arXiv subjects

Jingwen Xu

Publications and source records attributed to Jingwen Xu.

At least 19 recordsLinked to original sources

arXiv2026

From Raw Experience to Skill Consumption: A Systematic Study of Model-Generated Agent Skills

Language agents increasingly improve by reusing \emph{skills} -- structured procedural artifacts distilled from past experience. In particular, \emph{domain-level} and \emph{model-generated} skills are especially promising. They offer fast adaptation within a domain by encoding domain-specific recurring procedures, and they scale beyond labor-intensive hand-crafting. However, while extraction methods continue to proliferate, understanding remains limited, with no comprehensive study spanning the full skill lifecycle -- \textbf{experience generation}, \textbf{skill extraction}, and \textbf{skill consumption} -- to ask whether such skills actually work, when they work, and what makes them succeed or fail. To close this gap, we build a utility-grounded evaluation framework that provides systematic experimental results across extractors and target agents, covering five diverse agentic task domains. We find that model-generated skills are beneficial on average but exhibit non-trivial negative transfer, and that neither extractors nor targets behave uniformly. A model can be a strong extractor yet a weak consumer, or vice versa, with skill utility independent of model scale or baseline task strength. To explain these patterns, we then dissect each lifecycle stage in depth, analyzing how experience composition shapes skill quality, what properties characterize useful skills, and how the same skill transfers across different consumers. Finally, we translate these findings into a concrete \emph{meta-skill} that guides skill extraction toward the features tied to actual utility, which consistently improves skill quality across domains and substantially reduces negative transfer.

cs.AI↗

arXiv2026

Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges

Reinforcement Learning from Human Feedback (RLHF) and related alignment paradigms have become central to steering large language models (LLMs) and multimodal large language models (MLLMs) toward human-preferred behaviors. However, these approaches introduce a systemic vulnerability: reward hacking, where models exploit imperfections in learned reward signals to maximize proxy objectives without fulfilling true task intent. As models scale and optimization intensifies, such exploitation manifests as verbosity bias, sycophancy, hallucinated justification, benchmark overfitting, and, in multimodal settings, perception--reasoning decoupling and evaluator manipulation. Recent evidence further suggests that seemingly benign shortcut behaviors can generalize into broader forms of misalignment, including deception and strategic gaming of oversight mechanisms. In this survey, we propose the Proxy Compression Hypothesis (PCH) as a unifying framework for understanding reward hacking. We formalize reward hacking as an emergent consequence of optimizing expressive policies against compressed reward representations of high-dimensional human objectives. Under this view, reward hacking arises from the interaction of objective compression, optimization amplification, and evaluator--policy co-adaptation. This perspective unifies empirical phenomena across RLHF, RLAIF, and RLVR regimes, and explains how local shortcut learning can generalize into broader forms of misalignment, including deception and strategic manipulation of oversight mechanisms. We further organize detection and mitigation strategies according to how they intervene on compression, amplification, or co-adaptation dynamics. By framing reward hacking as a structural instability of proxy-based alignment under scale, we highlight open challenges in scalable oversight, multimodal grounding, and agentic autonomy.

cs.LG↗

arXiv2026

Silence the Judge: Reinforcement Learning with Self-Verifier via Latent Geometric Clustering

Group Relative Policy Optimization (GRPO) significantly enhances the reasoning performance of Large Language Models (LLMs). However, this success heavily relies on expensive external verifiers or human rules. Such dependency not only leads to significant computational costs and training latency, but also yields sparse rewards that hinder optimization efficiency. To address these challenges, we propose Latent-GRPO, a framework that derives intrinsic rewards directly from latent space geometry. Crucially, our empirical analysis reveals a compelling geometric property: terminal token representations of correct reasoning trajectories form dense clusters with high intra-class similarity, whereas incorrect trajectories remain scattered as outliers. In light of this discovery, we introduce the Iterative Robust Centroid Estimation (IRCE) algorithm, which generates dense, continuous rewards by mitigating magnitude fluctuations via spherical projection and estimating a robust ``truth centroid'' through iterative aggregation. Experimental results on multiple datasets show that our method maintains model performance while achieving a training speedup of over 2x compared to baselines. Furthermore, extensive results demonstrate strong generalization ability and robustness. The code will be released soon.

cs.CL↗

arXiv2026

Biologically Plausible Learning via Bidirectional Spike-Based Distillation

Developing biologically plausible learning algorithms that can achieve performance comparable to error backpropagation remains a longstanding challenge. Existing approaches often compromise biological plausibility by entirely avoiding the use of spikes for error propagation or relying on both positive and negative learning signals, while the question of how spikes can represent negative values remains unresolved. To address these limitations, we introduce Bidirectional Spike-based Distillation (BSD), a novel learning algorithm that jointly trains a feedforward and a backward spiking network. We formulate learning as a transformation between two spiking representations (i.e., stimulus encoding and concept encoding) so that the feedforward network implements perception and decision-making by mapping stimuli to actions, while the backward network supports memory recall by reconstructing stimuli from concept representations. Extensive experiments on diverse benchmarks, including image recognition, image generation, and sequential regression, show that BSD achieves performance comparable to networks trained with classical error backpropagation. These findings represent a significant step toward biologically grounded, spike-driven learning in neural networks. Our code is available at https://github.com/alden199/Bidirectional-Spike-Based-Distillation.

cs.NE↗

arXiv2026

Consistency Deep Equilibrium Models

Deep Equilibrium Models (DEQs) have emerged as a powerful paradigm in deep learning, offering the ability to model infinite-depth networks with constant memory usage. However, DEQs incur significant inference latency due to the iterative nature of fixed-point solvers. In this work, we introduce the Consistency Deep Equilibrium Model (C-DEQ), a novel framework that leverages consistency distillation to accelerate DEQ inference. We cast the DEQ iterative inference process as evolution along a fixed ODE trajectory toward the equilibrium. Along this trajectory, we train C-DEQs to consistently map intermediate states directly to the fixed point, enabling few-step inference while preserving the performance of the teacher DEQ. At the same time, it facilitates multi-step evaluation to flexibly trade computation for performance gains. Extensive experiments across various domain tasks demonstrate that C-DEQs achieve consistent 2-20$\times$ accuracy improvements over implicit DEQs under the same few-step inference budget. Our code is available at https://github.com/landrarwolf/CDEQ.

cs.LG↗

arXiv2026

Learning Query-Specific Rubrics from Human Preferences for DeepResearch Report Generation

Nowadays, developing reliable DeepResearch-style long-form report generation remains challenging, as training and evaluation lack verifiable reward signals. Accordingly, rubric-based evaluation has become a common practice. However, existing approaches either rely on coarse, pre-defined rubrics that lack sufficient granularity or depend on manually constructed query-specific rubrics that are costly and difficult to scale. In this paper, we propose a pipeline to train preference-grounded query-specific rubric generators tailored for DeepResearch report generation. We first construct a dataset of DeepResearch-style queries annotated with human preferences over paired reports, and train rubric generators via reinforcement learning with a hybrid reward combining preference consistency, format validity, and LLM-based rubric evaluation. We evaluate the resulting rubric generators in two stages. First, on a held-out human-preference test set, the learned rubrics discriminate preferred from rejected reports more effectively than generic, prompted, or SFT-trained rubric alternatives. Second, when used as reward signals to train DeepResearch systems, our rubric generators yield substantial performance gains under both a simple single-agent ReAct framework and a complex multi-agent workflow on the DeepResearch Bench.

cs.CL↗

arXiv2026

TRIP-Bench: A Benchmark for Long-Horizon Interactive Agents in Real-World Scenarios

As LLM-based agents are deployed in increasingly complex real-world settings, existing benchmarks underrepresent key challenges such as enforcing global constraints, coordinating multi-tool reasoning, and adapting to evolving user behavior over long, multi-turn interactions. To bridge this gap, we introduce \textbf{TRIP-Bench}, a long-horizon benchmark grounded in realistic travel-planning scenarios. TRIP-Bench leverages real-world data, offers 18 curated tools and 40+ travel requirements, and supports automated evaluation. It includes splits of varying difficulty; the hard split emphasizes long and ambiguous interactions, style shifts, feasibility changes, and iterative version revision. Dialogues span up to 15 user turns, can involve 150+ tool calls, and may exceed 200k tokens of context. Experiments show that even advanced models achieve at most 50\% success on the easy split, with performance dropping below 10\% on hard subsets. We further propose \textbf{GTPO}, an online multi-turn reinforcement learning method with specialized reward normalization and reward differencing. Applied to Qwen2.5-32B-Instruct, GTPO improves constraint satisfaction and interaction robustness, outperforming Gemini-3-Pro in our evaluation. We expect TRIP-Bench to advance practical long-horizon interactive agents, and GTPO to provide an effective online RL recipe for robust long-horizon training.

cs.AI↗

arXiv2026

BatCoder: Self-Supervised Bidirectional Code-Documentation Learning via Back-Translation

Training LLMs for code-related tasks typically depends on high-quality code-documentation pairs, which are costly to curate and often scarce for niche programming languages. We introduce BatCoder, a self-supervised reinforcement learning framework designed to jointly optimize code generation and documentation production. BatCoder employs a back-translation strategy: a documentation is first generated from code, and then the generated documentation is used to reconstruct the original code. The semantic similarity between the original and reconstructed code serves as an implicit reward, enabling reinforcement learning to improve the model's performance both in generating code from documentation and vice versa. This approach allows models to be trained using only code, substantially increasing the available training examples. Evaluated on HumanEval and MBPP with a 7B model, BatCoder achieved 83.5% and 81.0% pass@1, outperforming strong open-source baselines. Moreover, the framework demonstrates consistent scaling with respect to both training corpus size and model capacity.

cs.LG↗

arXiv2026

LTS-VoiceAgent: A Listen-Think-Speak Framework for Efficient Streaming Voice Interaction via Semantic Triggering and Incremental Reasoning

Real-time voice agents face a dilemma: end-to-end models often lack deep reasoning, while cascaded pipelines incur high latency by executing ASR, LLM reasoning, and TTS strictly in sequence, unlike human conversation where listeners often start thinking before the speaker finishes. Since cascaded architectures remain the dominant choice for complex tasks, existing cascaded streaming strategies attempt to reduce this latency via mechanical segmentation (e.g., fixed chunks, VAD-based splitting) or speculative generation, but they frequently either break semantic units or waste computation on predictions that must be rolled back. To address these challenges, we propose LTS-VoiceAgent, a Listen-Think-Speak framework that explicitly separates when to think from how to reason incrementally. It features a Dynamic Semantic Trigger to detect meaningful prefixes, and a Dual-Role Stream Orchestrator that coordinates a background Thinker (for state maintenance) and a foreground Speaker (for speculative solving). This parallel design enables "thinking while speaking" without blocking responses. We also introduce a Pause-and-Repair benchmark containing natural disfluencies to stress-test streaming robustness. Experiments across VERA, Spoken-MQA, BigBenchAudio, and our benchmark show that LTS-VoiceAgent achieves a stronger accuracy-latency-efficiency trade-off than serial cascaded baselines and existing streaming strategies.

cs.SD↗

arXiv2026

CSSG: Measuring Code Similarity with Semantic Graphs

Existing code similarity metrics, such as BLEU, CodeBLEU, and TSED, largely rely on surface-level string overlap or abstract syntax tree structures, and often fail to capture deeper semantic relationships between programs.We propose CSSG (Code Similarity using Semantic Graphs), a novel metric that leverages program dependence graphs to explicitly model control dependencies and variable interactions, providing a semantics-aware representation of code.Experiments on the CodeContests+ dataset show that CSSG consistently outperforms existing metrics in distinguishing more similar code from less similar code under both monolingual and cross-lingual settings, demonstrating that dependency-aware graph representations offer a more effective alternative to surface-level or syntax-based similarity measures.

cs.PL↗

arXiv2026

UserLM-R1: Modeling Human Reasoning in User Language Models with Multi-Reward Reinforcement Learning

User simulators serve as the critical interactive environment for agent post-training, and an ideal user simulator generalizes across domains and proactively engages in negotiation by challenging or bargaining. However, current methods exhibit two issues. They rely on static and context-unaware profiles, necessitating extensive manual redesign for new scenarios, thus limiting generalizability. Moreover, they neglect human strategic thinking, leading to vulnerability to agent manipulation. To address these issues, we propose UserLM-R1, a novel user language model with reasoning capability. Specifically, we first construct comprehensive user profiles with both static roles and dynamic scenario-specific goals for adaptation to diverse scenarios. Then, we propose a goal-driven decision-making policy to generate high-quality rationales before producing responses, and further refine the reasoning and improve strategic capabilities with supervised fine-tuning and multi-reward reinforcement learning. Extensive experimental results demonstrate that UserLM-R1 outperforms competitive baselines, particularly on the more challenging adversarial set.

cs.CL↗

arXiv2026

Controllable Memory Usage: Balancing Anchoring and Innovation in Long-Term Human-Agent Interaction

As LLM-based agents are increasingly used in long-term interactions, cumulative memory is critical for enabling personalization and maintaining stylistic consistency. However, most existing systems adopt an ``all-or-nothing'' approach to memory usage: incorporating all relevant past information can lead to \textit{Memory Anchoring}, where the agent is trapped by past interactions, while excluding memory entirely results in under-utilization and the loss of important interaction history. We show that an agent's reliance on memory can be modeled as an explicit and user-controllable dimension. We first introduce a behavioral metric of memory dependence to quantify the influence of past interactions on current outputs. We then propose \textbf{Stee}rable \textbf{M}emory Agent, \texttt{SteeM}, a framework that allows users to dynamically regulate memory reliance, ranging from a fresh-start mode that promotes innovation to a high-fidelity mode that closely follows interaction history. Experiments across different scenarios demonstrate that our approach consistently outperforms conventional prompting and rigid memory masking strategies, yielding a more nuanced and effective control for personalized human-agent collaboration.

cs.AI↗

arXiv2026

Benchmark^2: Systematic Evaluation of LLM Benchmarks

The rapid proliferation of benchmarks for evaluating large language models (LLMs) has created an urgent need for systematic methods to assess benchmark quality itself. We propose Benchmark^2, a comprehensive framework comprising three complementary metrics: (1) Cross-Benchmark Ranking Consistency, measuring whether a benchmark produces model rankings aligned with peer benchmarks; (2) Discriminability Score, quantifying a benchmark's ability to differentiate between models; and (3) Capability Alignment Deviation, identifying problematic instances where stronger models fail but weaker models succeed within the same model family. We conduct extensive experiments across 15 benchmarks spanning mathematics, reasoning, and knowledge domains, evaluating 11 LLMs across four model families. Our analysis reveals significant quality variations among existing benchmarks and demonstrates that selective benchmark construction based on our metrics can achieve comparable evaluation performance with substantially reduced test sets.

cs.CL↗

arXiv2025

Enhancing Model Privacy in Federated Learning with Random Masking and Quantization

The primary goal of traditional federated learning is to protect data privacy by enabling distributed edge devices to collaboratively train a shared global model while keeping raw data decentralized at local clients. The rise of large language models (LLMs) has introduced new challenges in distributed systems, as their substantial computational requirements and the need for specialized expertise raise critical concerns about protecting intellectual property (IP). This highlights the need for a federated learning approach that can safeguard both sensitive data and proprietary models. To tackle this challenge, we propose FedQSN, a federated learning approach that leverages random masking to obscure a subnetwork of model parameters and applies quantization to the remaining parameters. Consequently, the server transmits only a privacy-preserving proxy of the global model to clients during each communication round, thus enhancing the model's confidentiality. Experimental results across various models and tasks demonstrate that our approach not only maintains strong model performance in federated learning settings but also achieves enhanced protection of model parameters compared to baseline methods.

cs.LG↗

arXiv2025

Improving RL Exploration for LLM Reasoning through Retrospective Replay

Reinforcement learning (RL) has increasingly become a pivotal technique in the post-training of large language models (LLMs). The effective exploration of the output space is essential for the success of RL. We observe that for complex problems, during the early stages of training, the model exhibits strong exploratory capabilities and can identify promising solution ideas. However, its limited capability at this stage prevents it from successfully solving these problems. The early suppression of these potentially valuable solution ideas by the policy gradient hinders the model's ability to revisit and re-explore these ideas later. Consequently, although the LLM's capabilities improve in the later stages of training, it still struggles to effectively address these complex problems. To address this exploration issue, we propose a novel algorithm named Retrospective Replay-based Reinforcement Learning (RRL), which introduces a dynamic replay mechanism throughout the training process. RRL enables the model to revisit promising states identified in the early stages, thereby improving its efficiency and effectiveness in exploration. To evaluate the effectiveness of RRL, we conduct extensive experiments on complex reasoning tasks, including mathematical reasoning and code generation, and general dialogue tasks. The results indicate that RRL maintains high exploration efficiency throughout the training period, significantly enhancing the effectiveness of RL in optimizing LLMs for complicated reasoning tasks. Moreover, it also improves the performance of RLHF, making the model both safer and more helpful.

cs.LG↗

arXiv2025

Dendritic Localized Learning: Toward Biologically Plausible Algorithm

Backpropagation is the foundational algorithm for training neural networks and a key driver of deep learning's success. However, its biological plausibility has been challenged due to three primary limitations: weight symmetry, reliance on global error signals, and the dual-phase nature of training, as highlighted by the existing literature. Although various alternative learning approaches have been proposed to address these issues, most either fail to satisfy all three criteria simultaneously or yield suboptimal results. Inspired by the dynamics and plasticity of pyramidal neurons, we propose Dendritic Localized Learning (DLL), a novel learning algorithm designed to overcome these challenges. Extensive empirical experiments demonstrate that DLL satisfies all three criteria of biological plausibility while achieving state-of-the-art performance among algorithms that meet these requirements. Furthermore, DLL exhibits strong generalization across a range of architectures, including MLPs, CNNs, and RNNs. These results, benchmarked against existing biologically plausible learning algorithms, offer valuable empirical insights for future research. We hope this study can inspire the development of new biologically plausible algorithms for training multilayer networks and advancing progress in both neuroscience and machine learning. Our code is available at https://github.com/Lvchangze/Dendritic-Localized-Learning.

cs.NE↗

arXiv2025

RECAST: Expanding the Boundaries of LLMs' Complex Instruction Following with Multi-Constraint Data

Large language models (LLMs) are increasingly expected to tackle complex tasks, driven by their expanding applications and users' growing proficiency in crafting sophisticated prompts. However, as the number of explicitly stated requirements increases (particularly more than 10 constraints), LLMs often struggle to accurately follow such complex instructions, which limits their applicability in complex real-world scenarios. To the best of our knowledge, existing datasets do not exceed 10 constraints per instance. To address this challenge, we propose RECAST, an efficient and scalable framework for synthesizing datasets where each example incorporates far more constraints than those in existing benchmarks, aiming to challenge and extend the boundaries of models' ability to follow complex instructions. These constraints are extracted from real-world prompt-response pairs to ensure practical relevance. Using this framework, we construct RECAST-30K, a large-scale, high-quality dataset comprising 30k instances spanning 19 constraint types. Experimental results demonstrate that models finetuned on RECAST-30K substantially improve in following complex instructions while maintaining their general capabilities without degradation. Moreover, RECAST enables automatic verification of constraint satisfaction via rule-based validators for quantitative constraints and LLM-based validators for qualitative ones; the verifiability provided by RECAST enables the design of reward functions for reinforcement learning, which further boosts model performance on complex and challenging tasks.

cs.AI↗

arXiv2025

NTIRE 2025 Challenge on Short-form UGC Video Quality Assessment and Enhancement: Methods and Results

This paper presents a review for the NTIRE 2025 Challenge on Short-form UGC Video Quality Assessment and Enhancement. The challenge comprises two tracks: (i) Efficient Video Quality Assessment (KVQ), and (ii) Diffusion-based Image Super-Resolution (KwaiSR). Track 1 aims to advance the development of lightweight and efficient video quality assessment (VQA) models, with an emphasis on eliminating reliance on model ensembles, redundant weights, and other computationally expensive components in the previous IQA/VQA competitions. Track 2 introduces a new short-form UGC dataset tailored for single image super-resolution, i.e., the KwaiSR dataset. It consists of 1,800 synthetically generated S-UGC image pairs and 1,900 real-world S-UGC images, which are split into training, validation, and test sets using a ratio of 8:1:1. The primary objective of the challenge is to drive research that benefits the user experience of short-form UGC platforms such as Kwai and TikTok. This challenge attracted 266 participants and received 18 valid final submissions with corresponding fact sheets, significantly contributing to the progress of short-form UGC VQA and image superresolution. The project is publicly available at https://github.com/lixinustc/KVQE- ChallengeCVPR-NTIRE2025.

eess.IV↗