arXiv subjects

Fangzhou Xu

Publications and source records attributed to Fangzhou Xu.

11 recordsLinked to original sources

arXiv2026

PGMT: Perceptive General Motion Tracking for Humanoid Robots

Humanoid motion trackers can reproduce diverse whole-body motions, but their performance degrades on complex terrain where terrain-agnostic references become physically infeasible. We present PGMT, a Perceptive General Motion Tracking pipeline for humanoid robots that learns terrain adaptation from independently selected motion references and terrains. PGMT first learns a general tracking and recovery prior, then incorporates terrain perception through motion-conditioned terrain glimpses that selectively encode regions relevant to the current motion. Terrain-aware tracking relaxation allows necessary deviations from the reference while preserving its motion intent. Zero-shot deployment on a Unitree G1 demonstrates robust terrain-adaptive locomotion and whole-body motion execution over real-world terrain with obstacles up to 37 cm high, while supporting teleoperation, dynamic motion tracking, and fall recovery. PGMT extends general humanoid motion tracking beyond flat ground, providing a unified policy for terrain-adaptive locomotion, diverse whole-body behaviors, and teleoperation in complex environments. Project homepage: https://luyili.github.io/pgmt/

cs.RO↗

arXiv2026

TAGA: Terrain-aware Active Gaze Learning for Generalizable Agile Humanoid Locomotion

Agile humanoid locomotion across diverse challenging terrain demands both wide perceptual coverage and precise local geometry understanding. Motivated by the way humans selectively look at relevant terrain during locomotion, we introduce TAGA, a Terrain-aware Active Gaze learning framework for Attention-based humanoid control. By fusing vision, proprioception, and motion commands, our framework guides the model to learn anticipatory cues and actively attend to specific areas of the height scan, selectively using these informative regions for the downstream network. This adaptively increases the information density of observations under tight onboard computational constraints, thus enabling fine-grained perceptive locomotion over larger-scale terrains. We find that such gaze behaviors can naturally emerge through reinforcement learning alone, without requiring additional supervision or explicit guidance, significantly improve training efficiency. As a result, the trained policy demonstrates robust and generalizable locomotion in simulation and on hardware, including reliable terrain-aware foothold selection, elevated-platform traversal, competitive sparse-foothold traversal, and the largest reported real-world gap traversal distance of 1.2m among perceptive humanoid locomotion systems, while maintaining stability under severe perceptual disturbances and environmental interference.

cs.RO↗

arXiv2026

Efficient Retrieval Scaling with Hierarchical Indexing for Large Scale Recommendation

The increase in data volume, computational resources, and model parameters during training has led to the development of numerous large-scale industrial retrieval models for recommendation tasks. However, effectively and efficiently deploying these large-scale foundational retrieval models remains a critical challenge that has not been fully addressed. Common quick-win solutions for deploying these massive models include relying on offline computations (such as cached user dictionaries) or distilling large models into smaller ones. Yet, both approaches fall short of fully leveraging the representational and inference capabilities of foundational models. In this paper, we explore whether it is possible to learn a hierarchical organization over the memory of foundational retrieval models. Such a hierarchical structure would enable more efficient search by reducing retrieval costs while preserving exactness. To achieve this, we propose jointly learning a hierarchical index using cross-attention and residual quantization for large-scale retrieval models. We also present its real-world deployment at Meta, supporting daily advertisement recommendations for billions of Facebook and Instagram users. Interestingly, we discovered that the intermediate nodes in the learned index correspond to a small set of high-quality data. Fine-tuning the model on this set further improves inference performance, and concretize the concept of "test-time training" within the recommendation system domain. We demonstrate these findings using both internal and public datasets with strong baseline comparisons and hope they contribute to the community's efforts in developing the next generation of foundational retrieval models.

cs.IR↗

arXiv2025

Hierarchical Structured Neural Network: Efficient Retrieval Scaling for Large Scale Recommendation

Retrieval, the initial stage of a recommendation system, is tasked with down-selecting items from a pool of tens of millions of candidates to a few thousands. Embedding Based Retrieval (EBR) has been a typical choice for this problem, addressing the computational demands of deep neural networks across vast item corpora. EBR utilizes Two Tower or Siamese Networks to learn representations for users and items, and employ Approximate Nearest Neighbor (ANN) search to efficiently retrieve relevant items. Despite its popularity in industry, EBR faces limitations. The Two Tower architecture, relying on a single dot product interaction, struggles to capture complex data distributions due to limited capability in learning expressive interactions between users and items. Additionally, ANN index building and representation learning for user and item are often separate, leading to inconsistencies exacerbated by representation (e.g. continuous online training) and item drift (e.g. items expired and new items added). In this paper, we introduce the Hierarchical Structured Neural Network (HSNN), an efficient deep neural network model to learn intricate user and item interactions beyond the commonly used dot product in retrieval tasks, achieving sublinear computational costs relative to corpus size. A Modular Neural Network (MoNN) is designed to maintain high expressiveness for interaction learning while ensuring efficiency. A mixture of MoNNs operate on a hierarchical item index to achieve extensive computation sharing, enabling it to scale up to large corpus size. MoNN and the hierarchical index are jointly learnt to continuously adapt to distribution shifts in both user interests and item distributions. HSNN achieves substantial improvement in offline evaluation compared to prevailing methods.

cs.IR↗

arXiv2024

Human-Like Code Quality Evaluation through LLM-based Recursive Semantic Comprehension

Code quality evaluation involves scoring generated code quality based on a reference code for a specific problem statement. Currently, there are two main forms of evaluating code quality: match-based evaluation and execution-based evaluation. The former requires the collection of a large number of test cases, making a huge cost. The latter relies on superficial code matching as an evaluation metric, which fails to accurately capture code semantics. Moreover, extensive research has demonstrated that match-based evaluations do not truly reflect code quality. With the development of large language models (LLMs) in recent years, studies have proven the feasibility of using LLMs as evaluators for generative tasks. However, due to issues like hallucinations and uncertainty in LLMs, their correlation with human judgment remains at a lower level, making the direct use of LLMs for code quality evaluation challenging. To address these issues, we propose Human-Like Code Quality Evaluation through LLM-based Recursive Semantic Comprehension (HuCoSC). We employ a recursive approach to enable LLMs to comprehend portions of code semantics independently each time, obtaining the code semantics through multiple interactions with LLMs. We designed a Semantic Dependency Decoupling Storage to make independent analysis feasible, allowing LLMs to achieve more accurate semantics by breaking down complex problems. Finally, the generated code is scored based on a semantic comparison between the reference code and itself. Experimental results indicate that HuCoSC surpasses existing state-of-the-art methods in terms of correlation with human experts and correlation with code execution.

cs.SE↗

arXiv2024

Empowering Agile-Based Generative Software Development through Human-AI Teamwork

In software development, the raw requirements proposed by users are frequently incomplete, which impedes the complete implementation of application functionalities. With the emergence of large language models, recent methods with the top-down waterfall model employ a questioning approach for requirement completion, attempting to explore further user requirements. However, users, constrained by their domain knowledge, lack effective acceptance criteria, which fail to capture the implicit needs of the user. Moreover, the cumulative errors of the waterfall model can lead to discrepancies between the generated code and user requirements. The Agile methodologies reduce cumulative errors through lightweight iteration and collaboration with users, but the challenge lies in ensuring semantic consistency between user requirements and the code generated. We propose AgileGen, an agile-based generative software development through human-AI teamwork. AgileGen attempts for the first time to use testable requirements by Gherkin for semantic consistency between requirements and code. Additionally, we innovate in human-AI teamwork, allowing users to participate in decision-making processes they do well and enhancing the completeness of application functionality. Finally, to improve the reliability of user scenarios, a memory pool mechanism is used to collect user decision-making scenarios and recommend them to new users. AgileGen, as a user-friendly interactive system, significantly outperformed existing best methods by 16.4% and garnered higher user satisfaction.

cs.SE↗

arXiv2024

A test-free semantic mistakes localization framework in Neural Code Translation

In the task of code translation, neural network-based models have been shown to frequently produce semantically erroneous code that deviates from the original logic of the source code. This issue persists even with advanced large models. Although a recent approach proposed using test cases to identify these semantic errors, it relies heavily on the quality of the test cases and is not applicable to code snippets without test cases in real-world scenarios. Therefore, We present EISP, a static analysis framework based on the Large Language Model (LLM).First, the framework generates a semantic mapping between source code and translated code. Next, each sub-code fragment is identified by recursively traversing the abstract syntax tree of the source code, and its corresponding translated code fragment is found through the semantic mapping. Finally, EISP connects each pair of sub-code fragments with fine-grained knowledge hints through an AI chain to assist LLMs in discovering semantic mistakes in the translated code. In our benchmark evaluation, the EISP framework, based on GPT-4o mini, achieved an accuracy of 82.3\%, representing a 20.3\% improvement over baseline methods using the same base model, and a 7.4\% improvement compared to dynamic analysis methods that require test cases and manual intervention. To our knowledge, EISP is the first tool to locate semantic errors in translated code without test cases or compilable code. This innovative tool provides the software engineering community with a new way to deal with code fragments without test cases.

cs.SE↗

arXiv2021

Axion and dark photon limits from Crab Nebula high energy gamma-rays

The observation of cosmic sub-PeV gamma-rays from the Crab Nebula opens up the possibility of testing cosmic ray photon transparency at the multi-hundred TeV scale. Assuming no deviation from a source gamma-ray emission due to accelerated electron inverse-Compton scattering, higher event energies can extend constraints on the effects of new physics; We consider oscillation between gamma-rays and axions/dark photons, plus attenuation effects from gamma-ray absorption in the case of dark photon dark matter. Combining the recent AS$γ$ and HAWC sub-PeV data with earlier MAGIC and HEGRA data, axion-like particles are most constrained in the $10^{-7}-10^{-6}$ eV mass range, where the coupling $g_{aγ}$ is constrained to be below ${1.8}\times 10^{-10}$ GeV$^{-1}$. In comparison, gamma ray flux attenuation due to oscillation with a dark photon leads to a very weak constraint on the mixing parameter; $ε\lesssim$ 0.2 for dark photon mass between $10^{-7}$ and $10^{-6}$ eV. Direct scattering from dark photon dark matter limits $ε\lesssim$ 0.01 for masses between $6$ and $400$ eV.

astro-ph.HE↗

arXiv2020

$SU(3)_C\times SU(3)_L\times U(1)_X$ model from $SU(6)$

We propose the $SU(3)_C\times SU(3)_L\times U(1)_X$ model arising from $SU(6)$ breaking. One family of the Standard Model (SM) fermions arises from two $\bar{6}$ representations and one $15$ representation of $SU(6)$ gauge symmetry. To break the $SU(3)_C\times SU(3)_L\times U(1)_X$ gauge symmetry down to the SM, we introduce three $SU(3)_L$ triplet Higgs fields, where two of them come from the $\bar{6}$ representation while the other one from the $15$ representation. We study the gauge boson masses and Higgs boson mass in detail, and find that the vacuum expectation value (VEV) of the Higgs field for $SU(3)_L\times U(1)_X$ gauge symmetry breaking is around 10 TeV. The neutrino masses and mixing can be generated via the littlest inverse seesaw mechanism. In particular, we have normal hierarchy for neutrino masses and the lightest active neutrino is massless. Also, we consider constraints from the charged lepton flavor changing decays as well. Furthermore, introducing two $SU(3)_L$ adjoint fermions, one $SU(3)_C$ adjoint scalar, and one $SU(3)_L$ triplet scalar, we can achieve gauge coupling unification within 1\%. These extra particles can provide a dark matter candidate as well.

hep-ph↗

arXiv2020

BusTr: Predicting Bus Travel Times from Real-Time Traffic

We present BusTr, a machine-learned model for translating road traffic forecasts into predictions of bus delays, used by Google Maps to serve the majority of the world's public transit systems where no official real-time bus tracking is provided. We demonstrate that our neural sequence model improves over DeepTTE, the state-of-the-art baseline, both in performance (-30% MAPE) and training stability. We also demonstrate significant generalization gains over simpler models, evaluated on longitudinal data to cope with a constantly evolving world.

cs.LG↗

arXiv2018

Deep learning for the R-parity violating supersymmetry searches at the LHC

Supersymmetry with hadronic R-parity violation in which the lightest neutralino decays into three quarks is still weakly constrained. This work aims to further improve the current search for this scenario by the boosted decision tree method with additional information from jet substructure. In particular, we find a deep neural network turns out to perform well in characterizing the neutralino jet substructure. We first construct a Convolutional Neutral Network (CNN) which is capable of tagging the neutralino jet in any signal process by using the idea of jet image. When applied to pure jet samples, such a CNN outperforms the N-subjettiness variable by a factor of a few in tagging efficiency. Moreover, we find the method, which combines the CNN output and jet invariant mass, can perform better and is applicable to a wider range of neutralino mass than the CNN alone. Finally, the ATLAS search for the signal of gluino pair production with subsequent decay $\tilde{g} \to q q \tildeχ^0_1 (\to q q q)$ is recasted as an application. In contrast to the pure sample, the heavy contamination among jets in this complex final state renders the discriminating powers of the CNN and N-subjettiness similar. By analyzing the jets substructure in events which pass the ATLAS cuts with our CNN method, the exclusion limit on gluino mass can be pushed up by $\sim200$ GeV for neutralino mass $\sim 100$ GeV.

hep-ph↗