Searcharxiv⌕ Search

arXiv subjects

Liangjie Hong

Publications and source records attributed to Liangjie Hong.

At least 19 recordsLinked to original sources
arXiv2026

Self-Evolving Code-with-Image Reasoning

Multimodal models increasingly reach for tools when solving visual tasks (crop, zoom, rotate, brighten), a paradigm known as thinking-with-images. The central challenge is one of perception: tools mostly serve to expose visual evidence, reasoning over that evidence stays in language, and most targets are ones a human could in principle determine by inspection. Some visual questions, however, are not bottlenecked by perception: recovering their answers requires executing a multi-step visual algorithm over the pixels. On such questions a model often names the correct algorithm at once yet still answers wrong, because language can describe an algorithm without being able to run one. Code-with-Image crosses that line: given nothing but a Python interpreter, the model must implement a genuine visual algorithm in code to solve the task; the program itself becomes the reasoning. The bottleneck then shifts from executing code to deciding which algorithm to implement. So we let the model teach itself: a training-free reflection loop studies its own failed programs, tests repairs against constructive ground truth, and keeps what survives as portable skills. On our Code-with-Image Bench (CwI-Bench), thirty task families induced by hidden visual computations with disjoint learning and evaluation splits, even GPT-5.6-luna stays below 30% with tool-free chain of thought; given a bare interpreter it reaches 43%, and with skills evolved through its own executable reflection, 67%. The open 27B model climbs the same ladder (9% $\rightarrow$ 33% $\rightarrow$ 56%), and the skills are plain text, transferable across scales and families. When code carries the reasoning, debugging code becomes debugging reasoning.

cs.CV↗
arXiv2026

SPECTRA: Pushing the KV Cache Beyond the 2-Bit Cliff via Spectral Transform Coding

Large language models (LLMs) increasingly read long inputs in the agentic era, from whole documents and codebases to conversations across many turns. Their inference memory is then dominated by the key-value (KV) cache, the stored attention keys and values of every token the model has read and generated. Because the cache grows with context length and is re-read in full at every generated token, a longer context means more GPU memory. To reduce this cost, most existing methods compress the KV cache by lowering every stored value to the same low precision, a technique known as quantization. They can push this to nearly two bits per value, but rarely further, because quality drops sharply at this 2-bit cliff: four levels are too few for the cache's outlier-heavy values, where a few large entries consume the levels and collapse the rest into noise. A natural remedy is to spend more bits on the channels (feature dimensions) that matter and fewer on the rest, but the raw cache offers no handle: its channels are strongly correlated, so none stands out as more important. Our analysis shows that this handle appears once the cache is rotated into a coordinate system computed from its own statistics, removing these correlations. There, a small fraction of channels carries almost all the information, and spending the budget on those few is far more accurate than spreading it evenly. Guided by this analysis, we develop SPECTRA, a training-free, drop-in codec that re-encodes the cache into this coordinate system and concentrates the bit budget on the channels that carry the signal. On Llama-3.1-8B and Qwen2.5-7B over long-context benchmarks, SPECTRA is near-lossless at 4x compression, competitive at 8x where uniform quantization has collapsed, and reaches up to 12x, pushing usable compression past the 2-bit cliff so the same GPU holds longer contexts and larger batches.

cs.LG↗
arXiv2026

Toward Reliable Context Compression for Long-Horizon Agents: An Empirical Study of Execution Instability

Recurrent context compression controls context growth in long-horizon agents, but its behavioral effects remain poorly understood. In this preliminary empirical study, we show that compression can weaken the influence of recent interactions, increasing blocked actions, repeated exploration, and instability across runs. Motivated by these observations, we introduce TRACE, a verifier-guided framework that evaluates individual compaction events through paired closed-loop continuations from the same environment state and uses summary preferences to optimize a natural-language compression prompt while keeping all models frozen. Initial results on AppWorld show improvements over existing compression baselines in task performance, multi-run reliability, and context--execution efficiency. These findings provide early evidence for boundary-local evaluation as a promising direction for reliable agent context compression.

cs.LG↗
arXiv2026

Field-Aware Agent Skill Retrieval

As lifelong learning agents accumulate lifelong growing skill banks, retrieving the correct skill becomes an increasingly important bottleneck. Most current skill retrieval methods treat each skill as one flat document by concatenating fields such as the name, description, and body. However, skills are naturally structured, multi-field objects, where each field provides different information about when and how the skill should be used. In this work, we study whether preserving this structure improves skill retrieval. We represent each skill as its separate components, and compute sparse and dense similarities for each field independently, exposing a naturally tensorized, field-aware representation of the skill bank. We then combine these field-level scores either with uniform weights or with a small learned MLP. Across two different skill retrieval benchmarks, SkillRet and SRA-Bench, we find that keeping fields separate improves hybrid retrieval, and learning over the field-level scores gives the strongest and most consistent results. Our field-aware MLP reaches $77.95$ Recall@10 on SkillRet and $83.78$ Recall@10 on SRA-Bench, outperforming the corresponding concatenated learned baselines. We also find that the advantage grows as the skill bank becomes larger, suggesting that field-aware skill retrieval becomes especially useful in the setting where retrieval is most difficult. Our results show that skill representation itself matters, and that simply preserving the structure already present in skill files can substantially improve retrieval.

cs.IR↗
arXiv2026

Unified Semantic Modeling Framework for Large-Scale Job Understanding at LinkedIn

Job understanding is critical to LinkedIn's mission of connecting talent with opportunity. This task involves transforming unstructured and noisy job postings into standardized or derived job attributes that power numerous LinkedIn products. However, building a scalable, cost-efficient, and high-performing job understanding system remains challenging. In this paper, we present a unified semantic modeling framework powered by a small language model (SLM) to address the challenges. We begin by fine-tuning an open-source SLM using a suite of carefully curated synthetic tasks augmented with reasoning traces. These tasks jointly target taxonomy-guided classification and taxonomy-agnostic entity extraction. This allows the resulting model to acquire robust zero-shot generalization for job understanding in structured and unstructured contexts. Building upon this foundation, we introduce a multi-adapter architecture with attribute grouping to facilitate efficient task-specific adaptation while streamlining model management across diverse downstream attributes. Offline evaluations and online A/B tests demonstrate significant performance improvement while reducing operational complexity. Our work provides practical insights into building industry-scale text understanding systems.

cs.AI↗
arXiv2026

WiSP: A Working-Set View of Mixture-of-Experts Serving on Extremely Low-Resource Hardware

Modern local and agentic workloads often need large-model capacity at low concurrency, but run on GPUs that cannot keep a frontier-scale model resident. Mixture-of-Experts (MoE) models are a natural fit because they activate only a small subset of experts per token, but their sparsity saves computation, not residency: the full expert pool still has to be stored, and any expert used by a layer must be in GPU memory when that layer runs. Static layer-level CPU offload makes such models fit, but transfers the expert layer in bulk on every forward pass, losing much of the sparsity advantage. We view low-resource MoE serving as a working-set problem on the GPU. Routed expert weights and the KV cache are two memory-demand streams competing for the same limited VRAM. We implement this view in WiSP (Working-Set Paging), a routing-aware expert pager that plugs into an unmodified serving engine and preserves byte-identical outputs. On a real 24 GiB RTX 3090, WiSP achieves up to 2.0x the decode throughput of static offload at the same memory budget when the model does not fit. A natural next step is to predict future experts and prefetch them. We find that this does not help in single-stream decode: the bottleneck is PCIe bandwidth, not prediction quality, so speculative transfers compete with demand transfers instead of hiding them. This shifts the design question from prefetching to allocation: how should one VRAM budget be divided between resident experts and the KV cache? We answer with MV-WSA (Marginal-Value Working-Set Allocation), which splits memory by marginal latency benefit per byte while enforcing a KV-admission floor. As a startup configurator, MV-WSA is the only policy we test that stays near-best on both prefill and decode; as a live controller, it resizes both pools while serving and reduces end-to-end time by up to 1.19x over a fixed offline split, without changing model outputs.

cs.LG↗
arXiv2026

A Unified Structured Query Understanding Framework for Industrial Semantic Search

Query understanding in large-scale industrial search systems is typically implemented as a cascade of disparate, task-specific components. While individually optimizable, this fragmented architecture incurs high maintenance overhead and results in inconsistent behaviors, particularly for long-tail queries. In this work, we propose and deploy a unified structured query understanding system that consolidates these heterogeneous functions into a single Small Language Model (SLM) that performs schema-constrained generation. To address the data bottlenecks inherent in unified modeling, we introduce Query Illuminator, a dual-purpose framework serving as: (i) a teacher model for high-quality auto-annotation and distillation, and (ii) a surrogate judge for scalable evaluation where human labels are scarce. We validate this approach through extensive offline and online tests within LinkedIn's Job Search system. Furthermore, we demonstrate the framework's horizontal extensibility through a cross-domain case study on People Search. The results show improved user engagement and reduced operational costs, achieved while satisfying strict low-latency serving constraints on limited GPU resources.

cs.IR↗
arXiv2026

Towards Resilient and Autonomous Networks: A BlueSky Vision on AI-Native 6G

The proliferation of emerging applications, such as autonomous driving and immersive experiences, demands cellular networks that are not only faster, but fundamentally more resilient and autonomous. This paper presents a BlueSky vision on how Artificial Intelligence will be natively integrated into 6G, shifting the paradigm from \underline{Network for AI} to \underline{AI for Network}. We envision that, unlike 5G's reliance on scattered, ad-hoc models each trained for a single task, native AI in the 6G era will be anchored by a foundation model and and orchestrated via collaborative multi-agent systems, framing network management as a unified, multi-modal, multi-task optimization problem. Built on this vision, we outline two transformative directions. The first focuses on developing a 6G foundation model as a unified backbone, with task-specific knowledge distilled into compact models suited for diverse edge deployments. The second advances multi-agent systems designed to autonomously diagnose, maintain, and recover networks with minimal human intervention. These directions chart a roadmap for 6G to evolve into an intelligent, self-sustaining communication infrastructure.

cs.AI↗
arXiv2026

SAPO: Step-Aligned Policy Optimization for Reasoning-Based Generative Recommendation

Generative recommendation treats next-item prediction as autoregressive item-identifier generation. Specifically, items are encoded as semantic identifiers (SIDs), which are short coarse-to-fine token sequences whose early tokens capture broad semantics and later tokens refine them. Recent work augments this paradigm with reasoning traces and optimizes them via reinforcement learning with verifiable rewards, typically outcome-reward algorithm with exact-match feedback on the generated SID. However, in large-catalog recommendation, exact-match feedback on the generated SID only reports whether the final item is correct; when a generated SID mismatches, outcome-reward cannot identify which SID-token prediction caused the mismatch and may penalize matched SID-token positions together with the mismatched position. We identify that the natural unit of credit assignment in this setting is a single reasoning step (one thinking block paired with one SID token). We instantiate this idea in SAPO (Step-Aligned Policy Optimization): rather than broadcasting one advantage to the whole response, SAPO computes a separate group-relative advantage for each reasoning step and applies it only to the corresponding thinking block and SID token. Across three real-world recommendation datasets, SAPO stabilizes reinforcement-learning training and consistently improves over existing generative recommendation baselines, with the largest gains where sparse exact-match feedback makes reasoning-step credit assignment important. Our results suggest that reinforcement-learning objectives for structured generation should mirror the decoder's own decomposition of the output.

cs.AI↗
arXiv2026

SemCoT: Accelerating Chain-of-Thought Reasoning through Semantically-Aligned Implicit Tokens

The verbosity of Chain-of-Thought (CoT) reasoning hinders its mass deployment in efficiency-critical applications. Recently, implicit CoT approaches have emerged, which encode reasoning steps within LLM's hidden embeddings (termed ``implicit reasoning'') rather than explicit tokens. This approach accelerates CoT by reducing the reasoning length and bypassing some LLM components. However, existing implicit CoT methods face two significant challenges: (1) they fail to preserve the semantic alignment between the implicit reasoning (when transformed to natural language) and the ground-truth reasoning, resulting in a significant CoT performance degradation, and (2) they focus on reducing the length of the implicit reasoning; however, they neglect the considerable time cost for an LLM to generate one individual implicit reasoning token. To tackle these challenges, we propose a novel semantically-aligned implicit CoT framework termed SemCoT. In particular, for the first challenge, we design a contrastively trained sentence transformer that evaluates semantic alignment between implicit and explicit reasoning, which is used to enforce semantic preservation during implicit reasoning optimization. To address the second challenge, we introduce an efficient implicit reasoning generator by finetuning a lightweight language model using knowledge distillation. This generator is guided by our sentence transformer to distill ground-truth reasoning into semantically aligned implicit reasoning, while also optimizing for accuracy. SemCoT is the first approach that enhances CoT efficiency by jointly optimizing token-level generation speed and preserving semantic alignment with ground-truth reasoning. Extensive experiments demonstrate the superior performance of SemCoT compared to state-of-the-art methods in both efficiency and effectiveness. Our code can be found at https://github.com/YinhanHe123/SemCoT/.

cs.CL↗
arXiv2025

LANTERN: Scalable Distillation of Large Language Models for Job-Person Fit and Explanation

Large language models (LLMs) have achieved strong performance across a wide range of natural language processing tasks. However, deploying LLMs at scale for domain specific applications, such as job-person fit and explanation in job seeking platforms, introduces distinct challenges. At LinkedIn, the job person fit task requires analyzing a candidate's public profile against job requirements to produce both a fit assessment and a detailed explanation. Directly applying open source or finetuned LLMs to this task often fails to yield high quality, actionable feedback due to the complexity of the domain and the need for structured outputs. Moreover, the large size of these models leads to high inference latency and limits scalability, making them unsuitable for online use. To address these challenges, we introduce LANTERN, a novel LLM knowledge distillation framework tailored specifically for job person fit tasks. LANTERN involves modeling over multiple objectives, an encoder model for classification purpose, and a decoder model for explanation purpose. To better distill the knowledge from a strong black box teacher model to multiple downstream models, LANTERN incorporates multi level knowledge distillation that integrates both data and logit level insights. In addition to introducing the knowledge distillation framework, we share our insights on post training techniques and prompt engineering, both of which are crucial for successfully adapting LLMs to domain specific downstream tasks. Extensive experimental results demonstrate that LANTERN significantly improves task specific metrics for both job person fit and explanation. Online evaluations further confirm its effectiveness, showing measurable gains in job seeker engagement, including a 0.24\% increase in apply rate and a 0.28\% increase in qualified applications.

cs.CL↗
arXiv2025

Powering Job Search at Scale: LLM-Enhanced Query Understanding in Job Matching Systems

Query understanding is essential in modern relevance systems, where user queries are often short, ambiguous, and highly context-dependent. Traditional approaches often rely on multiple task-specific Named Entity Recognition models to extract structured facets as seen in job search applications. However, this fragmented architecture is brittle, expensive to maintain, and slow to adapt to evolving taxonomies and language patterns. In this paper, we introduce a unified query understanding framework powered by a Large Language Model (LLM), designed to address these limitations. Our approach jointly models the user query and contextual signals such as profile attributes to generate structured interpretations that drive more accurate and personalized recommendations. The framework improves relevance quality in online A/B testing while significantly reducing system complexity and operational overhead. The results demonstrate that our solution provides a scalable and adaptable foundation for query understanding in dynamic web applications.

cs.IR↗
arXiv2025

LLM-Enhanced User-Item Interactions: Leveraging Edge Information for Optimized Recommendations

Graph recommendation methods, representing a connected interaction perspective, reformulate user-item interactions as graphs to leverage graph structure and topology to recommend and have proved practical effectiveness at scale. Large language models, representing a textual generative perspective, excel at modeling user languages, understanding behavioral contexts, capturing user-item semantic relationships, analyzing textual sentiments, and generating coherent and contextually relevant texts as recommendations. However, there is a gap between the connected graph perspective and the text generation perspective as the task formulations are different. A research question arises: how can we effectively integrate the two perspectives for more personalized recsys? To fill this gap, we propose to incorporate graph-edge information into LLMs via prompt and attention innovations. We reformulate recommendations as a probabilistic generative problem using prompts. We develop a framework to incorporate graph edge information from the prompt and attention mechanisms for graph-structured LLM recommendations. We develop a new prompt design that brings in both first-order and second-order graph relationships; we devise an improved LLM attention mechanism to embed direct the spatial and connectivity information of edges. Our evaluation of real-world datasets demonstrates the framework's ability to understand connectivity information in graph data and to improve the relevance and quality of recommendation results.

cs.AI↗
arXiv2025

A Scalable and Efficient Signal Integration System for Job Matching

LinkedIn, one of the world's largest platforms for professional networking and job seeking, encounters various modeling challenges in building recommendation systems for its job matching product, including cold-start, filter bubbles, and biases affecting candidate-job matching. To address these, we developed the STAR (Signal Integration for Talent And Recruiters) system, leveraging the combined strengths of Large Language Models (LLMs) and Graph Neural Networks (GNNs). LLMs excel at understanding textual data, such as member profiles and job postings, while GNNs capture intricate relationships and mitigate cold-start issues through network effects. STAR integrates diverse signals by uniting LLM and GNN capabilities with industrial-scale paradigms including adaptive sampling and version management. It provides an end-to-end solution for developing and deploying embeddings in large-scale recommender systems. Our key contributions include a robust methodology for building embeddings in industrial applications, a scalable GNN-LLM integration for high-performing recommendations, and practical insights for real-world model deployment.

cs.LG↗
arXiv2024

Understanding and Modeling Job Marketplace with Pretrained Language Models

Job marketplace is a heterogeneous graph composed of interactions among members (job-seekers), companies, and jobs. Understanding and modeling job marketplace can benefit both job seekers and employers, ultimately contributing to the greater good of the society. However, existing graph neural network (GNN)-based methods have shallow understandings of the associated textual features and heterogeneous relations. To address the above challenges, we propose PLM4Job, a job marketplace foundation model that tightly couples pretrained language models (PLM) with job market graph, aiming to fully utilize the pretrained knowledge and reasoning ability to model member/job textual features as well as various member-job relations simultaneously. In the pretraining phase, we propose a heterogeneous ego-graph-based prompting strategy to model and aggregate member/job textual features based on the topological structure around the target member/job node, where entity type embeddings and graph positional embeddings are introduced accordingly to model different entities and their heterogeneous relations. Meanwhile, a proximity-aware attention alignment strategy is designed to dynamically adjust the attention of the PLM on ego-graph node tokens in the prompt, such that the attention can be better aligned with job marketplace semantics. Extensive experiments at LinkedIn demonstrate the effectiveness of PLM4Job.

cs.IR↗
arXiv2024

Learning to Retrieve for Job Matching

Web-scale search systems typically tackle the scalability challenge with a two-step paradigm: retrieval and ranking. The retrieval step, also known as candidate selection, often involves extracting standardized entities, creating an inverted index, and performing term matching for retrieval. Such traditional methods require manual and time-consuming development of query models. In this paper, we discuss applying learning-to-retrieve technology to enhance LinkedIns job search and recommendation systems. In the realm of promoted jobs, the key objective is to improve the quality of applicants, thereby delivering value to recruiter customers. To achieve this, we leverage confirmed hire data to construct a graph that evaluates a seeker's qualification for a job, and utilize learned links for retrieval. Our learned model is easy to explain, debug, and adjust. On the other hand, the focus for organic jobs is to optimize seeker engagement. We accomplished this by training embeddings for personalized retrieval, fortified by a set of rules derived from the categorization of member feedback. In addition to a solution based on a conventional inverted index, we developed an on-GPU solution capable of supporting both KNN and term matching efficiently.

cs.IR↗
arXiv2024

Collaborative Large Language Model for Recommender Systems

Recently, there has been growing interest in developing the next-generation recommender systems (RSs) based on pretrained large language models (LLMs). However, the semantic gap between natural language and recommendation tasks is still not well addressed, leading to multiple issues such as spuriously correlated user/item descriptors, ineffective language modeling on user/item data, inefficient recommendations via auto-regression, etc. In this paper, we propose CLLM4Rec, the first generative RS that tightly integrates the LLM paradigm and ID paradigm of RSs, aiming to address the above challenges simultaneously. We first extend the vocabulary of pretrained LLMs with user/item ID tokens to faithfully model user/item collaborative and content semantics. Accordingly, a novel soft+hard prompting strategy is proposed to effectively learn user/item collaborative/content token embeddings via language modeling on RS-specific corpora, where each document is split into a prompt consisting of heterogeneous soft (user/item) tokens and hard (vocab) tokens and a main text consisting of homogeneous item tokens or vocab tokens to facilitate stable and effective language modeling. In addition, a novel mutual regularization strategy is introduced to encourage CLLM4Rec to capture recommendation-related information from noisy user/item content. Finally, we propose a novel recommendation-oriented finetuning strategy for CLLM4Rec, where an item prediction head with multinomial likelihood is added to the pretrained CLLM4Rec backbone to predict hold-out items based on soft+hard prompts established from masked user-item interaction history, where recommendations of multiple items can be generated efficiently without hallucination. Codes are released at https://github.com/yaochenzhu/llm4rec.

cs.IR↗
arXiv2023

Path-Specific Counterfactual Fairness for Recommender Systems

Recommender systems (RSs) have become an indispensable part of online platforms. With the growing concerns of algorithmic fairness, RSs are not only expected to deliver high-quality personalized content, but are also demanded not to discriminate against users based on their demographic information. However, existing RSs could capture undesirable correlations between sensitive features and observed user behaviors, leading to biased recommendations. Most fair RSs tackle this problem by completely blocking the influences of sensitive features on recommendations. But since sensitive features may also affect user interests in a fair manner (e.g., race on culture-based preferences), indiscriminately eliminating all the influences of sensitive features inevitably degenerate the recommendations quality and necessary diversities. To address this challenge, we propose a path-specific fair RS (PSF-RS) for recommendations. Specifically, we summarize all fair and unfair correlations between sensitive features and observed ratings into two latent proxy mediators, where the concept of path-specific bias (PS-Bias) is defined based on path-specific counterfactual inference. Inspired by Pearl's minimal change principle, we address the PS-Bias by minimally transforming the biased factual world into a hypothetically fair world, where a fair RS model can be learned accordingly by solving a constrained optimization problem. For the technical part, we propose a feasible implementation of PSF-RS, i.e., PSF-VAE, with weakly-supervised variational inference, which robustly infers the latent mediators such that unfairness can be mitigated while necessary recommendation diversities can be maximally preserved simultaneously. Experiments conducted on semi-simulated and real-world datasets demonstrate the effectiveness of PSF-RS.

cs.IR↗