arXiv subjects

Huancheng Chen

Publications and source records attributed to Huancheng Chen.

At least 19 recordsLinked to original sources

arXiv2026

CacheRL:Multi-Turn Tool-Calling Agents via Cached Rollouts and Hybrid Reward

We present CacheRL, a system for training small agent foundation models that achieves 92 percent process accuracy on multi-step tool-calling tasks, approaching GPT-5's 94 percent while requiring 100 times less compute. Our approach addresses three challenges in practical agent training: transferring tool-calling knowledge from large models at scale, enabling reinforcement learning without costly live tool execution, and learning robustly from noisy cached environments. CacheRL introduces three key innovations. First, a hybrid thinking trajectory pipeline augments agent trajectories with LLM-generated reasoning traces, producing training examples that teach models not only what tools to call but also why. Second, the CacheAgentLoop eliminates live execution costs through a three-tier fuzzy cache while preserving trajectory fidelity using token-level masking. Third, a cache-tier-aware reward dynamically adjusts answer-quality weights to avoid penalizing models for cache-induced limitations. Through iterative supervised fine-tuning (SFT) and Group Relative Policy Optimization (GRPO), CacheRL improves Qwen3-4B-Thinking's validation reward from 0.43 to 0.78. On public agentic tool-calling benchmarks, our model achieves competitive performance against frontier models such as GPT-5. Ablation studies show that removing knowledge transfer reduces performance by 41 percent, while cache-aware rewards contribute a 17 percent improvement. Interestingly, reinforcement learning improves training stability but yields limited gains beyond strong supervised fine-tuning, suggesting that data quality and reward design play a more important role than complex optimization methods in building practical small agent models.

cs.CL↗

arXiv2026

Foundation-Preserving Adaptation via Generalized Rayleigh-Quotient Optimization

While finetuning effectively adapts foundation models to specialized downstream tasks, it can degrade nontarget capabilities acquired during pretraining. Existing forgetting aware methods typically seek safer updates through specialized initialization or fixed constraints, but do not regulate the adaptation preservation trade-off during training. We propose Foundation Preserving LoRA (FoLoRA), a forgetting aware optimization framework. Guided by a first order preservation condition, FoLoRA defines a forgetting penalty over pretraining-proxy activations and a task utility over downstream task activations. It then scores update directions by task utility per unit forgetting penalty via a generalized Rayleigh quotient. The resulting spectral coordinate system enables direction wise gated Adam updates, attenuating low utility to penalty directions during training. To estimate the forgetting penalty, FoLoRA constructs pretraining proxy calibration data by sampling from the pretrained model rather than relying on a single proxy dataset. Experiments on math, code, and instruction following adaptation show that FoLoRA achieves the strongest preservation adaptation balance over baselines, improving target task performance with best aggregate preservation of non target capabilities.

cs.LG↗

arXiv2026

Memex(RL): Scaling Long-Horizon LLM Agents via Indexed Experience Memory

Large language model (LLM) agents are fundamentally bottlenecked by finite context windows on long-horizon tasks. As trajectories grow, retaining tool outputs and intermediate reasoning in-context quickly becomes infeasible: the working context becomes prohibitively long, eventually exceeds the context budget, and makes distant evidence harder to use even when it is still present. Existing solutions typically shorten context through truncation or running summaries, but these methods are fundamentally lossy because they compress or discard past evidence itself. We introduce Memex, an indexed experience memory mechanism that instead compresses context without discarding evidence. Memex maintains a compact working context consisting of concise structured summaries and stable indices, while storing full-fidelity underlying interactions in an external experience database under those indices. The agent can then decide when to dereference an index and recover the exact past evidence needed for the current subgoal. We optimize both write and read behaviors with our reinforcement learning framework MemexRL, using reward shaping tailored to indexed memory usage under a context budget, so the agent learns what to summarize, what to archive, how to index it, and when to retrieve it. This yields a substantially less lossy form of long-horizon memory than summary-only approaches. We further provide a theoretical analysis showing the potential of the Memex loop to preserve decision quality with bounded dereferencing while keeping effective in-context computation bounded as history grows. Empirically, on challenging long-horizon tasks, Memex agent trained with MemexRL improves task success while using a significantly smaller working context.

cs.CL↗

arXiv2026

CoreQ: Learning-Free Mismatch Correction and Successive Rounding for Quantization

Post-training quantization (PTQ) enables efficient deployment of large language models by mapping pretrained weights to low-bit formats without retraining, typically using a small calibration set to minimize a layer-wise calibration objective. However, this sequential procedure induces a mismatch: errors from earlier quantized layers alter the inputs received by later layers, causing the activations to deviate from those of the full-precision model. Recent approaches introduce mismatch-aware calibration objectives to compensate for this effect, but leave open how much of the observed mismatch should shift each layer's calibration target. Fully applying this correction can overfit limited calibration data, while scaling the mismatch correction with a fixed coefficient ignores varying reliability of mismatch estimates across layers. To address these limitations, we propose CoreQ, a learning-free PTQ framework that applies a closed-form coefficient for mismatch correction derived from a geometric decomposition of the mismatch. The resulting coefficient adapts the correction across layers, reduces overfitting to finite calibration data, and requires no hyperparameter tuning. Given the corrected target, CoreQ minimizes the induced triangular least-squares objective with an efficient greedy successive-rounding solver and a bounded beam-search extension, K-CoreQ, that trades modest additional compute for improved performance. Across multiple LLM families, scales, bit-widths, and quantization settings, CoreQ improves perplexity and downstream accuracy over strong PTQ baselines.

cs.LG↗

arXiv2025

AI-Accelerated Qubit Readout at the Single-Photon Level for Scalable Atomic Quantum Processors

Quantum state readout with minimal resources is crucial for scalable quantum information processing. As a leading platform, neutral atom arrays rely on atomic fluorescence imaging for qubit readout, requiring short exposure, low photon count schemes to mitigate heating and atom loss while enabling mid-circuit feedback. However, a fundamental challenge arises in the single-photon regime where severe overlap in state distributions causes conventional threshold discrimination to fail. Here, we report an AI-accelerated Bayesian inference method for fluorescence readout in neutral atom arrays. Our approach leverages Bayesian inference to achieve reliable state detection at the single-photon level under short exposure. Specifically, we introduce a weakly anchored Bayesian scheme that requires calibration of only one state, addressing asymmetric calibration challenges common across quantum platforms. Furthermore, acceleration is achieved via a permutation-invariant neural network, which yields a 100-fold speedup by compressing iterative inference into a single forward pass. The approach achieves relative readout fidelity above 99% and 98% for histogram overlaps of 61% and 72%, respectively, enabling reliable extraction of Rabi oscillations and Ramsey interference results unattainable with conventional threshold based methods. This framework supports scalable, real-time readout of large atom arrays and paves the way toward AI-enhanced quantum technology in computation and sensing.

quant-ph↗

arXiv2025

Task-Agnostic Federated Continual Learning via Replay-Free Gradient Projection

Federated continual learning (FCL) enables collaborative model training across distributed clients on sequentially arriving tasks without revisiting past data. However, existing approaches often suffer from catastrophic forgetting, rely on replay buffers or generative models that may violate privacy constraints, or assume knowledge of task identities during inference. We propose FedProTIP (Federated Projection-based Continual Learning with Task Identity Prediction), a replay-free FCL framework that maintains shared task-specific feature subspaces across clients. Each client extracts low-rank core bases from intermediate activations using randomized singular value decomposition, capturing dominant feature directions associated with the current task. These bases are transmitted to the server and aggregated to construct global task subspaces that capture shared feature directions across clients without requiring data sharing. During training, client updates are projected onto the orthogonal complement of previously learned subspaces to reduce cross-task interference and mitigate catastrophic forgetting. The learned subspaces are also reused during inference to estimate task identity via subspace relevance, enabling task-agnostic prediction without requiring explicit task labels. Experiments on CIFAR100, ImageNet-R, and DomainNet demonstrate that FedProTIP consistently outperforms state-of-the-art federated continual learning baselines while maintaining lower training time, memory footprint, and communication cost.

cs.LG↗

arXiv2024

Training-Free Layout-to-Image Generation with Marginal Attention Constraints

Recently, many text-to-image diffusion models have excelled at generating high-resolution images from text but struggle with precise control over spatial composition and object counting. To address these challenges, prior works have developed layout-to-image (L2I) approaches that incorporate layout instructions into text-to-image models. However, existing L2I methods typically require fine-tuning of pre-trained parameters or training additional control modules for diffusion models. In this work, we propose a training-free L2I approach, MAC (Marginal Attention Constrained Generation), which eliminates the need for additional modules or fine-tuning. Specifically, we use text-visual cross-attention feature maps to quantify inconsistencies between the layout of the generated images and the provided instructions, and then compute loss functions to optimize latent features during the diffusion reverse process. To enhance spatial controllability and mitigate semantic failures under complex layout instructions, we leverage pixel-to-pixel correlations in self-attention feature maps to align cross-attention maps and combine three loss functions constrained by boundary attention to update latent features. Comprehensive experimental results on both L2I and non-L2I pretrained diffusion models demonstrate that our method outperforms existing training-free L2I techniques, both quantitatively and qualitatively, in terms of image composition on the DrawBench and HRS benchmarks.

cs.CV↗

arXiv2024

Replay-Free Continual Low-Rank Adaptation with Dynamic Memory

We revisit continual learning~(CL), which enables pre-trained vision transformers (ViTs) to sequentially fine-tune on new downstream tasks over time. However, as the scale of these models increases, catastrophic forgetting remains a more serious challenge. Recent studies highlight a crossover between CL techniques and parameter-efficient fine-tuning (PEFT), which focuses on fine-tuning only a small set of trainable parameters to adapt to downstream tasks, such as low-rank adaptation (LoRA). While LoRA achieves faster convergence and requires fewer trainable parameters, it has seldom been explored in the context of continual learning. To address this gap, we propose a novel PEFT-CL method called Dual Low-Rank Adaptation (DualLoRA), which introduces both an orthogonal LoRA adapter and a residual LoRA adapter parallel to pre-trained weights in each layer. These components are orchestrated by a dynamic memory mechanism to strike a balance between stability and plasticity. Additionally, we propose a scheme to predict task identity with confidence and calibrate the model's outputs accordingly. On ViT-based models, we demonstrate that DualLoRA offers significant advantages in accuracy, inference speed, and computation efficiency in training over existing CL methods across multiple benchmarks.

cs.CV↗

arXiv2024

Heterogeneity-Guided Client Sampling: Towards Fast and Efficient Non-IID Federated Learning

Statistical heterogeneity of data present at client devices in a federated learning (FL) system renders the training of a global model in such systems difficult. Particularly challenging are the settings where due to communication resource constraints only a small fraction of clients can participate in any given round of FL. Recent approaches to training a global model in FL systems with non-IID data have focused on developing client selection methods that aim to sample clients with more informative updates of the model. However, existing client selection techniques either introduce significant computation overhead or perform well only in the scenarios where clients have data with similar heterogeneity profiles. In this paper, we propose HiCS-FL (Federated Learning via Hierarchical Clustered Sampling), a novel client selection method in which the server estimates statistical heterogeneity of a client's data using the client's update of the network's output layer and relies on this information to cluster and sample the clients. We analyze the ability of the proposed techniques to compare heterogeneity of different datasets, and characterize convergence of the training process that deploys the introduced client selection method. Extensive experimental results demonstrate that in non-IID settings HiCS-FL achieves faster convergence than state-of-the-art FL client selection schemes. Notably, HiCS-FL drastically reduces computation cost compared to existing selection schemes and is adaptable to different heterogeneity scenarios.

cs.LG↗

arXiv2024

Recovering Labels from Local Updates in Federated Learning

Gradient inversion (GI) attacks present a threat to the privacy of clients in federated learning (FL) by aiming to enable reconstruction of the clients' data from communicated model updates. A number of such techniques attempts to accelerate data recovery by first reconstructing labels of the samples used in local training. However, existing label extraction methods make strong assumptions that typically do not hold in realistic FL settings. In this paper we present a novel label recovery scheme, Recovering Labels from Local Updates (RLU), which provides near-perfect accuracy when attacking untrained (most vulnerable) models. More significantly, RLU achieves high performance even in realistic real-world settings where the clients in an FL system run multiple local epochs, train on heterogeneous data, and deploy various optimizers to minimize different objective functions. Specifically, RLU estimates labels by solving a least-square problem that emerges from the analysis of the correlation between labels of the data points used in a training round and the resulting update of the output layer. The experimental results on several datasets, architectures, and data heterogeneity scenarios demonstrate that the proposed method consistently outperforms existing baselines, and helps improve quality of the reconstructed images in GI attacks in terms of both PSNR and LPIPS.

cs.LG↗

arXiv2023

Mixed-Precision Quantization for Federated Learning on Resource-Constrained Heterogeneous Devices

While federated learning (FL) systems often utilize quantization to battle communication and computational bottlenecks, they have heretofore been limited to deploying fixed-precision quantization schemes. Meanwhile, the concept of mixed-precision quantization (MPQ), where different layers of a deep learning model are assigned varying bit-width, remains unexplored in the FL settings. We present a novel FL algorithm, FedMPQ, which introduces mixed-precision quantization to resource-heterogeneous FL systems. Specifically, local models, quantized so as to satisfy bit-width constraint, are trained by optimizing an objective function that includes a regularization term which promotes reduction of precision in some of the layers without significant performance degradation. The server collects local model updates, de-quantizes them into full-precision models, and then aggregates them into a global model. To initialize the next round of local training, the server relies on the information learned in the previous training round to customize bit-width assignments of the models delivered to different clients. In extensive benchmarking experiments on several model architectures and different datasets in both iid and non-iid settings, FedMPQ outperformed the baseline FL schemes that utilize fixed-precision quantization while incurring only a minor computational overhead on the participating devices.

cs.LG↗

arXiv2023

Federated Learning in Non-IID Settings Aided by Differentially Private Synthetic Data

Federated learning (FL) is a privacy-promoting framework that enables potentially large number of clients to collaboratively train machine learning models. In a FL system, a server coordinates the collaboration by collecting and aggregating clients' model updates while the clients' data remains local and private. A major challenge in federated learning arises when the local data is heterogeneous -- the setting in which performance of the learned global model may deteriorate significantly compared to the scenario where the data is identically distributed across the clients. In this paper we propose FedDPMS (Federated Differentially Private Means Sharing), an FL algorithm in which clients deploy variational auto-encoders to augment local datasets with data synthesized using differentially private means of latent data representations communicated by a trusted server. Such augmentation ameliorates effects of data heterogeneity across the clients without compromising privacy. Our experiments on deep image classification tasks demonstrate that FedDPMS outperforms competing state-of-the-art FL methods specifically designed for heterogeneous data settings.

cs.LG↗

arXiv2023

The Best of Both Worlds: Accurate Global and Personalized Models through Federated Learning with Data-Free Hyper-Knowledge Distillation

Heterogeneity of data distributed across clients limits the performance of global models trained through federated learning, especially in the settings with highly imbalanced class distributions of local datasets. In recent years, personalized federated learning (pFL) has emerged as a potential solution to the challenges presented by heterogeneous data. However, existing pFL methods typically enhance performance of local models at the expense of the global model's accuracy. We propose FedHKD (Federated Hyper-Knowledge Distillation), a novel FL algorithm in which clients rely on knowledge distillation (KD) to train local models. In particular, each client extracts and sends to the server the means of local data representations and the corresponding soft predictions -- information that we refer to as ``hyper-knowledge". The server aggregates this information and broadcasts it to the clients in support of local training. Notably, unlike other KD-based pFL methods, FedHKD does not rely on a public dataset nor it deploys a generative model at the server. We analyze convergence of FedHKD and conduct extensive experiments on visual datasets in a variety of scenarios, demonstrating that FedHKD provides significant improvement in both personalized as well as global model performance compared to state-of-the-art FL methods designed for heterogeneous data settings.

cs.LG↗

arXiv2021

Skeleton-Graph: Long-Term 3D Motion Prediction From 2D Observations Using Deep Spatio-Temporal Graph CNNs

Several applications such as autonomous driving, augmented reality and virtual reality require a precise prediction of the 3D human pose. Recently, a new problem was introduced in the field to predict the 3D human poses from observed 2D poses. We propose Skeleton-Graph, a deep spatio-temporal graph CNN model that predicts the future 3D skeleton poses in a single pass from the 2D ones. Unlike prior works, Skeleton-Graph focuses on modeling the interaction between the skeleton joints by exploiting their spatial configuration. This is being achieved by formulating the problem as a graph structure while learning a suitable graph adjacency kernel. By the design, Skeleton-Graph predicts the future 3D poses without divergence in the long-term, unlike prior works. We also introduce a new metric that measures the divergence of predictions in the long term. Our results show an FDE improvement of at least 27% and an ADE of 4% on both the GTA-IM and PROX datasets respectively in comparison with prior works. Also, we are 88% and 93% less divergence on the long-term motion prediction in comparison with prior works on both GTA-IM and PROX datasets. Code is available at https://github.com/abduallahmohamed/Skeleton-Graph.git

cs.CV↗

arXiv2021

Large Magnetoresistance and Nontrivial Berry Phase in Nb3Sb Crystals with A15 Structure

Compounds with the A15 structure have attracted extensive attention due to their superconductivity and nontrivial topological band structure. We have successfully grown Nb$_3$Sb single crystals with a A15 structure and systematically measured the longitudinal resistivity, Hall resistivity and quantum oscillations in magnetization. Similar to other topological trivial/nontrivial semimetals, Nb$_3$Sb, exhibits large magnetoresistance (MR) at low temperatures (717$\%$, 2 K and 9 T), unsaturating quadratic field dependence of MR and up-turn behavior in $ρ_{xx}$(\emph{T}) curves under magnetic field, which is considered to result from a perfect hole-electron compensation, as evidenced by the Hall resistivity measurements. The nonzero Berry phase obtained from the de-Hass van Alphen (dHvA) oscillations demonstrates that Nb$_3$Sb is topologically nontrivial. These results indicate that Nb$_{3}$Sb superconductor is also a semimetal with large MR and nontrivial Berry phase, indicating that Nb$_{3}$Sb may be another platform to search for Majorana zero-energy mode.

cond-mat.supr-con↗

arXiv2021

Extremely large magnetoresistance in the "ordinary" metal ReO3

The extremely large magnetoresistance (XMR) observed in many topologically nontrivial and trivial semimetals has attracted much attention in relation to its underlying physical mechanism. In this paper, by combining the band structure and Fermi surface (FS) calculations with the Hall resistivity and de Haas-Van Alphen (dHvA) oscillation measurements, we studied the anisotropy of magnetoresistance (MR) of ReO$_3$ with a simple cubic structure, an "ordinary" nonmagnetic metal considered previously. We found that ReO$_3$ exhibits almost all the characteristics of XMR semimetals: the nearly quadratic field dependence of MR, a field-induced upturn in resistivity followed by a plateau at low temperatures, high mobilities of charge carriers. It was found that for magnetic field \emph{H} applied along the \emph{c} axis, the MR exhibits an unsaturated \emph{H}$^{1.75}$ dependence, which was argued to arise from the complete carrier compensation supported by the Hall resistivity measurements. For \emph{H} applied along the direction of 15$^\circ$ relative to the \emph{c} axis, an unsaturated \emph{H}$^{1.90}$ dependence of MR up to 9.43~$\times$~$10^3$$\%$ at 10~K and 9~T was observed, which was explained by the existence of electron open orbits extending along the $k_{x}$ direction. Two mechanisms responsible for XMR observed usually in the semimetals occur also in the simple metal ReO$_3$ due to its peculiar FS (two closed electron pockets and one open electron pocket), once again indicating that the details of FS topology are a key factor for the observed XMR in materials.

cond-mat.str-el↗

arXiv2020

Bulk Superconductivity in the Dirac Semimetal TlSb

A feasible strategy to realize the Majorana fermions is searching for a simple compound with both bulk superconductivity and Dirac surface states. In this paper, we performed calculations of electronic band structure, the Fermi surface and surface states, as well as measured the resistivity, magnetization, specific heat for TlSb compound with a CsCl-type structure. The band structure calculations show that TlSb is a Dirac semimetal when spin-orbit coupling is taken into account. Meanwhile, we first found that TlSb is a type-II superconductor with $T_c$ = 4.38 K, $H_{c1}$(0) = 148 Oe, $H_{c2}$(0) = 1.12 T and $κ_{GL}$ = 10.6, and confirmed it to be a moderately coupled s-wave superconductor. Although we can not determine which bands near the Fermi level $E_F$ to be responsible for superconductivity, its coexistence with the topological surface states implies that TlSb compound may be a simple material platform to realize the fault-tolerant quantum computations.

cond-mat.supr-con↗

arXiv2020

Linear and quadratic magnetoresistance in the semimetal SiP2

Multiple mechanisms for extremely large magnetoresistance (XMR) found in many topologically nontrivial/trivial semimetals have been theoretically proposed, but experimentally it is unclear which mechanism is responsible in a particular sample. In this article, by the combination of band structure calculations, numerical simulations of magnetoresistance (MR), Hall resistivity and de Haas-van Alphen (dHvA) oscillation measurements, we studied the MR anisotropy of SiP$_{2}$ which is verified to be a topologically trivial, incomplete compensation semimetal. It was found that as magnetic field, $H$, is applied along the $a$ axis, the MR exhibits an unsaturated nearly linear $H$ dependence, which was argued to arise from incomplete carriers compensation. For the $H$ $\parallel$ [101] orientation, an unsaturated nearly quadratic $H$ dependence of MR up to 5.88 $\times$ 10$^{4}$$\%$ (at 1.8 K, 31.2 T) and field-induced up-turn behavior in resistivity were observed, which was suggested due to the existence of hole open orbits extending along the $k_{x}$ direction. Good agreement of the experimental results with the simulations based on the calculated Fermi surface (FS) indicates that the topology of FS plays an important role in its MR.

cond-mat.mtrl-sci↗