Searcharxiv⌕ Search

arXiv subjects

Xiao Yang

Publications and source records attributed to Xiao Yang.

At least 37 records · Page 2Linked to original sources
arXiv2026

PUFFER: Incremental Fuzzy Deduplication for Continuously Evolving Corpora

Large language model training corpora grow through successive, often redundant releases, so each release must be deduplicated against both itself and the accumulated history. At trillion-token scale, this requires incremental ingestion, bounded resident memory, deterministic retry, and dataset-scoped lifecycle control without repeated corpus-wide rebuilding. We introduce PUFFER (Provenance-aware Updatable Fuzzy Filtering for Evolving Repositories), a MinHash-LSH fuzzy-deduplication pipeline built around two design choices. First, PUFFER stores each LSH band as immutable, dataset-tagged, memory-mapped sorted segments, enabling exact historical band-key membership checks without RAM proportional to corpus size. Second, T-fanout tiered compaction periodically merges segments to control screening fanout, trading lower query cost against additional index-maintenance writes while preserving membership decisions. Across N ingested keys and K equal-sized releases, PUFFER's cumulative maintenance cost is O(N log N log_T K), compared with Theta(KN) for repeated snapshot rebuilding. Dataset-tagged segments also support dataset-scoped withdrawal: removal is constant-time for uncompacted or protected datasets, while post-compaction withdrawal reconstructs only the affected merged segment, even if the original dataset is unavailable. In our implementation, PUFFER completed cumulative index-stage ingestion for one billion documents in about 1.75 hours in a single process, using 128 bytes per document for a 16-band index. A classical resident MinHash-LSH table required about 6.5 KB per document and exceeded a 900 GiB RAM cap. In a ten-hour comparison capped at one billion documents, PUFFER was 11x faster than LSHBloom and 35x faster than Milvus-LSH. PUFFER is deployed on more than 30 billion documents, and we release it as open-source software at https://github.com/Zyphra/puffer.

cs.DB↗
arXiv2026

DecoEvo: Score-Decoupled Co-Evolution of Solver and Rubric-Generator Skills in Text Space

Text-space optimization adapts large language models (LLMs) by editing external natural-language artifacts rather than model weights, so the optimized artifacts remain inspectable and the model can be treated as a black box. However, most existing text-space methods keep evaluation fixed. On open-ended tasks, this can become a bottleneck: once the solver improves on the criteria a rubric measures, omitted dimensions remain invisible to the optimization signal. Simply evolving the rubric is also unreliable when updates are selected by the current solver's score, because apparent progress can come from making the rubric easier to satisfy. We introduce DecoEvo (Decoupled Co-Evolution), which co-evolves a solver skill and a rubric-generator skill under decoupled objectives without using gold rubrics during optimization. The solver skill is updated using criterion-level feedback, while the rubric-generator skill is revised through complementary audits of requirement coverage and response discrimination that are independent of aggregate solver score. This separation focuses generator updates on newly exposed solver weaknesses, reducing repeated emphasis on criteria the solver already satisfies. Under each benchmark's official evaluation, DecoEvo outperforms all compared methods across five benchmarks and three LLM backbones, yielding 2.8--5.0\% relative gains over SkillOpt in the five-benchmark average.

cs.AI↗
arXiv2026

Are Heterogeneous Graph Neural Networks Truly Effective for Node Classification? A Causal Perspective

Graph neural networks (GNNs) have achieved remarkable success in node classification. Building on this progress, heterogeneous graph neural networks (HGNNs) integrate relation types and node and edge semantics to leverage heterogeneous information. Causal analysis for HGNNs is advancing rapidly, aiming to separate genuine causal effects from spurious correlations. However, whether HGNNs are intrinsically effective for node classification remains underexamined, and most studies implicitly assume rather than establish this effectiveness. In this work, we examine HGNNs for node classification from two perspectives: model architecture and heterogeneous information. We conduct a systematic reproduction across 21 datasets and 20 baselines, complemented by comprehensive hyperparameter retuning. To further disentangle the source of performance gains, we develop a causal mediation analysis framework that treats the introduction of heterogeneous relation information as the treatment, candidate structural properties as mediators, and node classification performance as the outcome. This framework first screens candidate mediators according to their treatment-induced changes and their associations with performance improvement, and then decomposes the total effect into mediated and direct effects. Our results lead to two conclusions. First, model architecture and complexity have no causal effect on node classification performance. Second, heterogeneous information exerts a positive causal effect primarily through increasing homophily and local-global distribution discrepancy, which makes node classes more distinguishable. The implementation is publicly available at https://github.com/YXNTU/CausalHGNN.

cs.LG↗
arXiv2026

RPM-Distill: Physiology-guided Adaptive Cross-modal Distillation for Robust Remote Physiological Measurement

Video-based remote physiological measurement (RPM) is highly accessible but remains fragile under varying illumination, skin tones, and motion. Radio frequency (RF) radar is largely invariant to illumination and appearance, providing complementary cardio-respiratory micro-motion cues; however, requiring radar at inference is often impractical due to its limited ubiquity and deployment overhead. We propose RPM-Distill, a physiology-guided cross-modal distillation framework that leverages synchronized radar only during training while retaining video-only inference. Our key observation is that although RGB and RF waveforms differ in sensing physics and time-domain morphology, they share similar latent periodic rhythm in the frequency domain. We thus distill physiology-structured spectral evidence to improve robustness, via losses that (i) anchor the fundamental peak, (ii) match the off-peak background distribution, and (iii) preserve spectral morphology and sharpness. To avoid negative transfer under sample-level teacher quality and alignment uncertainty, a spectral policy network predicts sample-level distillation gates and component weights from the student--teacher spectral relation map, learned with a meta bilevel objective on a small labeled validation split. Through extensive experiments in challenging conditions and cross-dataset settings, RPM-Distill brings 81\% MAE and 21\% correlation improvement over unimodal baselines. Code is at https://github.com/WJULYW/RPM-Distill.

cs.CV↗
arXiv2026

Self-Evolving Just-In-Time Memory for Proactive Embodied Safety

While Vision-Language Models (VLMs) have empowered embodied agents to execute complex household tasks, they struggle to proactively handle dynamically emerging hazards during closed-loop interactions. Existing safety approaches often rely on runtime guardrails to block unsafe actions or induce excessive caution, which severely stalls task progress instead of actively resolving the underlying risks. To break this safety-progress trade-off, we introduce the Self-Evolving Just-In-Time Memory framework, which reframes embodied safety from progress-stalling guardrails to proactive hazard mitigation. The framework consists of a Risk-Sufficient Topological Belief Graph (RSG) for persistent safety-relevant state tracking under partial observability, an Agency-Grounded Factual Memory for precise hazard anticipation, and an Experience Memory that injects procedural Meta-Skills to guide executable, progress-preserving mitigation. Furthermore, we propose an automated Test-Verify-Write loop, allowing agents to continually refine their mitigation Meta-Skills from execution traces at test time. Experiments on IS-Bench demonstrate that our framework substantially boosts the Safe-Success rate across multiple VLM backbones (e.g., +30.3% on Qwen3-VL-8B), enabling agents to proactively mitigate hazards without stalling task progress. Code is available at https://github.com/DyMessi/JIT-Memory.

cs.LG↗
arXiv2026

SpikeTimer: Exploring Active Copyright Protection in Spiking Neural Networks via Temporal Backdoor Regularization

Spiking Neural Networks (SNN) have emerged as a revolutionary paradigm compared to traditional Deep Neural Networks (DNN) in energy-efficient computing, showcasing exceptional capabilities in processing event-driven sensory data for real-time applications like robotics and edge AI systems. However, unlike extensive studies on DNN copyright solutions, SNN copyright protection remains largely underexplored due to their inherent temporal coding complexities and spike-driven computation. In this study, we propose a novel active copyright protection framework named SpikeTimer for SNNs via temporal backdoor learning. SpikeTimer partitions neuromorphic data into designated timeslices and exclusively embeds authorized tokens within authorized slices. Furthermore, the inherent temporal segmentation characteristic intrinsically enables SpikeTimer to support multi-user authorization mechanisms and accommodates token embedding of arbitrary morphology. Based on this, SpikeTimer precisely responds to authorized data containing a token within the correct timeslice, while producing erroneous responses to unauthorized data. Our key innovation lies in establishing a time-dependent authorization mechanism that protects the SNN copyright by temporal token validity. Additionally, SpikeTimer retains its defensive efficacy even under adversarial attempts. Evaluations on multiple neuromorphic datasets manifest that SpikeTimer achieves around 10% accuracy on unauthorized data with merely around 1.5% degradation on authorized inputs. Moreover, SpikeTimer demonstrates robust resistance against model finetuning and pruning threats.

cs.CR↗
arXiv2026

EmbodiedUS-FS: Fast Slow Intelligence for Ultrasound Robotics

Robotic ultrasound scanning in real clinical environments requires both high-level clinical workflow reasoning and low-level closed-loop execution. Physicians natural-language instructions often contain implicit anatomical targets, procedural logic, image-quality requirements, and safety constraints, while execution is affected by patient motion, contact variations, and target drift. We propose a fast and slow hierarchical embodied ultrasound system for safe and interpretable robotic ultrasound assistance. The Slow Brain performs intent parsing and stage-wise task planning with knowledge augmentation from an API and handbook corpus, and generates executable plans through task-graph construction and structured plan verification. The Fast Brain fuses multimodal feedback, including ultrasound images, robot pose and force states, and patient-motion information, to refine local actions and perform image-quality-guided recovery behaviors. The system further integrates a Safety Shield and a hierarchical escalation policy to constrain risky actions and trigger replanning or human confirmation under persistent failures or safety-bound violations. Experiments on planning evaluation, closed-loop execution under dynamic perturbations, and safety-mechanism validation demonstrate that the proposed hierarchical design improves task success rates while reducing safety violations.

cs.RO↗
arXiv2026

TruthRL: Incentivizing Truthful LLMs via Reinforcement Learning

While large language models (LLMs) have demonstrated strong performance on factoid question answering, they are still prone to hallucination and untruthful responses, particularly when tasks demand information outside their parametric knowledge. Indeed, truthfulness requires more than accuracy -- models must also recognize uncertainty and abstain when unsure to avoid hallucinations. This presents a fundamental challenge for existing methods: approaches that optimize for accuracy often amplify hallucinations, while those that encourage abstention can become overly conservative, sacrificing correct answers. Both extremes ultimately compromise truthfulness. In this work, we present TruthRL, a general reinforcement learning (RL) framework that directly optimizes the truthfulness of LLMs. Specifically, we implement TruthRL using GRPO with a simple yet effective ternary reward that distinguishes correct answers, hallucinations, and abstentions. It incentivizes models to reduce hallucinations not only by providing correct responses, but also by enabling abstention when uncertain, thereby improving truthfulness. Extensive experiments across four knowledge-intensive benchmarks show that TruthRL significantly reduces hallucinations (e.g., 43.5% $\rightarrow$ 19.4%) and improves truthfulness (e.g., 5.3% $\rightarrow$ 37.2%), with consistent gains across various backbone models. Analysis shows that the improvement of TruthRL arises from enhanced capability of LLMs to recognize their knowledge boundary, hence avoiding being overly conservative as the baselines are.

cs.CL↗
arXiv2026

PACT: Self-Evolving Physical Safety Alignment for Diffusion Policies in Embodied Manipulation

Diffusion policies have achieved remarkable success in robotic manipulation, yet they often fail to satisfy strict physical constraints required for safe deployment. Existing approaches impose safety either prematurely during training or reactively via external guardrails at test time, limiting policy expressivity and overall scalability. We propose Physical safety Alignment for Constrained Trajectories (PACT), a self-evolving post-training framework that projects pretrained diffusion policies onto constraint-feasible regions without accessing demonstration data or task rewards. PACT distills constraint gradients into the diffusion model through a reverse-KL objective with dense supervision across timesteps. It incorporates a curriculum that progressively tightens constraints while maintaining theoretically bounded policy shift and monotone improvement, mitigating the safety-performance trade-off from catastrophic forgetting. On simulated and real-world embodied manipulation benchmarks, PACT significantly reduces safety violations by 31.0% on average while improving task success by 30.7%.

cs.RO↗
arXiv2026

Ultralight Dark Matter Detection with a Ferromagnet Lattice

A levitated ferromagnet provides an exceptionally sensitive probe of ultralight dark matter (ULDM) through measuring weak magnetic-like field signals. We propose a ferromagnet lattice magnetometer that coherently combines multiple levitated ferromagnets to enhance effective sensitivity. By replacing a single ferromagnet with a lattice, we increase the total polarized spin while preserving the intrinsic dynamical response of each constituent ferromagnet. We show that magnetic dipole-dipole interactions within the lattice can be dynamically suppressed through a high-frequency magnetic field, rendering the system effectively noninteracting, at the cost of only a moderate reduction in signal amplitude due to the distinct renormalization of linear and quadratic spin responses. We analyze the noise properties of the lattice and demonstrate that collective readout leads to favorable scaling with the number of ferromagnets. Interpreted in terms of axion-electron, dark photon, and axion-photon couplings, our results yield projected sensitivities that significantly exceed existing single-ferromagnet implementations. In particular, for axion-photon interactions, we find a nontrivial lattice-induced enhancement of the signal itself, leading to sensitivities that surpass existing constraints over a broad mass range.

hep-ph↗
arXiv2026

Anatomy-Slot: Unsupervised Anatomical Factorization for Homologous Bilateral Reasoning in Retinal Diagnosis

Retinal diagnosis is inherently bilateral: clinicians compare homologous structures across eyes (e.g., optic disc asymmetry), yet most deep models operate on monocular representations. We investigate whether explicit structural correspondence improves diagnosis, and propose Anatomy-Slot to operationalize this hypothesis. Anatomy-Slot introduces an unsupervised anatomical bottleneck by decomposing patch tokens into a set of emergent, structurally-coherent slots that correspond to anatomical regions, then aligning these slots across eyes via bidirectional cross-attention. On ODIR-5K with $n=10$ seeds, the method improves AUC by $4.2$ points over a matched ViT-L baseline (95% CIs; Wilcoxon signed-rank test, $W=0$, $p=0.002$). Pairing disruption and stress testing under Gaussian noise provide controlled tests of correspondence dependence and robustness under corruption. We further report quantitative optic disc grounding on REFUGE and cross-attention localization analysis. Beyond the reported gains, these results indicate that object-centric anatomical correspondence offers a principled path toward interpretable diagnostic systems aligned with clinical bilateral comparison.

cs.CV↗
arXiv2026

MuChator: Enabling Active Music Discovery via Conversational Music LLMs in Douyin Music

Douyin Music, a large-scale platform with millions of daily users, adopts an immersive, feed-based discovery paradigm, where users passively explore music through continuous recommendations. While effective for passive music discovery, this paradigm restricts users to recommendation results and provides limited support for explicitly specifying listening intents. Unlike conventional search, where users express well-defined intents through explicit queries such as specific songs or artists, real-world active music discovery is often situational and colloquial, involving vague or underspecified requests. While LLMs enable natural language interaction, their direct use in music discovery remains limited by insufficient music-domain knowledge, lack of music-query collaborative reasoning, and shallow understanding of personalized preferences. To address these challenges, we introduce MuChator, an interactive MusicLLM-based framework that enables users to actively express situational music intents in natural language. MuChator incorporates three key components: (1) Music Knowledge Pre-training, a three-stage scheme that incrementally injects objective music knowledge, subjective music knowledge, and personalized music preferences into LLMs; (2) Context-aware Instruction Tuning, which constructs high-quality user-query-music triplets through an automated synthesis pipeline to align LLMs with active and situational user intents; and (3) Preference Alignment with Hybrid RM, which jointly models intent relevance, personalized preferences, and basic constraints, and is optimized using GRPO-based reinforcement learning. Extensive evaluations on industrial music recommendation datasets demonstrate that MuChator outperforms leading proprietary models, such as Gemini-3-Pro. The model has been deployed on Douyin Music App within ByteDance, with 46.49\% improvement of user active days in online A/B test.

cs.IR↗
arXiv2026

FT-Dojo: Towards Autonomous LLM Fine-Tuning with Language Agents

Fine-tuning large language models for vertical domains remains labor-intensive, requiring practitioners to curate data, configure training, and iteratively diagnose model behavior. Despite growing interest in autonomous machine learning and language agents, end-to-end LLM fine-tuning has not been systematically studied as an interactive agent task. We introduce FT-Dojo, an interactive benchmark environment for autonomous LLM fine-tuning, comprising 13 tasks across 5 domains. Rather than a new collection of static datasets, FT-Dojo standardizes a task interface, shared raw-data repository, sandboxed execution environment, structured feedback protocol, and held-out evaluation procedure. We further develop FT-Agent, a fine-tuning-oriented autonomous framework that uses structured iteration planning, fail-fast validation, and multi-level feedback analysis to refine data and training strategies. Experiments show that FT-Agent provides a strong initial baseline, achieving the best performance on 10 out of 13 tasks, with additional controlled comparisons against frontier agents, open-source planning backbones, and multi-run statistics supporting the main findings. Case studies show that agents can recover from failures through cumulative learning, while still exposing limitations in causal diagnosis and long-horizon planning. The implementation is available at https://github.com/microsoft/rd-agent.

cs.AI↗
arXiv2026

Make It Long, Keep It Fast: End-to-End 10K Long User Behavior Sequence Modeling for Billion-Scale Douyin Recommendation

Short-video recommenders such as Douyin must exploit extremely long user behavior histories without breaking latency or cost budgets. We present an end-to-end industrial recommender system that scales long-sequence recommendation modeling to 10K-length histories in production. First, we introduce Stacked Target-to-History Cross Attention (STCA), which replaces history self-attention with stacked cross-attention from the target to the history, reducing complexity from quadratic to linear in sequence length and enabling efficient end-to-end training over long user behavior sequences. Second, we propose Request Level Batching (RLB), a user-centric batching scheme that aggregates multiple targets for the same user/request to share the user-side encoding, substantially lowering sequence-related storage, communication, and compute without changing the learning objective. Third, we design a length-extrapolative training strategy -- train on shorter windows, infer on much longer ones -- so the model generalizes to 10K-scale histories without additional training cost. Across offline and online experiments, we observe predictable, monotonic gains as we scale history length and model capacity, mirroring the scaling law behavior observed in large language models. Deployed at full traffic on Douyin, our system delivers significant improvements on key engagement metrics while meeting production latency, demonstrating a practical path to scaling end-to-end ultra-long sequence recommendation to the 10K regime.

cs.LG↗
arXiv2026

Agent^2 RL-Bench: Can LLM Agents Engineer Agentic RL Post-Training?

We introduce Agent2 RL-Bench, a compact diagnostic benchmark for evaluating agentic RL post-training, which tests whether LLM agents can autonomously design, implement, debug, and execute post-training pipelines that improve foundation models. RL post-training increasingly drives model alignment and specialization, yet existing benchmarks are largely static, rewarding supervised fine-tuning or script generation without assessing an agent's ability to close an interactive RL loop. Agent2 RL-Bench provides a unified agent-facing interface: each run starts from an isolated workspace containing a base model, task data, instructions, and a grading API, and agents must iterate within a fixed budget by training models and submitting artifacts for evaluation. The benchmark spans six tasks across three levels, from static rule-based training to judge-based optimization and closed-loop online RL with trajectory collection. Two diagnostic skills, namely runtime recording and post-hoc summarization, enable structured analysis of agent behavior, facilitating smooth and effective iteration of the benchmark's evaluation framework. Across five agent systems and six driver LLMs, agents show intelligent behavior but clear limitations: one RL-oriented run improves ALFWorld from 4.85 to 93.28 via SFT warm-up and GRPO with online rollouts, yet DeepSearchQA remains difficult, most successful routes rely on supervised pipelines, and interactive outcomes show large single-run differences across agent stacks. Overall, Agent2 RL-Bench shows that current agents can sometimes engineer online RL, but stable agent-driven RL post-training remains rare under fixed budgets. It also demonstrates that our benchmark provides a strong and effective evaluation framework for future research in this direction. Code is available at https://github.com/microsoft/RD-Agent/blob/main/rdagent/scenarios/rl/autorl_bench/README.md

cs.AI↗
arXiv2026

AdaFocus: Adaptive Relevance-Diversity Sampling with Zero-Cache Look-back for Efficient Long Video Understanding

Long video understanding is heavily bottlenecked by a rigid one-shot paradigm: existing methods either densely encode videos at prohibitive memory and latency costs, or aggressively compress them into sparse frame sets that irreversibly discard fine-grained evidence needed for downstream reasoning. Consequently, current models struggle to simultaneously balance temporal coverage, visual details, and computational efficiency. We propose AdaFocus, an efficient framework that rethinks long-video understanding as progressive evidence acquisition rather than one-pass encoding. AdaFocus relies on two tightly coupled components. First, a Query-Aware Adaptive Relevance-Diversity sampler (AdaRD) produces a compact yet informative video preview, adaptively switching to global clustering when the query lacks reliable local grounding. Second, instead of caching exhaustive frame sequences in memory, AdaFocus introduces an uncertainty-triggered refinement mechanism. It performs targeted look-back only when the model is not confident, retrieving high-resolution evidence directly from disk via a zero-cache I/O design. This turns discarded visual details from an irreversible loss into on-demand recoverable evidence without paying the cost of exhaustive preloading. Experiments on seven standard long-video benchmarks show that AdaFocus delivers a substantially better efficiency-accuracy trade-off than strong baselines. Compared with conventional dense encoding, AdaFocus achieves improved task performance (e.g., +2.59 accuracy on VideoMME, +8.39 mIoU on Charades-STA over single-pass inference) while reducing visual token consumption by ~33x and eliminating the need for in-memory frame pre-caching through its zero-cache disk retrieval design. These findings suggest that progressive preview combined with zero-cache evidence refinement is a highly effective paradigm for scalable multimedia reasoning.

cs.CV↗
arXiv2026

U-HNO: A U-shaped Hybrid Neural Operator with Sparse-Point Adaptive Routing for Non-stationary PDE Dynamics

Solutions to many partial differential equations (PDEs) display coexisting smooth global transport and localized sharp features within a single trajectory: shock fronts, thin interfaces, and concentrated high-frequency content sit on top of slowly varying backgrounds. This poses a challenge for neural operators: Fourier-based architectures mix nonlocal interactions efficiently but tend to under-resolve localized non-smooth features, whereas spatially local architectures recover fine detail at the cost of long-range propagation and rollout stability. Existing hybrid operators paper over this tension with a fixed, spatially uniform fusion that forces the same trade-off everywhere. We propose U-HNO, a U-shaped hybrid neural operator whose central design is Sparse-Point Adaptive Routing (SPAR): at every spatial location, a per-pixel hard mask selects whether the global Fourier branch or the local multi-scale Gaussian branch should dominate, and the sparsity ratio is a function of the local contrast of the routing signal, so smooth and shock-aligned regions receive different mixtures of global and local computation. SPAR is embedded in a hierarchical encoder-bottleneck-decoder backbone with skip connections so that the dual branches and the gate operate at every resolution. Training combines pointwise supervision with a finite-difference H^1 gradient term and a band-wise spectral consistency regularizer. Across benchmarks spanning 1D Burgers, Kuramoto-Sivashinsky, KdV, 2D advection, Allen-Cahn, Navier-Stokes, Darcy flow, and 3D transonic compressible Navier-Stokes from PDEBench, U-HNO achieves state-of-the-art rollout accuracy on the majority of tasks in both relative L^2 and H^1 metrics, with the largest gains on problems dominated by sharp localized features. Ablations show that removing any single component substantially degrades rollout error.

cs.LG↗
arXiv2026

MILD: Mediator Agent System with Bidirectional Perception and Multi-Layered Alignment for Human-Vehicle Collaboration

Prior studies report that partial driving automation can increase the cognitive demands on human drivers. This effect largely arises from human drivers' lack of transparent insight into the vehicle's intentions and decision logic, as well as from automated systems' limited awareness of the driver's dynamic state and preferences. This bidirectional misalignment undermines shared situational awareness and exacerbates coordination failures in human-vehicle interaction. To address these limitations, we argue for a paradigm shift that elevates the human role from passive supervisor to active manager. We introduce the Mediator-in-the-Loop-Driving (MILD) system, based on an agentic system architecture to facilitate synergistic human-vehicle collaboration. MILD integrates a perception agent for joint in-cabin and out-of-cabin understanding with a lightweight strategy agent that generates compliant and explainable action suggestions. To ensure these strategies are strictly aligned with safety regulations and human values, we develop Evidence- and Constraint-weighted Policy Optimization (ECPO). ECPO leverages automatic validators to steer the agent toward behaviors that are not only accurate but also structurally complete, substantiated by evidence, and free from constraint violations. Furthermore, a retrieval-augmented generation module dynamically incorporates constraints from traffic regulations, speed recommendations, and driver preferences into the decision loop. Field experiments across three open datasets demonstrate that MILD consistently outperforms baselines in both perception accuracy and strategy quality under auditable offline metrics, and yields higher human-rated policy adequacy, comfort, and explanation than baselines. This work offers a practical pathway for building auditable and aligned agents for human-vehicle collaborative driving.

cs.AI↗