LLM-based memory construction.

cs.CL↗
arXiv2026

Federated Independent Component Analysis via Spectral Alignment and Robust Aggregation

This paper studies robust estimation for Independent Component Analysis (ICA) in the federated learning setting, where data are locally distributed across clients and may exhibit substantial heterogeneity. The goal is to recover a common global mixing matrix by aggregating local estimators computed by individual clients. The main difficulty is that local ICA estimators are identifiable only up to sign flips and column permutations and may have highly heterogeneous estimation quality. We propose a novel three-step aggregation method that first aligns the permutations across all local estimators via a particular spectral clustering approach, then aligns the signs within each estimated cluster via another tailored spectral approach, and finally applies the geometric median for robust aggregation. The proposed estimator is shown to remain accurate even when a substantial fraction of local estimators are of low quality or inconsistent, as long as each cluster contains a majority of accurate estimators. This contrasts with its analogue based on simple averaging, whose performance is determined by the worst local estimator. In the homogeneous setting, the robust estimator is also shown to be minimax optimal, up to a logarithmic factor, both when the number of clients remains fixed and when it diverges. The theoretical findings are corroborated by simulation studies and a real-data analysis.

cs.LG↗
arXiv2026

Towards Efficient Reasoning: Learning Causal Shortcuts for Diffusion Language Models

Diffusion Language Models (DLMs) have attracted significant attention for their strong reasoning ability. However, under a bidirectional attention mechanism, DLMs operate over an exponentially large exploration space compared to autoregressive models (ARMs), making it challenging to focus on reasoning-guiding tokens under random masking. We define causal shortcuts as token chains that cover the full sequence and provide explicit guidance towards correct reasoning trajectories. We analyze the effects of causal shortcuts on the reasoning accuracy and convergence speed of DLMs, and find that they largely improve answer convergence efficiency and generation accuracy. Motivated by this, we propose a Causal Shortcut Learning (CSL) Framework for DLMs. Specifically, we introduce a step-by-step token extraction procedure to extract causal shortcuts from data, and apply parallel prioritized masking on these tokens during training to enable efficient and accurate convergence to correct answers via causal shortcuts. Extensive experiments across multiple reasoning benchmarks and two base models demonstrate that CSL consistently outperforms existing SFT-variant baselines, achieving an average improvement of $1.92\%$ over SFT-only models, and up to $4.20\%$ on MATH-500. The code is available at the \href{https://github.com/ZJUDianJin/Causal-Shortcuts-Learning}{https://github.com/ZJUDianJin/Causal-Shortcuts-Learning

cs.CL↗
arXiv2026

Personalizing Large Language Model Agents with Small Policy Models

Large language model (LLM) agents can retrieve memory, call tools, ask clarifying questions, and vary response style, yet adapting these execution decisions to an individual user remains difficult. Fine-tuning a separate LLM is costly or impossible for proprietary systems, while prompts and memory primarily expose user information to the agent rather than adapt its execution decisions from feedback. We formulate personalization of a frozen agent as online learning of a per-user execution policy from scalar feedback observed only for the executed action. We propose FABLE (Factorized Adaptive Bandit Layer for Execution), a lightweight policy layer outside a potentially black-box host agent. FABLE factorizes memory, information-acquisition, and response decisions so feedback updates related choices; filters actions through an externally specified feasible set before exploration; and learns user-specific residual preferences relative to a fixed default-and-cost score via Bayesian contextual Thompson sampling. Under a linear residual-reward model, a calibrated variant inherits an expected-regret bound against the best feasible action. We also characterize preferences unidentifiable under persistent feasibility constraints and provide anytime-valid false-promotion control. Across personalized-reasoning, controlled-feedback, and executable tool-use evaluations, FABLE improves several preference-sensitive behaviors relative to rule-only control while remaining competitive on end-to-end task performance.

cs.AI↗
arXiv2026

Local Maxima of the Entrywise $\ell_4$ Norm on the Orthogonal Group

We classify the local maximizers of the entrywise fourth-power objective \[ Q\longmapsto \lVert Q\rVert_4^4 =\sum_{i,j=1}^r q_{ij}^4 \] over the real orthogonal group $\mathcal O(r)$. We prove that the signed permutation matrices are the only local maximizers, and hence the only global maximizers, in every dimension. More strongly, every other stationary point has an explicit rank-two tangent direction with strictly positive second variation. The proof is based on a maximum-entry pivot for the orthostochastic matrix $Q^{\circ2}$: the associated full Riemannian Hessian can be evaluated exactly and is positive at a largest nonunit squared entry. The argument is self-contained and handles zeros, repeated magnitudes, reducible support, and Hadamard-type stationary points.

eess.SP↗
arXiv2026

Accelerating discovery of infrared nonlinear optical materials with large shift current via high-throughput screening

Discovering nonlinear optical (NLO) materials with strong shift current response, particularly in the infrared (IR) regime, is essential for next-generation optoelectronics yet remains highly challenging in both experiments and theory, which still largely relies on case by case studies. Here, we employ a high-throughput screening strategy, applying a multi-step filter to the Materials Project database (>154,000 materials), which yielded 2,519 candidate materials for detailed first-principle evaluation. From these calculations, we identify 32 NLO materials with strong shift current response ($σ$ > 100 $μA/V^2$). Our work reveals that layered structures with $C_{3v}$ symmetry and heavy $p$-block elements (e.g. Te, Sb) exhibit apparent superiority in enhancing shift current. More importantly, 9 of these compounds show shift current response peaks in the IR region, with the strongest reaching 616 $μA/V^2$, holding significant application potential in fields such as IR photodetection, sensing, and energy harvesting. Beyond identifying promising candidates, this work establishes a comprehensive and high-quality first-principles dataset for NLO response, providing a solid foundation for future AI-driven screening and accelerated discovery of high-performance NLO materials, as demonstrated by a prototype machine-learning application.

cond-mat.mtrl-sci↗
arXiv2026

Optimizing Denoising Trajectories in dLLMs: A Lightweight Evolutionary Heuristic Approach

Diffusion Large Language Models (dLLMs) have recently emerged as a promising alternative to conventional Auto-Regressive (AR) Large Language Models (LLMs). By leveraging bidirectional attention and parallel decoding, dLLMs enable more efficient generation. However, they require a carefully designed denoising scheduler at inference time (absent during training) whose choice significantly impacts generation quality. While confidence-based heuristic schedulers have shown strong empirical performance, they suffer from two critical failure modes: EOS Overflow and Proximal Bias. Through in-depth analysis of the Transformer's attention patterns, we reveal that these failures stem from certain positions assigning disproportionately high attention weights to invalid tokens (e.g., [MASK] and [EOS]), which produce misleading confidence signals. Building on this insight, empirical evidence shows that valid attention scores can provide complementary guidance to conventional confidence-based heuristics, yet no single metric consistently excels across all scenarios, implying that the optimal denoising trajectory is highly context-dependent. To address this problem, we propose a lightweight evolutionary heuristic scheduler optimized using the Covariance Matrix Adaptation Evolution Strategy (CMA-ES). Our scheduler dynamically integrates multiple heuristic features with a contextual mean-field embedding, while requiring only 393 trainable parameters. Evaluated on LLaDA and Dream across four reasoning and planning benchmarks, our method consistently outperforms strong baselines, including conventional heuristics, block auto-regressive methods, and recent State-Of-The-Art (SOTA) approaches. To the best of our knowledge, it represents the most parameter-efficient neural scheduler to date. Our code is available at https://github.com/RS2002/Evo-Denoise .

cs.CL↗
arXiv2026

On the sensitivity of the subspace predictor to behavioral perturbations

Behavioral systems define discrete-time LTI systems in terms of a set of trajectories, which forms a linear subspace. This subspace underlies the subspace predictor used in data-driven prediction and control. In practice, such subspaces are typically represented through data matrices. For robustness certification and uncertainty quantification, however, these matrix representations are coordinate-dependent and therefore do not provide a coordinate-free way to quantify uncertainty. In this work, we derive an explicit prediction error bound in terms of behavioral distance between the true subspace and an estimate, showing that the predictor is locally Lipschitz with respect to behavioral perturbations. We also present a one-step prediction error bound that is relevant for receding-horizon implementations, which becomes computable when combined with existing behavioral-distance certificates. Numerical studies show that our bound is tighter than an existing data-matrix perturbation bound, and remains computable, though more conservative, when combined with an existing behavioral distance certificate.

math.OC↗
arXiv2026

A Novel Method with Encoder-Decoder for Cross-Sensor Adaptation in Surface Shape Sensing with Sparse Strain Sensors

Performance variations in sensor arrays, caused by intrinsic differences or installation conditions, can lead to inconsistent results during shape sensing. To obtain accurate results, a large amount of data is usually required, and a separate model must be retrained for each sensor array, thereby increasing the cost and time of data acquisition, transmission, and computation. To address this issue, this work proposes an encoder-decoder architecture for surface shape sensing based on sparse strain sensors and further incorporates meta-learning and few-shot adaptation strategies to enable adaptation across different groups of sensor arrays. Experimental results demonstrate that, after the cross-sensor adaptation, a newly deployed sensor array achieves a sensing error of approximately 4.0 mm relying on less than 5.0% newly labeled data and requiring an adaptation time of under 1 second, which represents a substantial improvement from 23.0 mm error without adaptation and 20-minute data collection time required to train a new model. Moreover, the number of points with errors below 5.0 mm increased by more than 65.0%. These results indicate that the proposed method can substantially reduce the cost and training burden of surface shape sensing, and it has broad potential applications in soft robotics and wearable devices.

cs.RO↗
arXiv2026

MTAVG-Bench 2.0: Diagnosing Failure Modes of Cinematic Expressiveness in Multi-Talker Audio-Video Generation

In recent years, Multi-Talker Audio-Video Generation (MTAVG) models have shown promising performance on fundamental metrics such as lip-sync and audio-visual alignment. However, these metrics remain insufficient for assessing cinematic expressiveness in scene-level generation. In multi-character scenes, generation models must go beyond audio-visual realism to convey coherent character performance and other higher-level cinematic qualities. To fill this gap, we introduce MTAVG-Bench 2.0, a benchmark for diagnosing failure modes of cinematic expressiveness in multi-talker audio-video generation. Unlike prior settings that mainly focus on the quality of basic multi-turn dialogue, MTAVG-Bench 2.0 targets short-drama and scene-level generation, and establishes a high-level failure taxonomy spanning acting, narrative, atmosphere, and audio-visual language. Based on this taxonomy, we construct more than 10,000 question-answering evaluation instances, together with subsets for short-drama-level assessment and temporal localization of failure modes, to systematically evaluate the ability of omni large language models to diagnose high-level audio-visual failures. Experimental results show that commercial omni models such as Gemini substantially outperform other evaluators, yet even the strongest models continue to struggle with complex failures in our benchmark. These results demonstrate that MTAVG-Bench 2.0 provides a systematic benchmark for failure diagnosis in cinematic multi-talker audio-video generation.

cs.AI↗
arXiv2026

Stage Light is Sequence$^2$: Multi-Light Control via Imitation Learning

Music-inspired Automatic Stage Lighting Control (ASLC) has gained increasing attention in recent years due to the substantial time and financial costs associated with hiring and training professional lighting engineers. However, existing methods suffer from several notable limitations: the low interpretability of rule-based approaches, the restriction to single-primary-light control in music-to-color-space methods, and the limited transferability of music-to-controlling-parameter frameworks. To address these gaps, we propose SeqLight, a hierarchical deep learning framework that maps music to multi-light Hue-Saturation-Value (HSV) space. Our approach first customizes SkipBART, an end-to-end single primary light generation model, to predict the full light color distribution for each frame, followed by hybrid Imitation Learning (IL) techniques to derive an effective decomposition strategy that distributes the global color distribution among individual lights. Notably, the light decomposition module can be trained under varying venue-specific lighting configurations using only mixed light data and no professional demonstrations, thereby flexibly adapting across diverse venues. In this stage, we formulate the light decomposition task as a Goal-Conditioned Markov Decision Process (GCMDP), construct an expert demonstration set inspired by Hindsight Experience Replay (HER), and introduce a three-phase IL training pipeline, achieving strong generalization capability. To validate our IL solution for the proposed GCMDP, we conduct a series of quantitative analysis and human study. The code and trained models are provided at https://github.com/RS2002/SeqLight .

cs.MM↗
arXiv2026

Zero-Effort Image-to-Music Generation: An Interpretable RAG-based VLM Approach

Recently, Image-to-Music (I2M) generation has garnered significant attention, with potential applications in fields such as gaming, advertising, and multi-modal art creation. However, due to the ambiguous and subjective nature of I2M tasks, most end-to-end methods lack interpretability, leaving users puzzled about the generation results. Even methods based on emotion mapping face controversy, as emotion represents only a singular aspect of art. Additionally, most learning-based methods require substantial computational resources and large datasets for training, hindering accessibility for common users. To address these challenges, we propose the first Vision Language Model (VLM)-based I2M framework that offers high interpretability and low computational cost. Specifically, we utilize ABC notation to bridge the text and music modalities, enabling the VLM to generate music using natural language. We then apply multi-modal Retrieval-Augmented Generation (RAG) and self-refinement techniques to allow the VLM to produce high-quality music without external training. Furthermore, we leverage the generated motivations in text and the attention maps from the VLM to provide explanations for the generated results in both text and image modalities. To validate our method, we conduct both human studies and machine evaluations, where our method outperforms others in terms of music quality and music-image consistency, indicating promising results. Our code is available at https://github.com/RS2002/Image2Music .

cs.SD↗
arXiv2026

Cost-optimal Sequential Testing via Doubly Robust Q-learning

Clinical decision-making often involves selecting tests that are costly, invasive, or time-consuming, motivating individualized, sequential strategies for what to measure and when to stop ascertaining. We study the problem of learning cost-optimal sequential decision policies from retrospective data, where test availability depends on prior results, inducing informative missingness. Under a sequential missing-at-random mechanism, we develop a doubly robust Q-learning framework for estimating optimal policies. The method introduces path-specific inverse probability weights that account for heterogeneous test trajectories and satisfy a normalization property conditional on the observed history. By combining these weights with auxiliary contrast models, we construct orthogonal pseudo-outcomes that enable unbiased policy learning when either the acquisition model or the contrast model is correctly specified. We establish oracle inequalities for the stage-wise contrast estimators, along with convergence rates, regret bounds, and misclassification rates for the learned policy. Simulations demonstrate improved cost-adjusted performance over weighted and complete-case baselines, and an application to a prostate cancer cohort study illustrates how the method reduces testing cost without compromising predictive accuracy.

stat.ML↗
arXiv2026

Online Subspace Learning on Flag Manifolds for System Identification

Data-driven control methods based on subspace representations are powerful but are often limited to linear time-invariant systems where the model order is known. A key challenge is developing online data-driven control algorithms for time-varying systems, especially when the system's complexity is unknown or changes over time. To address this, we propose a novel online subspace learning framework that operates on flag manifolds. Our algorithm leverages streaming data to recursively track an ensemble of nested subspaces, allowing it to adapt to varying system dimensions without prior knowledge of the true model order. We show that our algorithm is a generalization of the Grassmannian Recursive Algorithm for Tracking. The learned subspace models are then integrated into a data-driven simulation framework to perform prediction for unknown dynamical systems. The effectiveness of this approach is demonstrated through a case study where the proposed adaptive predictor successfully handles abrupt changes in system dynamics and outperforms several baselines.

math.OC↗