arXiv subjects

Zhongni Hou

Publications and source records attributed to Zhongni Hou.

13 recordsLinked to original sources

arXiv2026

ATLAS: Dual-Horizon Diagnostic Evaluation for Tool-Use Agents in Meituan Xiaotua

LLM-based assistants on web-scale service platforms increasingly address local-service needs through multi-step planning and tool use. Under dynamic business conditions and evolving user needs, outcome-level evaluation offers limited evidence of where execution fails or how service drifts across continued interaction. We present ATLAS, a dual-horizon diagnostic evaluation framework for industrial tool-use agents. Within-turn evaluation uses trajectory-wise signals across thinking and reflection, tool and skill execution, and response generation to identify execution failures and unmet capability requirements. Across-turn evaluation uses user-wise signals spanning multiple lifecycles to diagnose failures in maintaining or updating user intent, constraints, and corrections. Fine-grained signals use rule-based checks and LLM judgments with explicit evidence scopes and decision boundaries. To support reliable diagnosis at production scale, LLM judges are calibrated against dynamically maintained references from business logs, and selected judges are distilled into compact diagnostic models. The resulting diagnostic profiles help developers attribute failures, prioritize targeted interventions, and reassess improvements. Applied in Meituan Xiaotuan, a million-scale DAU local-services assistant, ATLAS achieves high diagnostic fidelity across calibrated judges and distilled models. Blinded human auditing and a 14-day online A/B test with nearly two million users per arm confirm that the ATLAS-guided agent improves recommendation relevance, factual reliability, user engagement (+2.84% answer read-through rate), and order volume (+6.13% paid orders), leading to full production deployment.

cs.AI↗

arXiv2026

On-Policy Self-Distillation with Teacher Evolution

On-Policy Self-Distillation (OPSD), which aligns a student model's token-level distribution with that of its self-teacher prompted by privileged context, has emerged as a promising direction to mitigate reward sparsity in sequence-level post-training paradigms. However, OPSD often fails to achieve continuous performance gains during training. In this paper, we find that the model's limited self-teaching capability underlies this phenomenon by generating unreliable guidance on certain samples. To address this problem, we propose Teacher-Evolution Self-Distillation (TESD), which enables the self-teacher to generate rollouts and receive environmental feedback based on task correctness and context independence. First, using these rewards as training objectives, we perform teacher-oriented optimization to continuously enhance the model's self-teaching capability during distillation. Second, by estimating teacher guidance reliability from these rewards, we apply adaptive student-oriented optimization to down-weight samples with unreliable guidance. Extensive experiments across multiple datasets and model scales show that TESD consistently outperforms representative token-level and sequence-level methods, highlighting the necessity of improving supervision quality in OPSD.

cs.LG↗

arXiv2026

LOOP: Recovery-Aligned Iterative Layer Pruning via On-Policy Distillation

Layer-level pruning is attractive for large language model (LLM) deployment as it reduces the model size while preserving a dense architecture that can be efficiently executed on standard hardware. Current methods typically follow a one-shot paradigm, which first removes selected layers to reach a target model size and then performs additional training to recover model capabilities. However, pruning directly to the target size can substantially degrade reasoning capabilities, making subsequent recovery more difficult under a limited training budget. Moreover, this paradigm separates layer selection from recovery, limiting the use of recovery information to guide selection and potentially hindering the recovery of reasoning capabilities. To address these limitations, we propose LOOP, a Layer-level Recovery-Oriented On-Policy Pruning framework that interleaves pruning and recovery across multiple shots. In each shot, LOOP removes a single layer and performs short on-policy distillation, mitigating the abrupt degradation in reasoning capabilities associated with one-shot pruning. To guide layer selection in subsequent shots, LOOP reuses on-policy trajectories and optimization signals from recovery to jointly assess each remaining layer's reasoning contribution and recovery sensitivity, thereby making layer selection recovery-aware. Evaluations on several benchmarks show that LOOP consistently outperforms one-shot baselines in average accuracy across different model scales and pruning ratios.

cs.LG↗

arXiv2026

Predictive Bandit Curriculum Reinforcement Learning via Gradient-Based Sample Group Discovery

Curriculum learning, which organizes training samples according to their learning priority, has been widely adopted in reinforcement learning to improve training efficiency. However, existing methods commonly construct curriculum units using predefined data properties and estimate their utility by simply aggregating a single proxy signal from recent training history.Such designs may group samples with different optimization effects and fail to anticipate how their utility evolves in relation to the policy. We propose Predictive Gradient-based Bandit (PGB) Curriculum, which first discovers optimization-aware learning units by clustering projected sample-level gradients. PGB then employs a lightweight temporal predictor to forecast the future utility of each unit and converts the predictions into a Boltzmann sampling distribution for adaptive data allocation. Experiments across multiple model scales and reasoning benchmarks show that PGB consistently outperforms standard GRPO, difficulty-based curricula, and history-based scheduling strategies, demonstrating improved data utilization and training efficiency.

cs.LG↗

arXiv2026

MemScope: Evolving Agent Memory Architectures through Localized Code Repair

Agent memory architectures define how LLM agents store and reuse experience across tasks, but they are typically fixed after deployment, limiting their adaptability across task types. Recent self-evolving approaches instead adapt these memory architectures using execution feedback, typically represented as evaluation scores or compressed summaries. However, such coarse-grained feedback is insufficient for precise failure localization in current agent systems, where long-horizon task execution involves multiple interacting components. Without a precise failure location, existing approaches lack a clear code-level target for modification, leaving the update scope weakly constrained and often leading to large-scale rewrites of an otherwise capable memory architecture. We propose MemScope, a framework for evidence-grounded, localized evolution of an existing memory architecture. Its tool-using Patch Agent jointly inspects raw execution traces, runtime memory state, and current architecture code to identify repair targets and apply bounded, in-place edits under provider-identity and memory-compatibility constraints. On GAIA, the trace-and-memory inspection variant achieves 86.67% accuracy, compared with 70.00% for trace-only and 63.33% for summary-only feedback. Under xBench tasks, the evolved memory architecture improves over its original version from 49 to 55 correct answers with DeepSeek V3.2 and from 50 to 53 with Kimi K2.7 Code. On disjoint ALFWorld unseen tasks, success rises from 50.75% to 71.64%. Across archived architecture-snapshot transitions, the median code-change ratio is 4.4%. These results support fine-grained execution evidence as a basis for localized, in-place memory-architecture evolution.

cs.AI↗

arXiv2026

On-Policy Parameter Update Direction Underlies Generalization in LLM Post-Training

The strong generalization performance of on-policy post-training paradigms has motivated studies of their parameter update behaviors. However, these studies treat the observed behaviors only as byproducts in on-policy training, overlooking their potential to serve as optimization principles for improving the generalization of other paradigms such as supervised fine-tuning (SFT). To address this limitation, we investigate whether there exists a specific on-policy update behavior that can achieve such improvements. First, our analyses reveal that SFT updates parameters along consistent directions, while the on-policy paradigm continuously adjusts the direction during training. This difference inspires us to focus on the cumulative update direction of each parameter as a promising behavior. Then, we evaluate its effectiveness for improving generalization by proposing On-Policy direction-constrained Supervised Fine-Tuning (OPSFT), which constrains SFT updates to the direction identified by on-policy paradigms. The strong performance of OPSFT indicates that the generalization advantage of on-policy paradigms can be transferred to SFT through the parameter update direction. Once such a direction is identified, even SFT can generalize with its updates constrained to this direction. This finding offers two practical benefits by combining the strong generalization of on-policy paradigms with the advantages of SFT, including the high training efficiency and ability to leverage high-quality trajectories. For efficiency, we identify update directions that support strong generalization using a few on-policy training steps, and subsequently apply OPSFT to achieve high training efficiency. For leveraging high-quality trajectories, OPSFT can utilize these trajectories to continue improving a post-trained model along its update direction without disrupting the ability learned from on-policy training.

cs.LG↗

arXiv2026

Beyond Dialogue Time: Temporal Semantic Memory for Personalized LLM Agents

Memory enables Large Language Model (LLM) agents to perceive, store, and use information from past dialogues, which is essential for personalization. However, existing methods fail to properly model the temporal dimension of memory in two aspects: 1) Temporal inaccuracy: memories are organized by dialogue time rather than their actual occurrence time; 2) Temporal fragmentation: existing methods focus on point-wise memory, losing durative information that captures persistent states and evolving patterns. To address these limitations, we propose Temporal Semantic Memory (TSM), a memory framework that models semantic time for point-wise memory and supports the construction and utilization of durative memory. During memory construction, it first builds a semantic timeline rather than a dialogue one. Then, it consolidates temporally continuous and semantically related information into a durative memory. During memory utilization, it incorporates the query's temporal intent on the semantic timeline, enabling the retrieval of temporally appropriate durative memories and providing time-valid, duration-consistent context to support response generation. Experiments on LongMemEval and LoCoMo show that TSM consistently outperforms existing methods and achieves up to 12.2% absolute improvement in accuracy, demonstrating the effectiveness of the proposed method.

cs.AI↗

arXiv2026

HiDiffTIR: Hierarchical Difficulty-Aware Policy Optimization for Multi-Turn Tool-Integrated Reasoning

Tool-Integrated Reasoning (TIR) is a fundamental capability for LLM agents to solve complex tasks by interacting with external tools iteratively. Reinforcement Learning (RL) has become the dominant paradigm for enabling this capability. However, existing approaches typically assign uniform trajectory-level advantages and treat all correct tool calls equally, ignoring the varying difficulty and learning value across trajectories and reasoning steps. This can lead to imprecise learning signals that do not adequately distinguish between trivial and challenging tool-use patterns. To address this limitation, we propose HiDiffTIR, a Hierarchical Difficulty-aware policy optimization framework for multi-turn TIR. HiDiffTIR performs difficulty-aware credit assignment at both trajectory and turn levels, enabling the policy to focus on more informative trajectories and harder reasoning steps. Notably, this fine-grained optimization is achieved without additional supervision, relying solely on group-level statistics derived from standard RL rollouts. Extensive experiments on three tool-using benchmarks demonstrate that HiDiffTIR consistently improves multi-turn TIR performance and tool invocation accuracy over strong RL baselines, highlighting the necessity of difficulty-aware credit assignment for effective policy optimization in tool-integrated LLM agents.

cs.CL↗

arXiv2026

R$^{2}$Adapter: A Routing and Rewriting Adapter for Efficient Hybrid RAG

Retrieval-Augmented Generation (RAG) has become a prevailing paradigm for enhancing Large Language Models (LLMs) with non-parametric knowledge. Vanilla RAG efficiently handles simple queries but struggles with relational or multi-hop reasoning. Graph-based RAG alleviates this issue but incurs higher inference complexity and latency. In practice, user queries can differ significantly in their complexity, rendering a fixed RAG strategy suboptimal. However, existing hybrid text-graph RAG methods typically rely on heuristic and LLM-based routing, resulting in unnecessary overhead and strong dependence on the underlying LLM. To address these challenges, we propose R$^{2}$Adapter, a lightweight plug-in Routing and Rewriting Adapter designed to allocate queries between vanilla and graph-based RAG dynamically. By routing only the queries that genuinely benefit from graph-based reasoning, R$^{2}$Adapter reduces unnecessary graph retrieval overhead. Additionally, uncertain graph-routed queries are rewritten to better expose their multi-hop reasoning requirements, improving retrieval quality without additional supervision. Extensive experiments on three multi-hop QA benchmarks demonstrate that R$^{2}$Adapter reduces graph-based RAG usage by up to 59% while maintaining comparable answer accuracy. This adapter is model-agnostic and can be seamlessly integrated into diverse vanilla and graph-based RAG pipelines, providing an efficient and adaptive solution for hybrid RAG systems.

cs.CL↗

arXiv2025

Mixture Policy based Multi-Hop Reasoning over N-tuple Temporal Knowledge Graphs

Temporal Knowledge Graphs (TKGs), which utilize quadruples in the form of (subject, predicate, object, timestamp) to describe temporal facts, have attracted extensive attention. N-tuple TKGs (N-TKGs) further extend traditional TKGs by utilizing n-tuples to incorporate auxiliary elements alongside core elements (i.e., subject, predicate, and object) of facts, so as to represent them in a more fine-grained manner. Reasoning over N-TKGs aims to predict potential future facts based on historical ones. However, existing N-TKG reasoning methods often lack explainability due to their black-box nature. Therefore, we introduce a new Reinforcement Learning-based method, named MT-Path, which leverages the temporal information to traverse historical n-tuples and construct a temporal reasoning path. Specifically, in order to integrate the information encapsulated within n-tuples, i.e., the entity-irrelevant information within the predicate, the information about core elements, and the complete information about the entire n-tuples, MT-Path utilizes a mixture policy-driven action selector, which bases on three low-level policies, namely, the predicate-focused policy, the core-element-focused policy and the whole-fact-focused policy. Further, MT-Path utilizes an auxiliary element-aware GCN to capture the rich semantic dependencies among facts, thereby enabling the agent to gain a deep understanding of each n-tuple. Experimental results demonstrate the effectiveness and the explainability of MT-Path.

cs.AI↗

arXiv2024

LongReward: Improving Long-context Large Language Models with AI Feedback

Though significant advancements have been achieved in developing long-context large language models (LLMs), the compromised quality of LLM-synthesized data for supervised fine-tuning (SFT) often affects the long-context performance of SFT models and leads to inherent limitations. In principle, reinforcement learning (RL) with appropriate reward signals can further enhance models' capacities. However, how to obtain reliable rewards in long-context scenarios remains unexplored. To this end, we propose LongReward, a novel method that utilizes an off-the-shelf LLM to provide rewards for long-context model responses from four human-valued dimensions: helpfulness, logicality, faithfulness, and completeness, each with a carefully designed assessment pipeline. By combining LongReward and offline RL algorithm DPO, we are able to effectively improve long-context SFT models. Our experiments indicate that LongReward not only significantly improves models' long-context performance but also enhances their ability to follow short instructions. We also find that long-context DPO with LongReward and conventional short-context DPO can be used together without hurting either one's performance.

cs.CL↗

arXiv2024

Selective Temporal Knowledge Graph Reasoning

Temporal Knowledge Graph (TKG), which characterizes temporally evolving facts in the form of (subject, relation, object, timestamp), has attracted much attention recently. TKG reasoning aims to predict future facts based on given historical ones. However, existing TKG reasoning models are unable to abstain from predictions they are uncertain, which will inevitably bring risks in real-world applications. Thus, in this paper, we propose an abstention mechanism for TKG reasoning, which helps the existing models make selective, instead of indiscriminate, predictions. Specifically, we develop a confidence estimator, called Confidence Estimator with History (CEHis), to enable the existing TKG reasoning models to first estimate their confidence in making predictions, and then abstain from those with low confidence. To do so, CEHis takes two kinds of information into consideration, namely, the certainty of the current prediction and the accuracy of historical predictions. Experiments with representative TKG reasoning models on two benchmark datasets demonstrate the effectiveness of the proposed CEHis.

cs.LG↗

arXiv2022

HiSMatch: Historical Structure Matching based Temporal Knowledge Graph Reasoning

A Temporal Knowledge Graph (TKG) is a sequence of KGs with respective timestamps, which adopts quadruples in the form of (\emph{subject}, \emph{relation}, \emph{object}, \emph{timestamp}) to describe dynamic facts. TKG reasoning has facilitated many real-world applications via answering such queries as (\emph{query entity}, \emph{query relation}, \emph{?}, \emph{future timestamp}) about future. This is actually a matching task between a query and candidate entities based on their historical structures, which reflect behavioral trends of the entities at different timestamps. In addition, recent KGs provide background knowledge of all the entities, which is also helpful for the matching. Thus, in this paper, we propose the \textbf{Hi}storical \textbf{S}tructure \textbf{Match}ing (\textbf{HiSMatch}) model. It applies two structure encoders to capture the semantic information contained in the historical structures of the query and candidate entities. Besides, it adopts another encoder to integrate the background knowledge into the model. TKG reasoning experiments on six benchmark datasets demonstrate the significant improvement of the proposed HiSMatch model, with up to 5.6\% performance improvement in MRR, compared to the state-of-the-art baselines.

cs.AI↗