arXiv subjects

Ye Jin

Publications and source records attributed to Ye Jin.

18 recordsLinked to original sources

arXiv2026

RareDxR1: Autonomous Medical Reasoning for Rare Disease Diagnosis Beyond Human Annotation

Rare disease differential diagnosis is a critical yet arduous clinical task, requiring physicians to identify precise phenotypes from complex, unstructured patient symptoms and execute intricate reasoning within a vast search space. However, existing AI approaches typically rely on pipeline-based phenotype extraction or retrieval-augmented generation, which suffer from critical information loss due to predefined ontologies, retrieval bottlenecks, and a lack of diagnostic logic. To address these challenges, we introduce RareDxR1, an end-to-end reasoning-centric large language model designed for open-domain rare disease diagnosis directly from unstructured clinical notes. We design a progressive end-to-end training framework by synergizing knowledge internalization with autonomous evolutionary learning, thereby bypassing reliance on structured phenotypes and closed-set decision-making. To overcome the limitations of RAG and phenotype restriction, we enabled the deep internalization of fragmented rare-disease knowledge directly into the model's parameters. Moreover, to bridge the gap between model generation and expert reasoning, we propose Reflection-Enhanced Reasoning Sampling (RERS), a strategy that synthesizes expert-level diagnostic trajectories by learning from failures without human annotation. Additionally, we propose a dual-level curriculum reinforcement learning approach for gradually mastering rare disease diagnosis. Experimental results demonstrate that RareDxR1 achieves state-of-the-art accuracy across different benchmarks, marking a significant breakthrough in open-domain rare disease diagnosis. Our code and dataset will be publicly available.

cs.AI↗

arXiv2026

G2DP: Diffusion Planning with Spatio-Temporal Grid Guidance

In autonomous driving, diffusion-based planners have emerged as a promising paradigm for robust motion planning in dense and interactive traffic, as they can effectively model diverse driving behaviors. However, their inherent stochasticity often requires explicit guidance during denoising to ensure safety and route adherence for robust closed-loop execution. Existing guidance typically relies on sparse, entity-centric geometric queries or post-hoc refinement, yielding limited situational awareness and fragile performance in interactive scenes. To address this issue, we propose G2DP (Grid-Guided Diffusion Planning), a diffusion-based planner that directly enforces dense environmental constraints through inference-time guidance. Specifically, G2DP constructs a differentiable spatio-temporal cost volume by fusing probabilistic future occupancy distributions with a route-progress map. By formulating this volume as a continuous safety energy functional, it injects dense gradients directly into the denoising loop, actively steering trajectory generation toward collision-free and progress-optimal regions. Extensive closed-loop evaluations show that G2DP achieves state-of-the-art performance on nuPlan, outperforming the strongest imitation-learning baseline by +7.2 points in reactive score. It further maintains top scores in zero-shot transfers to interPlan and DeepScenario benchmarks, with collision avoidance improving by +10.15 over the unguided approach on interPlan. These results demonstrate that spatio-temporal cost grids serve as an effective representation for robust guidance in diffusion-based planning. Code is available at https://github.com/HangYuu/G2DP.

cs.RO↗

arXiv2026

A specialized reasoning large language model for accelerating rare disease diagnosis: a randomized AI physician assistance trial

Rare diseases affect millions of individuals worldwide, yet timely diagnosis remains a major public health challenge due to scarcity of specialized clinical expertise. While large language models (LLMs) show promise to support rare disease diagnosis, current models are constrained by insufficient clinical deployability, limited clinically grounded evidence, and scarcity of training data. Here we present RaDaR (Rare Disease navigatoR), an open-source, compact reasoning LLM (32B parameters) for rare disease diagnosis. RaDaR was trained with 49,170 publicly available free-text cases and 104,666 synthetic cases with reasoning-enhanced training. RaDaR showed the strongest performance among evaluated open-source models, including the 671B DeepSeek-R1, across public benchmarks and four external validation centers. In a retrospective cohort, RaDaR prioritized the final diagnosis before documented clinical suspicion in 61.06 percent of cases, corresponding to a potential lead time of 1.87 months and 50.18 percent of the within-center interval. In a randomized physician-assistance trial, RaDaR assistance improved physicians' rare-disease diagnostic accuracy by 21.44 percentage points compared with internet search alone. Synthetic-data ablations suggested that phenotype-anchored narratives provide useful training signal for long-tail rare diseases, with a monotonic scaling trend within the tested data range. Together, RaDaR and its development and validation framework provide a deployable rare-disease reasoning model and a reproducible development framework for diagnostic AI under data scarcity.

cs.AI↗

arXiv2026

MemSlides: A Hierarchical Memory Driven Agent Framework for Personalized Slide Generation with Multi-turn Local Revision

Personalized presentation generation requires more than conditioning on a current prompt or template: agents must preserve stable user preferences across tasks, retain newly introduced preferences and constraints during multi-turn revision, and carry out local edits reliably. We propose MemSlides, a hierarchical memory framework for personalized presentation agents that separates long-term memory from working memory and further divides long-term memory into user profile memory and tool memory. User profile memory stores intent-conditioned profiles for round-0 personalization, working memory carries active preferences and session constraints across revision rounds, and tool memory stores reusable execution experience for reliable localized editing. MemSlides pairs this memory design with scoped slide-local revision, so targeted updates act on the smallest affected region instead of repeatedly regenerating the full deck. In controlled experiments, user profile memory improves persona-alignment judgments on a multi-persona, multi-intent profile bank, tool-memory injection improves closed-loop modify behavior in diagnostic matched-pair settings, and qualitative cases illustrate working memory's ability to carryover preferences. Taken together, these results suggest that effective personalization in presentation authoring depends on separating persistent user profiles, session-level working memory, and reusable execution experience across generation and localized revision.

cs.CL↗

arXiv2026

Volume-Refined Achievability and Converse Approximations for Noisy Permutation Channels

We study volume-refined achievability and converse bounds for noisy permutation channels generated by strictly positive DMCs, allowing the reachable output polytope to have arbitrary affine dimension $d\ge 1$. The reachable output polytope may be lower-dimensional than the output simplex, whereas existing refined achievability analyses and fixed-error converses are not adapted to this intrinsic affine geometry. On the achievability side, we develop an affine-coordinate simplex-lattice construction adapted to the reachable output polytope, together with a nearest-neighbor decoder and a geometric error-reduction argument in the same coordinate space. This yields a Gaussian achievability approximation with an $o(1)$ remainder. On the converse side, we first use a meta-converse combined with a KL covering and a local testing estimate to obtain a fixed-error converse with a bounded remainder, which implies the logarithmic $\epsilon$-capacity $d/2$. We then apply the meta-converse with a stratified Jeffreys-mixture auxiliary output distribution. Using a local Laplace approximation and a local likelihood-ratio approximation, this choice identifies the Fisher-volume term and an explicit Gaussian testing constant, yielding a constant-order converse approximation with an $o(1)$ remainder. The achievability and converse constants arise from different constructions and are not claimed to match in general.

cs.IT↗

arXiv2026

Hierarchical Long-Term Semantic Memory for LinkedIn's Hiring Agent

Large Language Model (LLM) agents are increasingly used in real-world products, where personalized and context-aware user interactions are essential. A central enabler of such capabilities is the agent's long-term semantic memory system, which extracts implicit and explicit signals from noisy longitudinal behavioral data, stores them in a structured form, and supports low-latency retrieval. Building industrial-grade long-term memory for LLM agents raises five challenges: scalability, low-latency retrieval, privacy constraints, adaptability, and observability. We introduce the Hierarchical Long-Term Semantic Memory (HLTM) framework, which organizes textual data into a schema-aligned memory tree that captures semantic knowledge at multiple levels of granularity, enabling scalable ingestion, privacy-aware storage, low-latency retrieval, and transparent provenance; HLTM further incorporates an adaptation mechanism to generalize across diverse use cases. Extensive evaluations on LinkedIn's Hiring Assistant show that HLTM improves answer correctness by more than 5% and retrieval F1 by more than 10%, while significantly advancing the Pareto frontier between query and indexing latency. HLTM has been fully deployed in LinkedIn's Hiring Assistant to power core personalization features in production hiring workflows.

cs.IR↗

arXiv2026

Co-Director: Agentic Generative Video Storytelling

While diffusion models generate high-fidelity video clips, transforming them into coherent storytelling engines remains challenging. Current agentic pipelines automate this via chained modules but suffer from semantic drift and cascading failures due to independent, handcrafted prompting. We present Co-Director, a hierarchical multi-agent framework formalizing video storytelling as a global optimization problem. To ensure semantic coherence, we introduce hierarchical parameterization: a multi-armed bandit globally identifies promising creative directions, while a local multimodal self-refinement loop mitigates identity drift and ensures sequence-level consistency. This balances the exploration of novel narrative strategies with the exploitation of effective creative configurations. For evaluation, we introduce GenAD-Bench, a 400-scenario dataset of fictional products for personalized advertising. Experiments demonstrate that Co-Director significantly outperforms state-of-the-art baselines, offering a principled approach that seamlessly generalizes to broader cinematic narratives. Project Page: https://co-director-agent.github.io/

cs.AI↗

arXiv2026

Snapping Actuators with Asymmetric and Sequenced Motion

Snapping instabilities in soft structures offer a powerful pathway to achieve rapid and energy-efficient actuation. In this study, an eccentric dome-shaped snapping actuator is developed to generate controllable asymmetric motion through geometry-induced instability. Finite element simulations and experiments reveal consistent asymmetric deformation and the corresponding pressure characteristics. By coupling four snapping actuators in a pneumatic network, a compact quadrupedal robot achieves coordinated wavelike locomotion using only a single pressure input. The robot exhibits frequency-dependent performance with a maximum speed of 72.78~mm/s at 7.5~Hz. These findings demonstrate the potential of asymmetric snapping mechanisms for physically controlled actuation and lay the groundwork for fully untethered and efficient soft robotic systems.

cs.RO↗

arXiv2025

Large Speech Model Enabled Semantic Communication

Existing speech semantic communication systems mainly based on Joint Source-Channel Coding (JSCC) architectures have demonstrated impressive performance, but their effectiveness remains limited by model structures specifically designed for particular tasks and datasets. Recent advances indicate that generative large models pre-trained on massive datasets, can achieve outstanding performance arexhibit exceptional performance across diverse downstream tasks with minimal fine-tuning. To exploit the rich semantic knowledge embedded in large models and enable adaptive transmission over lossy channels, we propose a Large Speech Model enabled Semantic Communication (LargeSC) system. Simultaneously achieving adaptive compression and robust transmission over lossy channels remains challenging, requiring trade-offs among compression efficiency, speech quality, and latency. In this work, we employ the Mimi as a speech codec, converting speech into discrete tokens compatible with existing network architectures. We propose an adaptive controller module that enables adaptive transmission and in-band Unequal Error Protection (UEP), dynamically adjusting to both speech content and packet loss probability under bandwidth constraints. Additionally, we employ Low-Rank Adaptation (LoRA) to finetune the Moshi foundation model for generative recovery of lost speech tokens. Simulation results show that the proposed system supports bandwidths ranging from 550 bps to 2.06 kbps, outperforms conventional baselines in speech quality under high packet loss rates and achieves an end-to-end latency of approximately 460 ms, thereby demonstrating its potential for real-time deployment.

cs.SD↗

arXiv2025

Integrated Sensing and Communication: Rate-Distortion Fundamental Limits of State Estimator

The state-dependent memoryless channel (SDMC) is employed to model the integrated sensing and communication (ISAC) system, where the transmitter conveys messages to the receiver while simultaneously estimating the state parameter of interest via the received echo signals. However, the performance of sensing has often been neglected in existing works. To address this gap, we establish the rate-distortion function for sensing performance in the SDMC model, which is defined based on standard information-theoretic principles to ensure clear operational meaning. In addition, we propose a modified Blahut-Arimoto type algorithm for solving the rate-distortion function and provide convergence proofs for the algorithm. We further define the capacity-rate-distortion tradeoff region, which unifies information-theoretic results for communication and sensing within a single optimization framework. Finally, we numerically evaluate the capacity-rate-distortion region and demonstrate the benefit of coding in terms of estimation rate for certain channels.

cs.IT↗

arXiv2024

RareAgents: Autonomous Multi-disciplinary Team for Rare Disease Diagnosis and Treatment

Rare diseases, despite their low individual incidence, collectively impact around 300 million people worldwide due to the vast number of diseases. The involvement of multiple organs and systems, and the shortage of specialized doctors with relevant experience, make diagnosing and treating rare diseases more challenging than common diseases. Recently, agents powered by large language models (LLMs) have demonstrated notable applications across various domains. In the medical field, some agent methods have outperformed direct prompts in question-answering tasks from medical examinations. However, current agent frameworks are not well-adapted to real-world clinical scenarios, especially those involving the complex demands of rare diseases. To bridge this gap, we introduce RareAgents, the first LLM-driven multi-disciplinary team decision-support tool designed specifically for the complex clinical context of rare diseases. RareAgents integrates advanced Multidisciplinary Team (MDT) coordination, memory mechanisms, and medical tools utilization, leveraging Llama-3.1-8B/70B as the base model. Experimental results show that RareAgents outperforms state-of-the-art domain-specific models, GPT-4o, and current agent frameworks in diagnosis and treatment for rare diseases. Furthermore, we contribute a novel rare disease dataset, MIMIC-IV-Ext-Rare, to facilitate further research in this field.

cs.CL↗

arXiv2024

DLM-VMTL:A Double Layer Mapper for heterogeneous data video Multi-task prompt learning

In recent years, the parameters of backbones of Video Understanding tasks continue to increase and even reach billion-level. Whether fine-tuning a specific task on the Video Foundation Model or pre-training the model designed for the specific task, incurs a lot of overhead. How to make these models play other values than their own tasks becomes a worthy question. Multi-Task Learning(MTL) makes the visual task acquire the rich shareable knowledge from other tasks while joint training. It is fully explored in Image Recognition tasks especially dense predict tasks. Nevertheless, it is rarely used in video domain due to the lack of multi-labels video data. In this paper, a heterogenous data video multi-task prompt learning (VMTL) method is proposed to address above problem. It's different from it in image domain, a Double-Layers Mapper(DLM) is proposed to extract the shareable knowledge into visual promptS and align it with representation of primary task. Extensive experiments prove that our DLM-VMTL performs better than baselines on 6 different video understanding tasks and 11 datasets.

cs.CV↗

arXiv2024

Synchronous Multi-modal Semantic Communication System with Packet-level Coding

Although the semantic communication with joint semantic-channel coding design has shown promising performance in transmitting data of different modalities over physical layer channels, the synchronization and packet-level forward error correction of multimodal semantics have not been well studied. Due to the independent design of semantic encoders, synchronizing multimodal features in both the semantic and time domains is a challenging problem. In this paper, we take the facial video and speech transmission as an example and propose a Synchronous Multimodal Semantic Communication System (SyncSC) with Packet-Level Coding. To achieve semantic and time synchronization, 3D Morphable Mode (3DMM) coefficients and text are transmitted as semantics, and we propose a semantic codec that achieves similar quality of reconstruction and synchronization with lower bandwidth, compared to traditional methods. To protect semantic packets under the erasure channel, we propose a packet-Level Forward Error Correction (FEC) method, called PacSC, that maintains a certain visual quality performance even at high packet loss rates. Particularly, for text packets, a text packet loss concealment module, called TextPC, based on Bidirectional Encoder Representations from Transformers (BERT) is proposed, which significantly improves the performance of traditional FEC methods. The simulation results show that our proposed SyncSC reduce transmission overhead and achieve high-quality synchronous transmission of video and speech over the packet loss network.

eess.IV↗

arXiv2023

What Makes a Fantastic Passenger-Car Driver in Urban Contexts?

The accurate evaluation of the quality of driving behavior is crucial for optimizing and implementing autonomous driving technology in practice. However, there is no comprehensive understanding of good driving behaviors currently. In this paper, we sought to understand driving behaviors from the perspectives of both drivers and passengers. We invited 10 expert drivers and 14 novice drivers to complete a 5.7-kilometer urban road driving task. After the experiments, we conducted semi-structured interviews with 24 drivers and 48 of their passengers (two passengers per driver). Through the analysis of interview data, we found passengers' assessing logic of driving behaviors, divers' considerations and efforts to achieve good driving, and gaps between these perspectives. Our research provided insights into a systematic evaluation of autonomous driving and the design implications for future autonomous vehicles.

cs.HC↗

arXiv2023

SurrealDriver: Designing LLM-powered Generative Driver Agent Framework based on Human Drivers' Driving-thinking Data

Leveraging advanced reasoning capabilities and extensive world knowledge of large language models (LLMs) to construct generative agents for solving complex real-world problems is a major trend. However, LLMs inherently lack embodiment as humans, resulting in suboptimal performance in many embodied decision-making tasks. In this paper, we introduce a framework for building human-like generative driving agents using post-driving self-report driving-thinking data from human drivers as both demonstration and feedback. To capture high-quality, natural language data from drivers, we conducted urban driving experiments, recording drivers' verbalized thoughts under various conditions to serve as chain-of-thought prompts and demonstration examples for the LLM-Agent. The framework's effectiveness was evaluated through simulations and human assessments. Results indicate that incorporating expert demonstration data significantly reduced collision rates by 81.04\% and increased human likeness by 50\% compared to a baseline LLM-based agent. Our study provides insights into using natural language-based human demonstration data for embodied tasks. The driving-thinking dataset is available at \url{https://github.com/AIR-DISCOVER/Driving-Thinking-Dataset}.

cs.HC↗

arXiv2022

Combining Localized Orbital Scaling Correction and Bethe-Salpeter Equation for Accurate Excitation Energies

We applied localized orbital scaling correction (LOSC) in Bethe-Salpeter equation (BSE) to predict accurate excitation energies for molecules. LOSC systematically eliminates the delocalization error in the density functional approximation and is capable of approximating quasiparticle (QP) energies with accuracy similar or better than the $GW$ Green's function approach and with much less computational cost. The QP energies from LOSC instead of commonly used $G_{0}W_{0}$ and ev$GW$ are directly used in BSE. We show that the BSE/LOSC approach greatly outperforms the commonly used BSE/$G_{0}W_{0}$ approach for predicting excitations with different characters. For the calculations for Truhlar-Gagliardi test set containing valence, charge transfer (CT) and Rydberg excitations, BSE/LOSC with the Tamm-Dancoff approximation provides a comparable accuracy to time-dependent density functional theory (TDDFT) and BSE/ev$GW$. For the calculations of Stein CT test set and Rydberg excitations of atoms, BSE/LOSC considerably outperforms both BSE/$G_{0}W_{0}$ and TDDFT approaches with a reduced starting point dependence. BSE/LOSC is thus a promising and efficient approach to calculate excitation energies for molecular systems.

physics.chem-ph↗

arXiv2022

Benchmark of $\boldsymbol{GW}$ Methods for Core-Level Binding Energies

The $GW$ approximation has recently gained increasing attention as a viable method for the computation of deep core-level binding energies as measured by X-ray photoelectron spectroscopy (XPS). We present a comprehensive benchmark study of different $GW$ methodologies (starting-point optimized, partial and full eigenvalue-self-consistent, Hedin shift and renormalized singles) for molecular inner-shell excitations. We demonstrate that all methods yield a unique solution and apply them to the CORE65 benchmark set and ethyl trifluoroacetate. Three $GW$ schemes clearly outperform the other methods for absolute core-level energies with a mean absolute error of 0.3 eV with respect to experiment. These are partial eigenvalue self-consistency, in which the eigenvalues are only updated in the Green's function, single-shot $GW$ calculations based on an optimized hybrid functional starting point and a Hedin shift in the Green's function. While all methods reproduce the experimental relative binding energies well, the eigenvalue self-consistent schemes and the Hedin shift yield with mean errors $<0.2$ eV the best results.

physics.chem-ph↗

arXiv2019

Query Expansion for Patent Searching using Word Embedding and Professional Crowdsourcing

The patent examination process includes a search of previous work to verify that a patent application describes a novel invention. Patent examiners primarily use keyword-based searches to uncover prior art. A critical part of keyword searching is query expansion, which is the process of including alternate terms such as synonyms and other related words, since the same concepts are often described differently in the literature. Patent terminology is often domain specific. By curating technology-specific corpora and training word embedding models based on these corpora, we are able to automatically identify the most relevant expansions of a given word or phrase. We compare the performance of several automated query expansion techniques against expert specified expansions. Furthermore, we explore a novel mechanism to extract related terms not just based on one input term but several terms in conjunction by computing their centroid and identifying the nearest neighbors to this centroid. Highly skilled patent examiners are often the best and most reliable source of identifying related terms. By designing a user interface that allows examiners to interact with the word embedding suggestions, we are able to use these interactions to power crowdsourced modes of related terms. Learning from users allows us to overcome several challenges such as identifying words that are bleeding edge and have not been published in the corpus yet. This paper studies the effectiveness of word embedding and crowdsourced models across 11 disparate technical areas.

cs.IR↗