arXiv subjects

Hongke Zhao

Publications and source records attributed to Hongke Zhao.

At least 19 recordsLinked to original sources

arXiv2026

From Understanding to Action: Feedback-Grounded Policy Discovery for Generative Recommendation

Semantic-ID-based generative recommenders enable efficient next-item generation, but their item-level supervision mainly captures behavioral co-occurrence and local transitions. Large language models (LLMs) can complement these models by reasoning over heterogeneous interaction histories to understand the user's current demand. However, LLMs are not inherently trained with recommendation-specific outcome feedback, and linguistically plausible reasoning therefore does not necessarily lead to effective recommendation decisions. We term this mismatch the Understanding-Action Gap. Accordingly, we distinguish intent knowledge, which captures the user's current demand, from policy knowledge, which specifies the recommendation direction and rejection boundary under that demand. To bridge this gap, we propose a feedback-driven agent framework that first induces task-oriented intent and then discovers recommendation policies according to their incremental utility over an intent-only baseline. Candidate policies are evaluated and refined using outcome-derived feedback rather than linguistic plausibility. We further transfer the resulting intent and policy knowledge into two latent tokens of a lightweight Semantic-ID generator through dual-space relational distillation, enabling LLM-free online inference. Experiments on public benchmarks show consistent improvements over baselines, while large-scale online A/B tests achieve gains of 4.506% in Revenue and 4.621% in ADVV.

cs.IR↗

arXiv2026

From Trajectories to Prefixes: Reusing Teacher Trajectories via Replayed Prefixes and Online Continuation

Small language models are attractive backbones for interactive agents, but direct distillation from strong teacher trajectories often turns rich multi-turn behavior into one-shot imitation targets. This is inefficient in long-horizon environments, where early decisions shape later states and rewards. We propose Prefix-GRPO, a reinforcement learning framework that decomposes teacher trajectories into replay-aligned prefix queries and online continuations. Each prefix is replayed in the environment to recover a valid intermediate state, after which the student continues online interaction and receives task reward. Unlike response-only GRPO, Prefix-GRPO also applies clipped policy updates to historical assistant tokens inside the replayed prefix, using a policy-distilled SFT checkpoint to estimate their old log-probabilities. This unifies prefix learning and continuation learning within the same policy-optimization form. Experiments on TextCraft, BabyAI, and ALFWorld show that Prefix-GRPO improves small-model agents over distillation and standard RL baselines, while ablations show that replay alone is insufficient without explicit prefix-token optimization. The implementation and reproduction scripts are available at https://github.com/HappynessI/Prefix_GRPO.

cs.LG↗

arXiv2026

LemonHarness Technical Report

As large language model (LLM) agents are applied to longer tasks, they increasingly modify workspace state across multiple rounds of iteration. However, agents typically observe only tool outputs and log fragments, while the actual state changes occur in the file system. Without explicit workspace boundaries, state-changing operations such as file writes and temporary artifact generation may scatter changes across paths. Over time, these weakly constrained changes accumulate, making states such as modified files difficult to track. This paper presents LemonHarness, an integrated execution framework for long-horizon agents. LemonHarness establishes an explicit execution boundary by constraining state-changing operations within a clearly defined workspace and bringing model invocation, tool execution, and rule knowledge within a single controlled boundary. State-changing operations, including file writes, dependency installation, and temporary artifact creation, are executed through structured tool interfaces, with execution feedback recorded as observations available to subsequent model decisions. The system also introduces a reusable rule knowledge base, which turns recurring execution rules and acceptance criteria into runtime knowledge. LemonHarness further adds a time-aware execution mechanism that exposes elapsed and remaining budget to the model, so it can rebalance exploration, implementation, and validation effort as time pressure shifts and avoid timeouts from long waits or excessive verification. On Terminal-Bench 2.0, LemonHarness_GPT-5.3-CodeX reached 84.49% accuracy over 445 trials; pairing the same framework with the stronger GPT-5.5 backbone raised the average accuracy to 86.52% across five jobs. The results suggest that a unified runtime boundary, callable rule knowledge, and time-aware execution can improve the stability of long-horizon agent execution.

cs.AI↗

arXiv2026

Missing Old Logits in Asynchronous Agentic RL: Semantic Mismatch and Repair Methods for Off-Policy Correction

Asynchronous reinforcement learning improves rollout throughput for large language model agents by decoupling sample generation from policy optimization, but it also introduces a critical failure mode for PPO-style off-policy correction. In heterogeneous training systems, the total importance ratio should ideally be decomposed into two semantically distinct factors: a \emph{training--inference discrepancy term} that aligns inference-side and training-side distributions at the same behavior-policy version, and a \emph{policy-staleness term} that constrains the update from the historical policy to the current policy. We show that practical asynchronous pipelines with delayed updates and partial rollouts often lose the required historical training-side logits, or old logits. This missing-old-logit problem entangles discrepancy repair with staleness correction, breaks the intended semantics of decoupled correction, and makes clipping and masking thresholds interact undesirably. To address this issue, we study both exact and approximate correction routes. We propose three exact old-logit acquisition strategies: snapshot-based version tracking, a dedicated old-logit model, and synchronization via partial rollout interruption, and compare their system trade-offs. From the perspective of approximate correction, we focus on preserving the benefits of decoupled correction through a more appropriate approximate policy when exact old logits cannot be recovered at low cost, without incurring extra system overhead. Following this analysis, we adopt a revised PPO-EWMA method, which achieves significant gains in both training speed and optimization performance.

cs.LG↗

arXiv2026

Lishu: A Real-Source Research Workbench for Elite Business Journal Search, Analysis, and Writing Support

This paper presents Lishu, a deployable web artifact for searching, monitoring, and interpreting literature from elite business and management journals. The system integrates the UTD-24 and Financial Times 50 (FT50) journal pools and combines Crossref, OpenAlex, Unpaywall, and optional CORE enrichment to support a broader research workflow than article retrieval alone. In the current implementation, users can search across curated journal pools, apply multi-journal filters, preview open full-text excerpts when available, generate citations and exports, inspect topic and affiliation structure, produce review drafts, simulate virtual peer review, and assemble grant-oriented research narratives. Unlike static journal directories or general-purpose academic search engines, the artifact is explicitly scoped to high-status management outlets and is designed to support sensemaking tasks that matter to researchers, doctoral students, and lab managers: identifying recent work, surfacing topical concentration, comparing themes, and converting search output into actionable research material. Architecturally, the system emphasizes source transparency, modularity, and low-cost public deployability through a lightweight Node.js service layer, a multi-page client interface, optional large-language-model enhancement for interpretation and writing support, and a free-tier persistence path through Supabase. The paper contributes both a functioning design artifact and an extensible architectural pattern for journal-pool-specific scholarly discovery and writing support, with implications for digital research infrastructure in information systems and business scholarship.

cs.DL↗

arXiv2026

From Exposure to Internalization: Dual-Stream Calibration for In-context Clinical Reasoning

Contextual clinical reasoning demands robust inference grounded in complex, heterogeneous clinical records. While state-of-the-art fine-tuning, in-context learning (ICL), and retrieval-augmented generation (RAG) enable knowledge exposure, they often fall short of genuine contextual internalization: dynamically adjusting a model's internal representations to the subtle nuances of individual cases at inference time. To address this, we propose Dual-Stream Calibration (DSC), a test-time training framework that transcends superficial knowledge exposure to achieve deep internalization during inference. DSC facilitates input internalization by synergistically aligning two calibration streams. Unlike passive context exposure, the Semantic Calibration Stream enforces a deliberative reflection on core evidence, internalizing semantic anchors by minimizing entropy to stabilize generative trajectories. Simultaneously, the Structural Calibration Stream assimilates latent inferential dependencies through an iterative meta-learning objective. By training on specialized support sets at test-time, this stream enables the model to bridge the gap between external evidence and internal logic, synthesizing fragmented data into a coherent response. Our approach shifts the reasoning paradigm from passive attention-based matching to an active refinement of the latent inferential space. Validated against thirteen clinical datasets, DSC demonstrates superiority across three distinct task paradigms, consistently outstripping state-of-the-art baselines ranging from training-dependent models to test-time learning frameworks.

q-bio.QM↗

arXiv2026

RL-VLA$^3$: A Flexible and Asynchronous Reinforcement Learning Framework for VLA Training

Reinforcement learning (RL) has emerged as a critical paradigm for post-training Vision-Language-Action (VLA) models, enabling embodied agents to adapt and improve through environmental interaction. However, existing RL frameworks for VLAs inherit synchronous design principles from traditional LLM training, treating entire rollouts as indivisible units and alternating strictly between data collection and policy optimization. This fundamentally mismatches the unique characteristics of VLA training, as physical simulators introduce highly variable, resource-intensive latencies. To address this, we introduce RL-VLA$^3$, a fully asynchronous distributed RL framework that enables fine-grained asynchronous interaction between simulation, inference, and training components through dynamic batching schedulers and flexible environment sharding strategies. Extensive experiments across diverse simulation backends, VLA architectures, and RL algorithms demonstrate that RL-VLA$^3$ achieves throughput improvements of up to 85.2\% over synchronous baselines while maintaining identical sample efficiency, with scalability validated from 8 to 256 GPUs. To our knowledge, RL-VLA$^3$ is the first fully asynchronous RL training framework tailored specifically for the system-level challenges of VLA training.

cs.AI↗

arXiv2025

Reinventing Clinical Dialogue: Agentic Paradigms for LLM Enabled Healthcare Communication

Clinical dialogue represents a complex duality requiring both the empathetic fluency of natural conversation and the rigorous precision of evidence-based medicine. While Large Language Models possess unprecedented linguistic capabilities, their architectural reliance on reactive and stateless processing often favors probabilistic plausibility over factual veracity. This structural limitation has catalyzed a paradigm shift in medical AI from generative text prediction to agentic autonomy, where the model functions as a central reasoning engine capable of deliberate planning and persistent memory. Moving beyond existing reviews that primarily catalog downstream applications, this survey provides a first-principles analysis of the cognitive architecture underpinning this shift. We introduce a novel taxonomy structured along the orthogonal axes of knowledge source and agency objective to delineate the provenance of clinical knowledge against the system's operational scope. This framework facilitates a systematic analysis of the intrinsic trade-offs between creativity and reliability by categorizing methods into four archetypes: \textit{Latent Space Clinicians}, \textit{Emergent Planners}, \textit{Grounded Synthesizers}, and \textit{Verifiable Workflow Automators}. For each paradigm, we deconstruct the technical realization across the entire cognitive pipeline, encompassing strategic planning, memory management, action execution, collaboration, and evolution to reveal how distinct architectural choices balance the tension between autonomy and safety.

q-bio.OT↗

arXiv2025

Grounded by Experience: Generative Healthcare Prediction Augmented with Hierarchical Agentic Retrieval

Accurate healthcare prediction is critical for improving patient outcomes and reducing operational costs. Bolstered by growing reasoning capabilities, large language models (LLMs) offer a promising path to enhance healthcare predictions by drawing on their rich parametric knowledge. However, LLMs are prone to factual inaccuracies due to limitations in the reliability and coverage of their embedded knowledge. While retrieval-augmented generation (RAG) frameworks, such as GraphRAG and its variants, have been proposed to mitigate these issues by incorporating external knowledge, they face two key challenges in the healthcare scenario: (1) identifying the clinical necessity to activate the retrieval mechanism, and (2) achieving synergy between the retriever and the generator to craft contextually appropriate retrievals. To address these challenges, we propose GHAR, a \underline{g}enerative \underline{h}ierarchical \underline{a}gentic \underline{R}AG framework that simultaneously resolves when to retrieve and how to optimize the collaboration between submodules in healthcare. Specifically, for the first challenge, we design a dual-agent architecture comprising Agent-Top and Agent-Low. Agent-Top acts as the primary physician, iteratively deciding whether to rely on parametric knowledge or to initiate retrieval, while Agent-Low acts as the consulting service, summarising all task-relevant knowledge once retrieval was triggered. To tackle the second challenge, we innovatively unify the optimization of both agents within a formal Markov Decision Process, designing diverse rewards to align their shared goal of accurate prediction while preserving their distinct roles. Extensive experiments on three benchmark datasets across three popular tasks demonstrate our superiority over state-of-the-art baselines, highlighting the potential of hierarchical agentic RAG in advancing healthcare systems.

cs.AI↗

arXiv2025

Hierarchical Semantic RL: Tackling the Problem of Dynamic Action Space for RL-based Recommendations

Recommender Systems (RS) are fundamental to modern online services. While most existing approaches optimize for short-term engagement, recent work has begun to explore reinforcement learning (RL) to model long-term user value. However, these efforts face significant challenges due to the vast, dynamic action spaces inherent in RS, which hinder stable policy learning. To resolve this bottleneck, we introduce Hierarchical Semantic RL (HSRL), which reframes RL-based recommendation over a fixed Semantic Action Space (SAS). HSRL encodes items as Semantic IDs (SIDs) for policy learning, and maps SIDs back to their original items via a fixed lookup during execution. To align decision-making with SID generation, the Hierarchical Policy Network (HPN) operates in a coarse-to-fine manner, employing hierarchical residual state modeling to refine each level's context from the previous level's residual, thereby reducing representation-decision mismatch. In parallel, a Multi-level Critic (MLC) provides token-level value estimates, enabling fine-grained credit assignment. Across public benchmarks and a large-scale production dataset from a leading short-video advertising platform, HSRL consistently surpasses state-of-the-art baselines. In online deployment over a 7-day A/B testing, it delivers an 18.421% ADVV lift and a 1.251% increase in Revenue, supporting HSRL as a scalable paradigm for RL-based recommendation.

cs.IR↗

arXiv2025

Recall-Extend Dynamics: Enhancing Small Language Models through Controlled Exploration and Refined Offline Integration

Many existing studies have achieved significant improvements in the reasoning capabilities of large language models (LLMs) through reinforcement learning with verifiable rewards (RLVR), while the enhancement of reasoning abilities in small language models (SLMs) has not yet been sufficiently explored. Combining distilled data from larger models with RLVR on small models themselves is a natural approach, but it still faces various challenges and issues. Therefore, we propose \textit{\underline{R}}ecall-\textit{\underline{E}}xtend \textit{\underline{D}}ynamics(RED): Enhancing Small Language Models through Controlled Exploration and Refined Offline Integration. In this paper, we explore the perspective of varying exploration spaces, balancing offline distillation with online reinforcement learning. Simultaneously, we specifically design and optimize for the insertion problem within offline data. By monitoring the ratio of entropy changes in the model concerning offline and online data, we regulate the weight of offline-SFT, thereby addressing the issues of insufficient exploration space in small models and the redundancy and complexity during the distillation process. Furthermore, to tackle the distribution discrepancies between offline data and the current policy, we design a sample-accuracy-based policy shift mechanism that dynamically chooses between imitating offline distilled data and learning from its own policy.

cs.LG↗

arXiv2025

Diffmv: A Unified Diffusion Framework for Healthcare Predictions with Random Missing Views and View Laziness

Advanced healthcare predictions offer significant improvements in patient outcomes by leveraging predictive analytics. Existing works primarily utilize various views of Electronic Health Record (EHR) data, such as diagnoses, lab tests, or clinical notes, for model training. These methods typically assume the availability of complete EHR views and that the designed model could fully leverage the potential of each view. However, in practice, random missing views and view laziness present two significant challenges that hinder further improvements in multi-view utilization. To address these challenges, we introduce Diffmv, an innovative diffusion-based generative framework designed to advance the exploitation of multiple views of EHR data. Specifically, to address random missing views, we integrate various views of EHR data into a unified diffusion-denoising framework, enriched with diverse contextual conditions to facilitate progressive alignment and view transformation. To mitigate view laziness, we propose a novel reweighting strategy that assesses the relative advantages of each view, promoting a balanced utilization of various data views within the model. Our proposed strategy achieves superior performance across multiple health prediction tasks derived from three popular datasets, including multi-view and multi-modality scenarios.

cs.LG↗

arXiv2025

Attention Mechanisms Perspective: Exploring LLM Processing of Graph-Structured Data

Attention mechanisms are critical to the success of large language models (LLMs), driving significant advancements in multiple fields. However, for graph-structured data, which requires emphasis on topological connections, they fall short compared to message-passing mechanisms on fixed links, such as those employed by Graph Neural Networks (GNNs). This raises a question: ``Does attention fail for graphs in natural language settings?'' Motivated by these observations, we embarked on an empirical study from the perspective of attention mechanisms to explore how LLMs process graph-structured data. The goal is to gain deeper insights into the attention behavior of LLMs over graph structures. We uncovered unique phenomena regarding how LLMs apply attention to graph-structured data and analyzed these findings to improve the modeling of such data by LLMs. The primary findings of our research are: 1) While LLMs can recognize graph data and capture text-node interactions, they struggle to model inter-node relationships within graph structures due to inherent architectural constraints. 2) The attention distribution of LLMs across graph nodes does not align with ideal structural patterns, indicating a failure to adapt to graph topology nuances. 3) Neither fully connected attention nor fixed connectivity is optimal; each has specific limitations in its application scenarios. Instead, intermediate-state attention windows improve LLM training performance and seamlessly transition to fully connected windows during inference. Source code: \href{https://github.com/millioniron/LLM_exploration}{LLM4Exploration}

cs.AI↗

arXiv2024

Collaborative Knowledge Fusion: A Novel Approach for Multi-task Recommender Systems via LLMs

Owing to the impressive general intelligence of large language models (LLMs), there has been a growing trend to integrate them into recommender systems to gain a more profound insight into human interests and intentions. Existing LLMs-based recommender systems primarily leverage item attributes and user interaction histories in textual format, improving the single task like rating prediction or explainable recommendation. Nevertheless, these approaches overlook the crucial contribution of traditional collaborative signals in discerning users' profound intentions and disregard the interrelatedness among tasks. To address these limitations, we introduce a novel framework known as CKF, specifically developed to boost multi-task recommendations via personalized collaborative knowledge fusion into LLMs. Specifically, our method synergizes traditional collaborative filtering models to produce collaborative embeddings, subsequently employing the meta-network to construct personalized mapping bridges tailored for each user. Upon mapped, the embeddings are incorporated into meticulously designed prompt templates and then fed into an advanced LLM to represent user interests. To investigate the intrinsic relationship among diverse recommendation tasks, we develop Multi-Lora, a new parameter-efficient approach for multi-task optimization, adept at distinctly segregating task-shared and task-specific information. This method forges a connection between LLMs and recommendation scenarios, while simultaneously enriching the supervisory signal through mutual knowledge transfer among various tasks. Extensive experiments and in-depth robustness analyses across four common recommendation tasks on four large public data sets substantiate the effectiveness and superiority of our framework.

cs.IR↗

arXiv2024

GANPrompt: Enhancing Robustness in LLM-Based Recommendations with GAN-Enhanced Diversity Prompts

In recent years, Large Language Models (LLMs) have demonstrated remarkable proficiency in comprehending and generating natural language, with a growing prevalence in the domain of recommendation systems. However, LLMs still face a significant challenge called prompt sensitivity, which refers to that it is highly susceptible to the influence of prompt words. This inconsistency in response to minor alterations in prompt input may compromise the accuracy and resilience of recommendation models. To address this issue, this paper proposes GANPrompt, a multi-dimensional LLMs prompt diversity framework based on Generative Adversarial Networks (GANs). The framework enhances the model's adaptability and stability to diverse prompts by integrating GANs generation techniques with the deep semantic understanding capabilities of LLMs. GANPrompt first trains a generator capable of producing diverse prompts by analysing multidimensional user behavioural data. These diverse prompts are then used to train the LLMs to improve its performance in the face of unseen prompts. Furthermore, to ensure a high degree of diversity and relevance of the prompts, this study introduces a mathematical theory-based diversity constraint mechanism that optimises the generated prompts to ensure that they are not only superficially distinct, but also semantically cover a wide range of user intentions. Through extensive experiments on multiple datasets, we demonstrate the effectiveness of the proposed framework, especially in improving the adaptability and robustness of recommendation systems in complex and dynamic environments. The experimental results demonstrate that GANPrompt yields substantial enhancements in accuracy and robustness relative to existing state-of-the-art methodologies.

cs.IR↗

arXiv2024

LANE: Logic Alignment of Non-tuning Large Language Models and Online Recommendation Systems for Explainable Reason Generation

The explainability of recommendation systems is crucial for enhancing user trust and satisfaction. Leveraging large language models (LLMs) offers new opportunities for comprehensive recommendation logic generation. However, in existing related studies, fine-tuning LLM models for recommendation tasks incurs high computational costs and alignment issues with existing systems, limiting the application potential of proven proprietary/closed-source LLM models, such as GPT-4. In this work, our proposed effective strategy LANE aligns LLMs with online recommendation systems without additional LLMs tuning, reducing costs and improving explainability. This innovative approach addresses key challenges in integrating language models with recommendation systems while fully utilizing the capabilities of powerful proprietary models. Specifically, our strategy operates through several key components: semantic embedding, user multi-preference extraction using zero-shot prompting, semantic alignment, and explainable recommendation generation using Chain of Thought (CoT) prompting. By embedding item titles instead of IDs and utilizing multi-head attention mechanisms, our approach aligns the semantic features of user preferences with those of candidate items, ensuring coherent and user-aligned recommendations. Sufficient experimental results including performance comparison, questionnaire voting, and visualization cases prove that our method can not only ensure recommendation performance, but also provide easy-to-understand and reasonable recommendation logic.

cs.IR↗

arXiv2024

Performative Debias with Fair-exposure Optimization Driven by Strategic Agents in Recommender Systems

Data bias, e.g., popularity impairs the dynamics of two-sided markets within recommender systems. This overshadows the less visible but potentially intriguing long-tail items that could capture user interest. Despite the abundance of research surrounding this issue, it still poses challenges and remains a hot topic in academic circles. Along this line, in this paper, we developed a re-ranking approach in dynamic settings with fair-exposure optimization driven by strategic agents. Designed for the producer side, the execution of agents assumes content creators can modify item features based on strategic incentives to maximize their exposure. This iterative process entails an end-to-end optimization, employing differentiable ranking operators that simultaneously target accuracy and fairness. Joint objectives ensure the performance of recommendations while enhancing the visibility of tail items. We also leveraged the performativity nature of predictions to illustrate how strategic learning influences content creators to shift towards fairness efficiently, thereby incentivizing features of tail items. Through comprehensive experiments on both public and industrial datasets, we have substantiated the effectiveness and dominance of the proposed method especially on unveiling the potential of tail items.

cs.IR↗

arXiv2024

Cross-domain Transfer of Valence Preferences via a Meta-optimization Approach

Cross-domain recommendation offers a potential avenue for alleviating data sparsity and cold-start problems. Embedding and mapping, as a classic cross-domain research genre, aims to identify a common mapping function to perform representation transformation between two domains. Nevertheless, previous coarse-grained preference representations, non-personalized mapping functions, and excessive reliance on overlapping users limit their performance, especially in scenarios where overlapping users are sparse. To address aforementioned challenges, we propose a novel cross-domain approach, namely CVPM. CVPM formalizes cross-domain interest transfer as a hybrid architecture of parametric meta-learning and self-supervised learning, which not only transfers user preferences at a finer level, but also enables signal enhancement with the knowledge of non-overlapping users. Specifically, with deep insights into user preferences and valence preference theory, we believe that there exists significant difference between users' positive preferences and negative behaviors, and thus employ differentiated encoders to learn their distributions. In particular, we further utilize the pre-trained model and item popularity to sample pseudo-interaction items to ensure the integrity of both distributions. To guarantee the personalization of preference transfer, we treat each user's mapping as two parts, the common transformation and the personalized bias, where the network used to generate the personalized bias is output by a meta-learner. Furthermore, in addition to the supervised loss for overlapping users, we design contrastive tasks for non-overlapping users from both group and individual-levels to avoid model skew and enhance the semantics of representations. Exhaustive data analysis and extensive experimental results demonstrate the effectiveness and advancement of our proposed framework.

cs.IR↗