arXiv subjects

Jing Zheng

Publications and source records attributed to Jing Zheng.

16 recordsLinked to original sources

arXiv2026

Localization Boosting for Growth Markets: Mitigating Cross-Locale Behavioral Bias in Learning-to-Rank

Adobe Express is expanding internationally, but the US has a disproportionately large content supply and interaction volume. Learning-to-rank (LTR) models trained primarily on behavioral feedback inherit this imbalance: templates popular in US are over-served in non-US locales. This cross-locale exposure bias suppresses local content discoverability and degrades ranking quality in growth locales. We show that click-only training suppresses semantically informative localization features. Adding vision-language model (VLM) graded relevance labels as auxiliary supervision alongside clicks improves semantic alignment but does not preserve local content visibility. We propose a multi-objective framework combining behavioral supervision, VLM-derived relevance signals, and locale-aware boosting. Across five locales, the resulting model improves relevance while restoring stable localization, demonstrating the importance of disentangling exposure from semantic supervision.

cs.LG↗

arXiv2026

Resource-efficient parallel entanglement generation for multinode quantum networks via time-bin multiplexing

Nonlocal entanglement generation among multiple remote quantum nodes provides a critical foundation for a variety of counterintuitive quantum applications. The exponential loss of photons transmitting over optical fibers sets an upper limit for entangling these quantum nodes. Here, we propose a resource-efficient and parallel protocol for entangling multiple remote quantum nodes via time-bin multiplexing. The transmission of a single photon with qudit-encoding in the time-bin mode enables entangling multiple stationary qubits in parallel, when single photons and individual stationary qubits interfaces are used and photon-state modulations are properly introduced before subsequently impinging the photon into each interface. Our protocol can generate parallel multipartite entanglement among ($N\geq3$) quantum nodes with the dimension of the photonic time bins independent of $N$, exponentially reducing the requirements for the coherence time of the stationary qubits and for the complexity of the photonic modulations. These distinct features make our protocol particularly advantageous for the development of multinode quantum networks.

quant-ph↗

arXiv2026

Equipping LLM with Directional Multi-Talker Speech Understanding Capabilities

Recent studies have demonstrated that prompting large language models (LLM) with audio encodings enables effective speech understanding capabilities. However, most speech LLMs are trained on single-channel, single-talker data, which makes it challenging to directly apply them to multi-talker and multi-channel speech understanding task. In this work, we present a comprehensive investigation on how to enable directional multi-talker speech understanding capabilities for LLMs, specifically in smart glasses usecase. We propose two novel approaches to integrate directivity into LLMs: (1) a cascaded system that leverages a source separation front-end module, and (2) an end-to-end system that utilizes serialized output training. All of the approaches utilize a multi-microphone array embedded in smart glasses to optimize directivity interpretation and processing in a streaming manner. Experimental results demonstrate the efficacy of our proposed methods in endowing LLMs with directional speech understanding capabilities, achieving strong performance in both speech recognition and speech translation tasks.

cs.CL↗

arXiv2025

WearVox: An Egocentric Multichannel Voice Assistant Benchmark for Wearables

Wearable devices such as AI glasses are transforming voice assistants into always-available, hands-free collaborators that integrate seamlessly with daily life, but they also introduce challenges like egocentric audio affected by motion and noise, rapid micro-interactions, and the need to distinguish device-directed speech from background conversations. Existing benchmarks largely overlook these complexities, focusing instead on clean or generic conversational audio. To bridge this gap, we present WearVox, the first benchmark designed to rigorously evaluate voice assistants in realistic wearable scenarios. WearVox comprises 3,842 multi-channel, egocentric audio recordings collected via AI glasses across five diverse tasks including Search-Grounded QA, Closed-Book QA, Side-Talk Rejection, Tool Calling, and Speech Translation, spanning a wide range of indoor and outdoor environments and acoustic conditions. Each recording is accompanied by rich metadata, enabling nuanced analysis of model performance under real-world constraints. We benchmark leading proprietary and open-source speech Large Language Models (SLLMs) and find that most real-time SLLMs achieve accuracies on WearVox ranging from 29% to 59%, with substantial performance degradation on noisy outdoor audio, underscoring the difficulty and realism of the benchmark. Additionally, we conduct a case study with two new SLLMs that perform inference with single-channel and multi-channel audio, demonstrating that multi-channel audio inputs significantly enhance model robustness to environmental noise and improve discrimination between device-directed and background speech. Our results highlight the critical importance of spatial audio cues for context-aware voice assistants and establish WearVox as a comprehensive testbed for advancing wearable voice AI research.

cs.CL↗

arXiv2025

Scaling Multimodal Search and Recommendation with Small Language Models via Upside-Down Reinforcement Learning

In this work, we investigate how small language models (SLMs) can be scaled to support multimodal search and recommendation use cases while remaining efficient enough for real-time, resource-constrained deployments. We present a framework that combines upside-down reinforcement learning with synthetic data distillation from a large language model (Llama-3) to train a 100M-parameter GPT-2 model for multitask prompt generation. Despite being up to 80 times smaller than state-of-the-art large language models (LLMs), our SLM achieves relevance and diversity scores within 6% of competitive baselines such as Llama-3 8B, Qwen3 8B, and Ministral 8B. These results demonstrate that SLMs can effectively handle multimodal search and recommendation tasks, while dramatically reducing inference latency and memory overhead. Our study highlights the potential of lightweight models as practical engines for scalable multimodal discovery, bridging the gap between cutting-edge research and real-world multimodal applications such as media recommendations and creative content generation.

cs.CL↗

arXiv2024

Reverse Chain: A Generic-Rule for LLMs to Master Multi-API Planning

While enabling large language models to implement function calling (known as APIs) can greatly enhance the performance of Large Language Models (LLMs), function calling is still a challenging task due to the complicated relations between different APIs, especially in a context-learning setting without fine-tuning. This paper introduces ``Reverse Chain'', a controllable, target-driven approach designed to empower LLMs with the capability to operate external APIs only via prompts. Recognizing that most LLMs have limited tool-use capabilities, Reverse Chain limits LLMs to executing simple tasks, e.g., API Selection and Argument Completion. Furthermore, to manage a controllable multi-function calling, Reverse Chain adopts a generic rule based on a backward reasoning process. This rule determines when to do API selection or Argument completion. To evaluate the multi-tool-use capability of LLMs, we have released a compositional multi-tool task dataset, available at \url{https://anonymous.4open.science/r/reverse-chain-8681}. Extensive numerical experiments validate the remarkable proficiency of Reverse Chain in managing multiple API calls.

cs.SE↗

arXiv2024

From Beginner to Expert: Modeling Medical Knowledge into General LLMs

Recently, large language model (LLM) based artificial intelligence (AI) systems have demonstrated remarkable capabilities in natural language understanding and generation. However, these models face a significant challenge when it comes to sensitive applications, such as reasoning over medical knowledge and answering medical questions in a physician-like manner. Prior studies attempted to overcome this challenge by increasing the model size (>100B) to learn more general medical knowledge, while there is still room for improvement in LLMs with smaller-scale model sizes (<100B). In this work, we start from a pre-trained general LLM model (AntGLM-10B) and fine-tune it from a medical beginner towards a medical expert (called AntGLM-Med-10B), which leverages a 3-stage optimization procedure, i.e., general medical knowledge injection, medical domain instruction tuning, and specific medical task adaptation. Our contributions are threefold: (1) We specifically investigate how to adapt a pre-trained general LLM in medical domain, especially for a specific medical task. (2) We collect and construct large-scale medical datasets for each stage of the optimization process. These datasets encompass various data types and tasks, such as question-answering, medical reasoning, multi-choice questions, and medical conversations. (3) Specifically for multi-choice questions in the medical domain, we propose a novel Verification-of-Choice approach for prompting engineering, which significantly enhances the reasoning ability of LLMs. Remarkably, by combining the above approaches, our AntGLM-Med-10B model can outperform the most of LLMs on PubMedQA, including both general and medical LLMs, even when these LLMs have larger model size.

cs.CL↗

arXiv2023

AdapterDistillation: Non-Destructive Task Composition with Knowledge Distillation

Leveraging knowledge from multiple tasks through introducing a small number of task specific parameters into each transformer layer, also known as adapters, receives much attention recently. However, adding an extra fusion layer to implement knowledge composition not only increases the inference time but also is non-scalable for some applications. To avoid these issues, we propose a two-stage knowledge distillation algorithm called AdapterDistillation. In the first stage, we extract task specific knowledge by using local data to train a student adapter. In the second stage, we distill the knowledge from the existing teacher adapters into the student adapter to help its inference. Extensive experiments on frequently asked question retrieval in task-oriented dialog systems validate the efficiency of AdapterDistillation. We show that AdapterDistillation outperforms existing algorithms in terms of accuracy, resource consumption and inference time.

cs.LG↗

arXiv2023

Contextual Font Recommendations based on User Intent

Adobe Fonts has a rich library of over 20,000 unique fonts that Adobe users utilize for creating graphics, posters, composites etc. Due to the nature of the large library, knowing what font to select can be a daunting task that requires a lot of experience. For most users in Adobe products, especially casual users of Adobe Express, this often means choosing the default font instead of utilizing the rich and diverse fonts available. In this work, we create an intent-driven system to provide contextual font recommendations to users to aid in their creative journey. Our system takes in multilingual text input and recommends suitable fonts based on the user's intent. Based on user entitlements, the mix of free and paid fonts is adjusted. The feature is currently used by millions of Adobe Express users with a CTR of >25%.

cs.HC↗

arXiv2022

Activity-assisted barrier-crossing of self-propelled colloids over parallel microgrooves

We report a systematic study of the dynamics of self-propelled particles (SPPs) over a one-dimensional periodic potential landscape, which is fabricated on a microgroove-patterned polydimethylsiloxane (PDMS) substrate. From the measured non-equilibrium probability density function of the SPPs, we find that the escape dynamics of the slow-rotating SPPs across the potential landscape can be described by an effective potential, once the self-propulsion force is included into the potential under the fixed angle approximation. This work demonstrates that the parallel microgrooves provide a versatile platform for a quantitative understanding of the interplay among the self-propulsion force, spatial confinement by the potential landscape, and thermal noise, as well as its effects on activity-assisted escape dynamics and transport of the SPPs.

cond-mat.soft↗

arXiv2022

R2D2: Recursive Transformer based on Differentiable Tree for Interpretable Hierarchical Language Modeling

Human language understanding operates at multiple levels of granularity (e.g., words, phrases, and sentences) with increasing levels of abstraction that can be hierarchically combined. However, existing deep models with stacked layers do not explicitly model any sort of hierarchical process. This paper proposes a recursive Transformer model based on differentiable CKY style binary trees to emulate the composition process. We extend the bidirectional language model pre-training objective to this architecture, attempting to predict each word given its left and right abstraction nodes. To scale up our approach, we also introduce an efficient pruned tree induction algorithm to enable encoding in just a linear number of composition steps. Experimental results on language modeling and unsupervised parsing show the effectiveness of our approach.

cs.CL↗

arXiv2022

Model-independent reconstruction of the cosmological scale factor as a function of lookback time

We present a model-independent method of reconstructing scale factor against lookback time from the Observational Hubble parameter Data (OHD). The reconstruction method is independent of dynamical models and is only based on the Friedmann-Robertson-Walker metric. We also calculate the propagation of error in the reconstruction process. The reconstruction data errors mainly come from trapezoidal rule approximation and the uncertainty from OHD. Furthermore, the model discrimination ability of original OHD and reconstructed a-t data is discussed under a dimensionless standard method. a-t data can present the differences between cosmology models more clearly than H-z data by comparing their coefficients of variations. Finally, we add fifty simulated H(z) data to estimate the influence of future observation. More Hubble measurements in the future will help constrain cosmological parameters more accurately.

astro-ph.CO↗

arXiv2021

Reranking Machine Translation Hypotheses with Structured and Web-based Language Models

In this paper, we investigate the use of linguistically motivated and computationally efficient structured language models for reranking N-best hypotheses in a statistical machine translation system. These language models, developed from Constraint Dependency Grammar parses, tightly integrate knowledge of words, morphological and lexical features, and syntactic dependency constraints. Two structured language models are applied for N-best rescoring, one is an almost-parsing language model, and the other utilizes more syntactic features by explicitly modeling syntactic dependencies between words. We also investigate effective and efficient language modeling methods to use N-grams extracted from up to 1 teraword of web documents. We apply all these language models for N-best re-ranking on the NIST and DARPA GALE program 2006 and 2007 machine translation evaluation tasks and find that the combination of these language models increases the BLEU score up to 1.6% absolutely on blind test sets.

cs.CL↗

arXiv2021

Graph-based Pyramid Global Context Reasoning with a Saliency-aware Projection for COVID-19 Lung Infections Segmentation

Coronavirus Disease 2019 (COVID-19) has rapidly spread in 2020, emerging a mass of studies for lung infection segmentation from CT images. Though many methods have been proposed for this issue, it is a challenging task because of infections of various size appearing in different lobe zones. To tackle these issues, we propose a Graph-based Pyramid Global Context Reasoning (Graph-PGCR) module, which is capable of modeling long-range dependencies among disjoint infections as well as adapt size variation. We first incorporate graph convolution to exploit long-term contextual information from multiple lobe zones. Different from previous average pooling or maximum object probability, we propose a saliency-aware projection mechanism to pick up infection-related pixels as a set of graph nodes. After graph reasoning, the relation-aware features are reversed back to the original coordinate space for the down-stream tasks. We further construct multiple graphs with different sampling rates to handle the size variation problem. To this end, distinct multi-scale long-range contextual patterns can be captured. Our Graph-PGCR module is plug-and-play, which can be integrated into any architecture to improve its performance. Experiments demonstrated that the proposed method consistently boost the performance of state-of-the-art backbone architectures on both of public and our private COVID-19 datasets.

eess.IV↗

arXiv2019

A Model-Independent Measurement of the Spatial Curvature using Cosmic Chronometers and the HII Hubble Diagram

We propose a model-independent way to determine the cosmic curvature using the Hubble parameter $H(z)$ measured with cosmic chronometers and the comoving distance $D(z)$ inferred from HII galaxies. We employ Gaussian processes to smooth the measure of distance and match it to $30$ values of $H(z)$. The curvature parameter $Ω_k$ may be obtained individually for each such pair. The weighted average for the complete sample is $Ω_k=-0.0013\pm0.0004$, suggesting a bias towards negative values. The accuracy of the curvature measurement improves with increased redshift, however, given possible systematic effects associated with local inhomogeneities. We therefore also analyze a high-redshift ($z>1.5$) sub-sample on its own, which is more likely to reflect the geometry of the Universe on large, smooth scales. We find for this set of data that $Ω_k=-0.0111\pm0.0416$, consistent with zero to better than $1σ$. This result is in agreement with the spatially flat universe inferred from the cosmic microwave background observations. We expect this method to yield even tighter constraints on the curvature parameter with future, more accurate observations of HII galaxies at high $z$.

astro-ph.CO↗

arXiv2010

The local properties of the Markov processes of Ornstein-Uhlenbeck type

We prove the existence of a local time, the continuity of the local time about $t$, and the regular property for $a.e.$ $x\in R$ of a Ornstein-Uhlenbeck type $\{X_t,\ t\in R^+\}$ driven by a general Lévy process, under mild regularity conditions. We discuss the asymptotic behaviour of the local time when $X$ is ergodic. We also investigate the first passage problem. These results give precise information about the local properties of the sample functions.

math.PR↗