arXiv subjects

Zheng Pan

Publications and source records attributed to Zheng Pan.

At least 19 recordsLinked to original sources

arXiv2026

SleepWalking: Privileged Representation Shaping for End-to-End Blind Locomotion in Legged Robots

Partially observable locomotion requires a policy to act when task-relevant properties of the robot--environment state are not fully specified by instantaneous observations. Existing approaches often address this challenge by explicitly estimating missing physical variables or processing extended observation histories through structured architectures. We take a different view: partial observability is fundamentally an information-retention problem. The decisive question is not how task-relevant information enters the network, but whether the policy's internal state retains it. Guided by this perspective, we propose SleepWalking for Robot Locomotion (SWAQ), a one-stage end-to-end framework that uses next-step privileged physical reconstruction to shape what a recurrent history representation retains during policy learning, while the deployed actor uses only a direct history-to-action pathway. Under aligned training settings, SWAQ achieves a 15.0\% higher peak mean terrain level than DWAQ, the strongest non-exteroceptive baseline, while using 44.4\% fewer inference MACs per control step. Layerwise probes further show that information associated with the reconstructed physical variables remains linearly decodable through the policy head up to the layer preceding the action output. Complementary theoretical analysis relates privileged-variable recoverability to the achievable-return gap between history-based and privileged-information policy classes. These results suggest that semantic objectives can structure learning without requiring a corresponding architectural decomposition of the deployed controller.

cs.RO↗

arXiv2026

SLIDER: Sparse History-Guided Aerial Robot Target Search using Sliding Local Maps

Efficient exploration and target search in large-scale unknown environments remain challenging for aerial robots due to the demands of broad spatial coverage, fine-grained perception, and real-time decision-making. This paper presents SLIDER, a lightweight and memory-efficient framework that avoids reliance on globally dense maps by combining a local sliding map with sparse global history information. A novel observation quality evaluation method is proposed, leveraging historical poses and sensor models to assess point cloud data in real-time, enabling efficient frontier detection. To support scalable and responsive planning, an incremental viewpoint clustering strategy dynamically adapts to local updates, significantly reducing the number of candidate targets and decreasing computational load. A sparse global topological map is incrementally maintained to assist global planning and cost evaluation. Extensive simulations and real-world experiments demonstrate that the proposed system outperforms state-of-the-art methods in memory usage, decision latency, and search efficiency.

cs.RO↗

arXiv2026

Multi-Stakeholder LLM Alignment: Decomposing Estimation from Aggregation

Multi-stakeholder tasks require one output to satisfy users with conflicting preferences. Holistic LLM judges conflate utility estimation and utility aggregation, yielding unstable implicit weights. We show empirically and theoretically that this aggregation-specific \emph{weighting noise} can create large score shifts when stakeholder satisfaction is dispersed; in our experiments, these weight-induced shifts also increase with stakeholder count. We propose \textsc{DecompR}: counterfactual-calibrated weights are fixed from query structure before candidate scoring, while per-role utilities are estimated independently, removing candidate-dependent weight drift and reducing estimation noise.

cs.AI↗

arXiv2026

GroupTravelBench: Benchmarking LLM Agents on Multi-Person Travel Planning

Travel planning in the real world is overwhelmingly a \textit{group} activity, yet existing LLM travel-planning benchmarks reduce it to a single user, where the field is approaching saturation. This single-user assumption sidesteps what makes group planning hard for an agent: discovering private preferences across multiple users, surfacing conflicts, and balancing utility against fairness. To bring the task back to its multi-user reality, we introduce \textbf{\textit{GroupTravelBench}}, the first benchmark for \textbf{multi-user, multi-turn} travel planning. Built from real user profiles, POI data, and ticket prices, it comprises 650 tasks across three difficulty levels, each running in a synchronous group-chat sandbox with cached tool data for reproducible offline evaluation. Beyond the multi-step reasoning and tool use that single-user benchmarks already test, GroupTravelBench probes three group-specific capabilities: \textit{(i) elicitation} of private preferences through multi-turn dialogue; \textit{(ii) coordination} of inter-user conflicts via compromise or subgrouping; and \textit{(iii) planning} that balances group utility against fairness. We pair this with a complementary evaluation framework combining rule-based outcome metrics and LLM-judge process metrics. Across a wide range of frontier models, even the strongest agents fall short on all four rule-based outcome metrics, with plan validity below 12\%, suggesting that group-level outcome quality is a key open challenge for LLM travel-planning agents.

cs.CL↗

arXiv2026

TRACE: Distilling Where It Matters via Token-Routed Self On-Policy Alignment

On-policy self-distillation (self-OPD) densifies reinforcement learning with verifiable rewards (RLVR) by letting a policy teach itself under privileged context. We find that when this guidance spans the full response, all-token KL spends gradients on mostly redundant positions and amplifies privileged-information leakage, causing entropy rise, shortened reasoning, and out-of-distribution degradation in long-horizon math training. We propose Token-Routed Alignment for Critical rEasoning (TRACE), which distills only on annotator-marked critical spans: forward KL on key spans of correct rollouts, optional reverse KL on localized error spans, and GRPO on all remaining tokens, with the KL channel annealed away after a short warm-up. Our analysis explains TRACE through two effects: forward KL provides non-vanishing lift to teacher-supported tokens that the student under-allocates, while span masking and decay keep cumulative privileged-gradient exposure finite. On four held-out math benchmarks plus GPQA-Diamond, TRACE improves over GRPO by 2.76 percentage points on average and preserves the Qwen3-8B base OOD score on GPQA-Diamond, where GRPO and all-token self-OPD baselines degrade. Gains persist under online self-annotation (+1.90 percentage points, about 69% of the strong-API gain), reducing the concern that TRACE merely imports external annotator capability. Across scales, the best routed action is base-dependent: on Qwen3-8B it is forward KL on key spans, while on Qwen3-1.7B it shifts to reverse KL on error spans.

cs.AI↗

arXiv2026

Aligning Agents via Planning: A Benchmark for Trajectory-Level Reward Modeling

In classical Reinforcement Learning from Human Feedback (RLHF), Reward Models (RMs) serve as the fundamental signal provider for model alignment. As Large Language Models evolve into agentic systems capable of autonomous tool invocation and complex reasoning, the paradigm of reward modeling faces unprecedented challenges -- most notably, the lack of benchmarks specifically designed to assess RM capabilities within tool-integrated environments. To address this gap, we present Plan-RewardBench, a trajectory-level preference benchmark designed to evaluate how well judges distinguish preferred versus distractor agent trajectories in complex tool-using scenarios. Plan-RewardBench covers four representative task families -- (i) Safety Refusal, (ii) Tool-Irrelevance / Unavailability, (iii) Complex Planning, and (iv) Robust Error Recovery -- comprising validated positive trajectories and confusable hard negatives constructed via multi-model natural rollouts, rule-based perturbations, and minimal-edit LLM perturbations. We benchmark representative RMs (generative, discriminative, and LLM-as-Judge) under a unified pairwise protocol, reporting accuracy trends across varying trajectory lengths and task categories. Furthermore, we provide diagnostic analyses of prevalent failure modes. Our results reveal that all three evaluator families face substantial challenges, with performance degrading sharply on long-horizon trajectories, underscoring the necessity for specialized training in agentic, trajectory-level reward modeling. Ultimately, Plan-RewardBench aims to serve as both a practical evaluation suite and a reusable blueprint for constructing agentic planning preference data.

cs.AI↗

arXiv2026

AIDABench: AI Data Analytics Benchmark

As AI-driven document understanding and processing tools become increasingly prevalent in real-world applications, the need for rigorous evaluation standards has grown increasingly urgent. Existing benchmarks and evaluations often focus on isolated capabilities or simplified scenarios, failing to capture the end-to-end task effectiveness required in practical settings. To address this gap, we introduce AIDABench, a comprehensive benchmark for evaluating AI systems on complex data analytics tasks in an end-to-end manner. AIDABench encompasses 600+ diverse document analysis tasks across three core capability dimensions: question answering, data visualization, and file generation. These tasks are grounded in realistic scenarios involving heterogeneous data types, including spreadsheets, databases, financial reports, and operational records, and reflect analytical demands across diverse industries and job functions. Notably, the tasks in AIDABench are sufficiently challenging that even human experts require 1-2 hours per question when assisted by AI tools, underscoring the benchmark's difficulty and real-world complexity. We evaluate 11 state-of-the-art models on AIDABench, spanning both proprietary (e.g., Claude Sonnet 4.5, Gemini 3 Pro Preview) and open-source (e.g., Qwen3-Max-2026-01-23-Thinking) families. Our results reveal that complex, real-world data analytics tasks remain a significant challenge for current AI systems, with the best-performing model achieving only 59.43% pass-at-1. We provide a detailed analysis of failure modes across each capability dimension and identify key challenges for future research. AIDABench offers a principled reference for enterprise procurement, tool selection, and model optimization, and is publicly available at https://github.com/MichaelYang-lyx/AIDABench.

cs.AI↗

arXiv2026

No More Stale Feedback: Co-Evolving Critics for Open-World Agent Learning

Critique-guided reinforcement learning (RL) has emerged as a powerful paradigm for training LLM agents by augmenting sparse outcome rewards with natural-language feedback. However, current methods often rely on static or offline critic models, which fail to adapt as the policy evolves. In on-policy RL, the agent's error patterns shift over time, causing stationary critics to become stale and providing feedback of diminishing utility. To address this, we introduce ECHO (Evolving Critic for Hindsight-Guided Optimization)}, a framework that jointly optimizes the policy and critic through a synchronized co-evolutionary loop. ECHO utilizes a cascaded rollout mechanism where the critic generates multiple diagnoses for an initial trajectory, followed by policy refinement to enable group-structured advantage estimation. We address the challenge of learning plateaus via a saturation-aware gain shaping objective, which rewards the critic for inducing incremental improvements in high-performing trajectories. By employing dual-track GRPO updates, ECHO ensures the critic's feedback stays synchronized with the evolving policy. Experimental results show that ECHO yields more stable training and higher long-horizon task success across open-world environments.

cs.AI↗

arXiv2025

AMAP Agentic Planning Technical Report

We present STAgent, an agentic large language model tailored for spatio-temporal understanding, designed to solve complex tasks such as constrained point-of-interest discovery and itinerary planning. STAgent is a specialized model capable of interacting with ten distinct tools within spatio-temporal scenarios, enabling it to explore, verify, and refine intermediate steps during complex reasoning. Notably, STAgent effectively preserves its general capabilities. We empower STAgent with these capabilities through three key contributions: (1) a stable tool environment that supports over ten domain-specific tools, enabling asynchronous rollout and training; (2) a hierarchical data curation framework that identifies high-quality data like a needle in a haystack, curating high-quality queries by retaining less than 1\% of the raw data, emphasizing both diversity and difficulty; and (3) a cascaded training recipe that starts with a seed SFT stage acting as a guardian to measure query difficulty, followed by a second SFT stage fine-tuned on queries with high certainty, and an ultimate RL stage that leverages data of low certainty. Initialized with Qwen3-30B-A3B to establish a strong SFT foundation and leverage insights into sample difficulty, STAgent yields promising performance on TravelBench while maintaining its general capabilities across a wide range of general benchmarks, thereby demonstrating the effectiveness of our proposed agentic model.

cs.AI↗

arXiv2025

Beyond Itinerary Planning-A Real-World Benchmark for Multi-Turn and Tool-Using Travel Tasks

Travel planning is a natural real-world task to test large language models' (LLMs) planning and tool-use abilities. Although prior work has studied LLM performance on travel planning, existing settings still differ from real-world needs, mainly due to limited domain coverage, insufficient modeling of users' implicit preferences in multi-turn conversations, and a lack of evaluation of agents' capability boundaries. To mitigate these gaps, we propose $\textbf{TravelBench}$, a benchmark for $\textit{truly real-world}$ travel planning. We collect user queries, user preferences, and tools from real scenarios, and construct three subtasks -- $\textit{Single-Turn}$, $\textit{Multi-Turn}$, and $\textit{Unsolvable}$ -- to evaluate agents' three core capabilities in real settings: (1) solving problems independently, (2) interacting with users to elicit implicit preferences, and (3) recognizing the capability boundaries. To enable stable tool invocation and reproducible evaluation, we cache real tool-call results and build a sandbox environment which integrates ten travel-related tools, enabling agents to combine these tools to solve most practical travel planning problems. We evaluate multiple LLMs on TravelBench and find that even advanced models exhibit imbalanced performance across different capabilities. Our further systematic verification demonstrates the stability of the proposed benchmark. TravelBench provides a practical and reproducible benchmark to advance research on LLM agents for real-world travel planning.

cs.AI↗

arXiv2025

Setting the Stage: Text-Driven Scene-Consistent Image Generation

We focus on the foundational task of Scene Staging: given a reference scene image and a text condition specifying an actor category to be generated in the scene and its spatial relation to the scene, the goal is to synthesize an output image that preserves the same scene identity as the reference image while correctly generating the actor according to the spatial relation described in the text. Existing methods struggle with this task, largely due to the scarcity of high-quality paired data and unconstrained generation objectives. To overcome the data bottleneck, we propose a novel data construction pipeline that combines real-world photographs, entity removal, and image-to-video diffusion models to generate training pairs with diverse scenes, viewpoints and correct entity-scene relationships. Furthermore, we introduce a novel correspondence-guided attention loss that leverages cross-view cues to enforce spatial alignment with the reference scene. Experiments on our scene-consistent benchmark show that our approach achieves better scene alignment and text-image alignment than state-of-the-art baselines, according to both automatic metrics and human preference studies. Our method generates images with diverse viewpoints and compositions while faithfully following the textual instructions and preserving the reference scene identity.

cs.CV↗

arXiv2025

Online Navigation Refinement: Achieving Lane-Level Guidance by Associating Standard-Definition and Online Perception Maps

Lane-level navigation is critical for geographic information systems and navigation-based tasks, offering finer-grained guidance than road-level navigation by standard definition (SD) maps. However, it currently relies on expansive global HD maps that cannot adapt to dynamic road conditions. Recently, online perception (OP) maps have become research hotspots, providing real-time geometry as an alternative, but lack the global topology needed for navigation. To address these issues, Online Navigation Refinement (ONR), a new mission is introduced that refines SD-map-based road-level routes into accurate lane-level navigation by associating SD maps with OP maps. The map-to-map association to handle many-to-one lane-to-road mappings under two key challenges: (1) no public dataset provides lane-to-road correspondences; (2) severe misalignment from spatial fluctuations, semantic disparities, and OP map noise invalidates traditional map matching. For these challenges, We contribute: (1) Online map association dataset (OMA), the first ONR benchmark with 30K scenarios and 2.6M annotated lane vectors; (2) MAT, a transformer with path-aware attention to aligns topology despite spatial fluctuations and semantic disparities and spatial attention for integrates noisy OP features via global context; and (3) NR P-R, a metric evaluating geometric and semantic alignment. Experiments show that MAT outperforms existing methods at 34 ms latency, enabling low-cost and up-to-date lane-level navigation.

cs.CV↗

arXiv2025

SeqGrowGraph: Learning Lane Topology as a Chain of Graph Expansions

Accurate lane topology is essential for autonomous driving, yet traditional methods struggle to model the complex, non-linear structures-such as loops and bidirectional lanes-prevalent in real-world road structure. We present SeqGrowGraph, a novel framework that learns lane topology as a chain of graph expansions, inspired by human map-drawing processes. Representing the lane graph as a directed graph $G=(V,E)$, with intersections ($V$) and centerlines ($E$), SeqGrowGraph incrementally constructs this graph by introducing one vertex at a time. At each step, an adjacency matrix ($A$) expands from $n \times n$ to $(n+1) \times (n+1)$ to encode connectivity, while a geometric matrix ($M$) captures centerline shapes as quadratic B\'ezier curves. The graph is serialized into sequences, enabling a transformer model to autoregressively predict the chain of expansions, guided by a depth-first search ordering. Evaluated on nuScenes and Argoverse 2 datasets, SeqGrowGraph achieves state-of-the-art performance.

cs.CV↗

arXiv2025

FutureSightDrive: Thinking Visually with Spatio-Temporal CoT for Autonomous Driving

Vision-Language-Action (VLA) models offer significant potential for end-to-end driving, yet their reasoning is often constrained by textual Chains-of-Thought (CoT). This symbolic compression of visual information creates a modality gap between perception and planning by blurring spatio-temporal relations and discarding fine-grained cues. We introduce FSDrive, a framework that empowers VLAs to "think visually" using a novel visual spatio-temporal CoT. FSDrive first operates as a world model, generating a unified future frame that combines a predicted background with explicit, physically-plausible priors like future lane dividers and 3D object boxes. This imagined scene serves as the visual spatio-temporal CoT, capturing both spatial structure and temporal evolution in a single representation. The same VLA then functions as an inverse-dynamics model to plan trajectories conditioned on current observations and this visual CoT. We enable this with a unified pre-training paradigm that expands the model's vocabulary with visual tokens and jointly optimizes for semantic understanding (VQA) and future-frame prediction. A progressive curriculum first generates structural priors to enforce physical laws before rendering the full scene. Evaluations on nuScenes and NAVSIM show FSDrive improves trajectory accuracy and reduces collisions, while also achieving competitive FID for video generation with a lightweight autoregressive model and advancing scene understanding on DriveLM. These results confirm that our visual spatio-temporal CoT bridges the perception-planning gap, enabling safer, more anticipatory autonomous driving. Code is available at https://github.com/MIV-XJTU/FSDrive.

cs.CV↗

arXiv2024

Driving by the Rules: A Benchmark for Integrating Traffic Sign Regulations into Vectorized HD Map

Ensuring adherence to traffic sign regulations is essential for both human and autonomous vehicle navigation. While current online mapping solutions often prioritize the construction of the geometric and connectivity layers of HD maps, overlooking the construction of the traffic regulation layer within HD maps. Addressing this gap, we introduce MapDR, a novel dataset designed for the extraction of Driving Rules from traffic signs and their association with vectorized, locally perceived HD Maps. MapDR features over $10,000$ annotated video clips that capture the intricate correlation between traffic sign regulations and lanes. Built upon this benchmark and the newly defined task of integrating traffic regulations into online HD maps, we provide modular and end-to-end solutions: VLE-MEE and RuleVLM, offering a strong baseline for advancing autonomous driving technology. It fills a critical gap in the integration of traffic sign rules, contributing to the development of reliable autonomous driving systems. Code is available at https://github.com/MIV-XJTU/MapDR.

cs.CV↗

arXiv2024

PriorDrive: Enhancing Online HD Mapping with Unified Vector Priors

High-Definition Maps (HD maps) are essential for the precise navigation and decision-making of autonomous vehicles, yet their creation and upkeep present significant cost and timeliness challenges. The online construction of HD maps using on-board sensors has emerged as a promising solution; however, these methods can be impeded by incomplete data due to occlusions and inclement weather, while their performance in distant regions remains unsatisfying. This paper proposes PriorDrive to address these limitations by directly harnessing the power of various vectorized prior maps, significantly enhancing the robustness and accuracy of online HD map construction. Our approach integrates a variety of prior maps uniformly, such as OpenStreetMap's Standard Definition Maps (SD maps), outdated HD maps from vendors, and locally constructed maps from historical vehicle data. To effectively integrate such prior information into online mapping models, we introduce a Hybrid Prior Representation (HPQuery) that standardizes the representation of diverse map elements. We further propose a Unified Vector Encoder (UVE), which employs fused prior embedding and a dual encoding mechanism to encode vector data. To improve the UVE's generalizability and performance, we propose a segment-level and point-level pre-training strategy that enables the UVE to learn the prior distribution of vector data. Through extensive testing on the nuScenes, Argoverse 2 and OpenLane-V2, we demonstrate that PriorDrive is highly compatible with various online mapping models and substantially improves map prediction capabilities. The integration of prior maps through PriorDrive offers a robust solution to the challenges of single-perception data, paving the way for more reliable autonomous vehicle navigation. Code is available at https://github.com/MIV-XJTU/PriorDrive.

cs.CV↗

arXiv2023

Adaptive Approximate Implicitization of Planar Parametric Curves via Weak Gradient Constraints

Converting a parametric curve into the implicit form, which is called implicitization, has always been a popular but challenging problem in geometric modeling and related applications. However, the existing methods mostly suffer from the problems of maintaining geometric features and choosing a reasonable implicit degree. The present paper has two contributions. We first introduce a new regularization constraint(called the weak gradient constraint) for both polynomial and non-polynomial curves, which efficiently possesses shape preserving. We then propose two adaptive algorithms of approximate implicitization for polynomial and non-polynomial curves respectively, which find the ``optimal'' implicit degree based on the behavior of the weak gradient constraint. More precisely, the idea is gradually increasing the implicit degree, until there is no obvious improvement in the weak gradient loss of the outputs. Experimental results have shown the effectiveness and high quality of our proposed methods.

cs.GR↗

arXiv2022

Multi-Graph based Multi-Scenario Recommendation in Large-scale Online Video Services

Recently, industrial recommendation services have been boosted by the continual upgrade of deep learning methods. However, they still face de-biasing challenges such as exposure bias and cold-start problem, where circulations of machine learning training on human interaction history leads algorithms to repeatedly suggest exposed items while ignoring less-active ones. Additional problems exist in multi-scenario platforms, e.g. appropriate data fusion from subsidiary scenarios, which we observe could be alleviated through graph structured data integration via message passing. In this paper, we present a multi-graph structured multi-scenario recommendation solution, which encapsulates interaction data across scenarios with multi-graph and obtains representation via graph learning. Extensive offline and online experiments on real-world datasets are conducted where the proposed method demonstrates an increase of 0.63% and 0.71% in CTR and Video Views per capita on new users over deployed set of baselines and outperforms regular method in increasing the number of outer-scenario videos by 25% and video watches by 116%, validating its superiority in activating cold videos and enriching target recommendation.

cs.AI↗