arXiv subjects

Haochen Liu

Publications and source records attributed to Haochen Liu.

At least 37 records · Page 2Linked to original sources

arXiv2026

Discrete-WAM: Unified Discrete Vision-Action Token Editing for World-Policy Learning

Autonomous driving requires reasoning about how ego actions shape future world evolution, rather than merely mapping observations to actions. However, most end-to-end methods rely on direct state-to-action imitation, while existing world models often remain weakly aligned with downstream policy generation. We introduce Discrete-WAM, a unified discrete vision-action world-policy framework that represents visual observations, future states, high-level decisions, and ego actions within a shared token space. Built on this discrete alignment, Discrete-WAM jointly trains world modeling, world-policy modeling, and policy modeling through multi-task and multi-stage pretraining, allowing action-conditioned future prediction to directly support policy generation. For downstream planning, Discrete-WAM further decomposes policy generation into hierarchical decision prediction and parallel action-token editing, where the decision token provides a high-level planning skeleton and confidence-based scheduling refines dense future actions efficiently. Experiments on large-scale autonomous-driving benchmarks show that Discrete-WAM achieves strong planning performance while supporting controllable future generation, counterfactual evaluation, surprise-based world-model analysis, and efficient parallel policy decoding. These results suggest that discrete representation alignment, unified world-policy training, and hierarchical token editing provide a promising design paradigm for physical AI.

cs.RO↗

arXiv2026

MINER: Mining Multimodal Internal Representation for Efficient Retrieval

Visual document retrieval has become essential for accessing information in visually rich documents. Existing approaches fall into two camps. Late-interaction retrievers achieve strong quality through fine-grained token-level matching but store hundreds of vectors per page, incurring large index footprints and high serving costs. By contrast, dense single-vector retrievers retain storage and latency advantages but consistently lag in quality because they compress all information into a single final-layer embedding. In this work, we first conduct a layerwise diagnostic on single-vector retrievers, revealing that retrieval-relevant signal resides in internal representations. Motivated by these findings, we propose MINER (Mining Multimodal Internal RepreseNtation for Efficient Retrieval), a lightweight plug-in module that probes and fuses internal signals across transformer layers into a single compact embedding without modifying the backbone or sacrificing single-vector efficiency. The first Retrieval-Aligned Layer Probing stage attaches a lightweight probe at each layer, surfacing which dimensions carry retrieval-relevant information. The subsequent Adaptive Sparse Multi-Layer Fusion stage applies performance-adaptive neuron-level masking to the selected layers and fuses the surviving signals into the final dense vector. Across ViDoRe V1/V2/V3, MINER outperforms existing dense single-vector retrievers on the majority of benchmarks, with up to 4.5% nDCG@5 improvement over its corresponding backbone. Compared to strong late-interaction baselines, in some settings MINER substantially narrows the nDCG@$5$ gap to $0.2$ while preserving the storage and serving advantages of dense retrieval.

cs.LG↗

arXiv2026

LegalDrill: Diagnosis-Driven Synthesis for Legal Reasoning in Small Language Models

Small language models (SLMs) are promising for real-world deployment due to their efficiency and low operational cost. However, their limited capacity struggles with high-stakes legal reasoning tasks that require coherent statute interpretation and logically consistent deduction. Furthermore, training SLMs for such tasks demands high-quality, concise reasoning trajectories, which are prohibitively expensive to manually collect and difficult to curate via standard rejection sampling, lacking granularity beyond final verdicts. To address these challenges, we propose {LegalDrill}, a diagnosis-driven synthesis framework that extracts and iteratively refines reasoning trajectories from a capable teacher via fine-grained prompting, then a self-reflective verification is employed to adaptively select the most effective data for the SLM student. The resulting data empower SLM training through supervised fine-tuning and direct preference optimization. Extensive experiments on several legal benchmarks demonstrate that {LegalDrill} significantly bolsters the legal reasoning capabilities of representative SLMs while bypassing the need for scarce expert annotations, paving a scalable path toward practical legal reasoning systems.

cs.CL↗

arXiv2026

SimScale: Learning to Drive via Real-World Simulation at Scale

Achieving fully autonomous driving systems requires learning rational decisions in a wide span of scenarios, including safety-critical and out-of-distribution ones. However, such cases are underrepresented in real-world corpus collected by human experts. To complement for the lack of data diversity, we introduce a novel and scalable simulation framework capable of synthesizing massive unseen states upon existing driving logs. Our pipeline utilizes advanced neural rendering with a reactive environment to generate high-fidelity multi-view observations controlled by the perturbed ego trajectory. Furthermore, we develop a pseudo-expert trajectory generation mechanism for these newly simulated states to provide action supervision. Upon the synthesized data, we find that a simple co-training strategy on both real-world and simulated samples can lead to significant improvements in both robustness and generalization for various planning methods on challenging real-world benchmarks, up to +8.6 EPDMS on navhard and +2.9 on navtest. More importantly, such policy improvement scales smoothly by increasing simulation data only, even without extra real-world data streaming in. We further reveal several crucial findings of such a sim-real learning system, which we term SimScale, including the design of pseudo-experts and the scaling properties for different policy architectures. Simulation data and code have been released at https://github.com/OpenDriveLab/SimScale.

cs.CV↗

arXiv2026

KAT-Coder-V2 Technical Report

We present KAT-Coder-V2, an agentic coding model developed by the KwaiKAT team at Kuaishou. KAT-Coder-V2 adopts a "Specialize-then-Unify" paradigm that decomposes agentic coding into five expert domains - SWE, WebCoding, Terminal, WebSearch, and General - each undergoing independent supervised fine-tuning and reinforcement learning, before being consolidated into a single model via on-policy distillation. We develop KwaiEnv, a modular infrastructure sustaining tens of thousands of concurrent sandbox instances, and scale RL training along task complexity, intent alignment, and scaffold generalization. We further propose MCLA for stabilizing MoE RL training and Tree Training for eliminating redundant computation over tree-structured trajectories with up to 6.2x speedup. KAT-Coder-V2 achieves 79.6% on SWE-bench Verified (vs. Claude Opus 4.6 at 80.8%), 88.7 on PinchBench (surpassing GLM-5 and MiniMax M2.7), ranks first across all three frontend aesthetics scenarios, and maintains strong generalist scores on Terminal-Bench Hard (46.8) and tau^2-Bench (93.9). Our model is publicly available at https://streamlake.com/product/kat-coder.

cs.CL↗

arXiv2026

A finite-element Delta-Sternheimer approach for computing accurate all-electron RPA correlation energies of polyatomic molecules

Attaining a reliable complete basis set (CBS) limit remains a significant challenge in ab initio correlated electronic-structure calculations. Building on our previous work for atoms and diatomic molecules, we present a finite-element (FE) Delta Sternheimer approach for numerically accurate random phase approximation (RPA) calculations applicable to general molecules. This approach seamlessly integrates atomic orbital basis sets with FE grids, enabling an arbitrary precision representation of first order wavefunctions. As a result, the density response function and RPA correlation energies can be computed with fully controlled numerical precision. The Delta Sternheimer approach thus provides direct access to RPA correlation energies at the CBS limit, eliminating reliance on conventional extrapolation schemes. We apply this approach to two problems: The energy hierarchy of 20 water-dimer configurations and the atomization energies of 50 molecules from the G2 set. For the water dimer, we examine the basis set dependence of the isomer energy ordering. For the G2 set, we investigate the residual numerical uncertainty in the conventional extrapolated CBS limit, both with and without correction for basis-set superposition error (BSSE).

cond-mat.mtrl-sci↗

arXiv2026

DecompGAIL: Learning Realistic Traffic Behaviors with Decomposed Multi-Agent Generative Adversarial Imitation Learning

Realistic traffic simulation is critical for the development of autonomous driving systems and urban mobility planning, yet existing imitation learning approaches often fail to model realistic traffic behaviors. Behavior cloning suffers from covariate shift, while Generative Adversarial Imitation Learning (GAIL) is notoriously unstable in multi-agent settings. We identify a key source of this instability: irrelevant interaction misguidance, where a discriminator penalizes an ego vehicle's realistic behavior due to unrealistic interactions among its neighbors. To address this, we propose Decomposed Multi-agent GAIL (DecompGAIL), which explicitly decomposes realism into ego-map and ego-neighbor components, filtering out misleading neighbor: neighbor and neighbor: map interactions. We further introduce a social PPO objective that augments ego rewards with distance-weighted neighborhood rewards, encouraging overall realism across agents. Integrated into a lightweight SMART-based backbone, DecompGAIL achieves state-of-the-art performance on the WOMD Sim Agents 2025 benchmark.

cs.LG↗

arXiv2025

Continuous Finite Element Method For Maxwell Eigenvalue Problems With Regular Decomposition Technique

With the regular decomposition technique, we decompose the space $\mathbf{H}_0^s(\mathbf{curl}; Ω)$ into the sum of a vector potential space and the gradient of a scalar space, both possessing higher regularity. Based on this new high order regular decomposition, a novel numerical method using standard high order Lagrange finite elements is designed for solving Maxwell eigenvalue problems. Specifically, the full convergence orders of the eigenpair approximations are proved for the proposed numerical method. Finally, numerical examples are provided to validate the proposed scheme and confirm the theoretical convergence results.

math.NA↗

arXiv2025

Representations of 3D Rotations: Mathematical Foundations and Comparative Analysis

Rotation representations are foundational in fields such as computer graphics, robotics, and machine learning, where precise and efficient modeling of 3D orientations is critical. This paper comprehensively investigates diverse representations of the special orthogonal group $SO(3)$, such as Euler angles, axis-angle vectors, quaternions, rotation matrices, exponential maps, and emerging continuous and probabilistic methods, evaluating their mathematical formulations, continuity, susceptibility to gimbal lock, computational efficiency, storage requirements, interpolation properties, and composition operations, while integrating detailed algebraic insights with practical applications in fields like animation, pose estimation, inertial navigation, 3D shape registration, and neural networks. Empirical evidence highlights quaternions' dominance due to their compactness and computational efficiency, while alternatives like 6D continuous representations and matrix Fisher distributions provide enhanced continuity and uncertainty modeling. Future research could explore hybrid methods and thorough large-scale evaluations to help build a solid foundation for improving rotation representation techniques.

cs.GR↗

arXiv2025

NeRC: Neural Ranging Correction through Differentiable Moving Horizon Location Estimation

GNSS localization using everyday mobile devices is challenging in urban environments, as ranging errors caused by the complex propagation of satellite signals and low-quality onboard GNSS hardware are blamed for undermining positioning accuracy. Researchers have pinned their hopes on data-driven methods to regress such ranging errors from raw measurements. However, the grueling annotation of ranging errors impedes their pace. This paper presents a robust end-to-end Neural Ranging Correction (NeRC) framework, where localization-related metrics serve as the task objective for training the neural modules. Instead of seeking impractical ranging error labels, we train the neural network using ground-truth locations that are relatively easy to obtain. This functionality is supported by differentiable moving horizon location estimation (MHE) that handles a horizon of measurements for positioning and backpropagates the gradients for training. Even better, as a blessing of end-to-end learning, we propose a new training paradigm using Euclidean Distance Field (EDF) cost maps, which alleviates the demands on labeled locations. We evaluate the proposed NeRC on public benchmarks and our collected datasets, demonstrating its distinguished improvement in positioning accuracy. We also deploy NeRC on the edge to verify its real-time performance for mobile devices.

cs.LG↗

arXiv2025

Cooperative Base Station Assignment and Resource Allocation for 6G ISAC Network

In the upcoming 6G networks, integrated sensing and communications (ISAC) will be able to provide a performance boost in both perception and wireless connectivity. This paper considers a multiple base station (BS) architecture to support the comprehensive services of data transmission and multi-target sensing. In this context, a cooperative BS assignment and resource allocation (CBARA) strategy is proposed in this paper, aiming at jointly optimizing the communication and sensing (C&S) performance. The posterior Cramer-Rao lower bound and the achievable rate with respect to transmit power and bandwidth are derived and utilized as optimization criteria for the CBARA scheme. We develop a heuristic alternating optimization algorithm to obtain an effective sub-optimal solution for the non-convex optimization problem caused by multiple coupled variables. Numerical results show the effectiveness of the proposed solution, which achieves a performance improvement of 117% in communication rate and 40% in sensing accuracy, compared to the classic scheme.

cs.IT↗

arXiv2025

A Transformer-Based Conditional GAN with Multiple Instance Learning for UAV Signal Detection and Classification

Unmanned Aerial Vehicles (UAVs) are increasingly used in surveillance, logistics, agriculture, disaster management, and military operations. Accurate detection and classification of UAV flight states, such as hovering, cruising, ascending, or transitioning, which are essential for safe and effective operations. However, conventional time series classification (TSC) methods often lack robustness and generalization for dynamic UAV environments, while state of the art(SOTA) models like Transformers and LSTM based architectures typically require large datasets and entail high computational costs, especially with high-dimensional data streams. This paper proposes a novel framework that integrates a Transformer-based Generative Adversarial Network (GAN) with Multiple Instance Locally Explainable Learning (MILET) to address these challenges in UAV flight state classification. The Transformer encoder captures long-range temporal dependencies and complex telemetry dynamics, while the GAN module augments limited datasets with realistic synthetic samples. MIL is incorporated to focus attention on the most discriminative input segments, reducing noise and computational overhead. Experimental results show that the proposed method achieves superior accuracy 96.5% on the DroneDetect dataset and 98.6% on the DroneRF dataset that outperforming other SOTA approaches. The framework also demonstrates strong computational efficiency and robust generalization across diverse UAV platforms and flight states, highlighting its potential for real-time deployment in resource constrained environments.

cs.LG↗

arXiv2025

PASE: A Massively Parallel Augmented Subspace Eigensolver for Large Scale Eigenvalue Problems

In this paper, we present a novel parallel augmented subspace method and build a package Parallel Augmented Subspace Eigensolver (PASE) for solving large scale eigenvalue problems by the massively parallel finite element discretization. Based on the augmented subspace, solving high dimensional eigenvalue problems can be transformed to solving the corresponding linear equations and low dimensional eigenvalue problems on the augmented subspace. Thus the complexity of solving the eigenvalue problems by augmented subspace method will be comparable to that of solving the same dimensinal linear equations. In order to improve the scalability and efficiency, we also present some implementing techniques for the parallel augmented subspace method. Based on parallel augmented subspace method and the concerned implementing techniques, a package PASE is built for solving large scale eigenvalue problems. Some numerical examples are provided to validate the efficiency and scalability of the proposed numerical methods.

math.NA↗

arXiv2025

Reinforced Refinement with Self-Aware Expansion for End-to-End Autonomous Driving

End-to-end autonomous driving has emerged as a promising paradigm for directly mapping sensor inputs to planning maneuvers using learning-based modular integrations. However, existing imitation learning (IL)-based models suffer from generalization to hard cases, and a lack of corrective feedback loop under post-deployment. While reinforcement learning (RL) offers a potential solution to tackle hard cases with optimality, it is often hindered by overfitting to specific driving cases, resulting in catastrophic forgetting of generalizable knowledge and sample inefficiency. To overcome these challenges, we propose Reinforced Refinement with Self-aware Expansion (R2SE), a novel learning pipeline that constantly refines hard domain while keeping generalizable driving policy for model-agnostic end-to-end driving systems. Through reinforcement fine-tuning and policy expansion that facilitates continuous improvement, R2SE features three key components: 1) Generalist Pretraining with hard-case allocation trains a generalist imitation learning (IL) driving system while dynamically identifying failure-prone cases for targeted refinement; 2) Residual Reinforced Specialist Fine-tuning optimizes residual corrections using reinforcement learning (RL) to improve performance in hard case domain while preserving global driving knowledge; 3) Self-aware Adapter Expansion dynamically integrates specialist policies back into the generalist model, enhancing continuous performance improvement. Experimental results in closed-loop simulation and real-world datasets demonstrate improvements in generalization, safety, and long-horizon policy robustness over state-of-the-art E2E systems, highlighting the effectiveness of reinforce refinement for scalable autonomous driving.

cs.RO↗

arXiv2025

UAV-Enabled Joint Sensing, Communication, Powering and Backhaul Transmission in Maritime Monitoring Networks

This paper addresses the challenge of energy-constrained maritime monitoring networks by proposing an unmanned aerial vehicle (UAV)-enabled integrated sensing, communication, powering and backhaul transmission scheme with a tailored time-division duplex frame structure. Within each time slot, the UAV sequentially implements sensing, wireless charging and uplink receiving with buoys, and lastly forwards part of collected data to the central ship via backhaul links. Considering the tight coupling among these functions, we jointly optimize time allocation, UAV trajectory, UAV-buoy association, and power scheduling to maximize the performance of data collection, with the practical consideration of sea clutter effects during UAV sensing. A novel optimization framework combining alternating optimization, quadratic transform and augmented first-order Taylor approximation is developed, which demonstrates good convergence behavior and robustness. Simulation results show that under sensing quality-of-service constraint, buoys are able to achieve an average data rate over 22bps/Hz using around 2mW harvested power per active time slot, validating the scheme's effectiveness for open-sea monitoring. Additionally, it is found that under the influence of sea clutters, the optimal UAV trajectory always keeps a certain distance with buoys to strike a balance between sensing and other multi-functional transmissions.

eess.SP↗

arXiv2025

iPad: Iterative Proposal-centric End-to-End Autonomous Driving

End-to-end (E2E) autonomous driving systems offer a promising alternative to traditional modular pipelines by reducing information loss and error accumulation, with significant potential to enhance both mobility and safety. However, most existing E2E approaches directly generate plans based on dense bird's-eye view (BEV) grid features, leading to inefficiency and limited planning awareness. To address these limitations, we propose iterative Proposal-centric autonomous driving (iPad), a novel framework that places proposals - a set of candidate future plans - at the center of feature extraction and auxiliary tasks. Central to iPad is ProFormer, a BEV encoder that iteratively refines proposals and their associated features through proposal-anchored attention, effectively fusing multi-view image data. Additionally, we introduce two lightweight, proposal-centric auxiliary tasks - mapping and prediction - that improve planning quality with minimal computational overhead. Extensive experiments on the NAVSIM and CARLA Bench2Drive benchmarks demonstrate that iPad achieves state-of-the-art performance while being significantly more efficient than prior leading methods.

cs.CV↗

arXiv2025

Question-Aware Knowledge Graph Prompting for Enhancing Large Language Models

Large Language Models (LLMs) often struggle with tasks requiring external knowledge, such as knowledge-intensive Multiple Choice Question Answering (MCQA). Integrating Knowledge Graphs (KGs) can enhance reasoning; however, existing methods typically demand costly fine-tuning or retrieve noisy KG information. Recent approaches leverage Graph Neural Networks (GNNs) to generate KG-based input embedding prefixes as soft prompts for LLMs but fail to account for question relevance, resulting in noisy prompts. Moreover, in MCQA tasks, the absence of relevant KG knowledge for certain answer options remains a significant challenge. To address these issues, we propose Question-Aware Knowledge Graph Prompting (QAP), which incorporates question embeddings into GNN aggregation to dynamically assess KG relevance. QAP employs global attention to capture inter-option relationships, enriching soft prompts with inferred knowledge. Experimental results demonstrate that QAP outperforms state-of-the-art methods across multiple datasets, highlighting its effectiveness.

cs.CL↗

arXiv2025

KG-CF: Knowledge Graph Completion with Context Filtering under the Guidance of Large Language Models

Large Language Models (LLMs) have shown impressive performance in various tasks, including knowledge graph completion (KGC). However, current studies mostly apply LLMs to classification tasks, like identifying missing triplets, rather than ranking-based tasks, where the model ranks candidate entities based on plausibility. This focus limits the practical use of LLMs in KGC, as real-world applications prioritize highly plausible triplets. Additionally, while graph paths can help infer the existence of missing triplets and improve completion accuracy, they often contain redundant information. To address these issues, we propose KG-CF, a framework tailored for ranking-based KGC tasks. KG-CF leverages LLMs' reasoning abilities to filter out irrelevant contexts, achieving superior results on real-world datasets. The code and datasets are available at \url{https://anonymous.4open.science/r/KG-CF}.

cs.AI↗