arXiv subjects

Bo Deng

Publications and source records attributed to Bo Deng.

At least 19 recordsLinked to original sources

arXiv2026

BLPO: Beyond-Local Policy Optimization for Long-Horizon LLM Agents

Long-horizon tasks require large language model (LLM) agents to coordinate interdependent actions, a capability central to automating complex workflows. Stepwise group-based reinforcement learning provides finer credit than trajectory-level supervision, but reliable comparisons must account for differences in interaction history. Enforcing this consistency within the limited rollouts of each task can leave insufficient support for estimating action preferences. We propose Beyond-Local Policy Optimization (BLPO), which estimates task-progress advantages from experience shared across tasks and training updates. We summarize trajectory prefixes into task-progress states and abstract actions by their function, allowing corresponding decisions from different tasks to contribute to shared estimates. To control bias and noise from heterogeneous tasks, we accumulate context-relative returns in a global memory with balanced aggregation and reliability weighting. These global advantages complement local comparisons over normalized actions and trajectory-level credit without a learned critic or additional rollouts. Experiments on ALFWorld and WebShop with 1.5B and 7B policies show improved task success over the evaluated baselines under matched training budgets. Further analyses show faster learning and shorter trajectories with little computational overhead.

cs.AI↗

arXiv2026

Size and spectral radius conditions for the existence of $k$-factor in graphs with given minimum degree

In recent years, characterizing the conditions for the existence of an $[a,b]$-factor in graphs has received significant academic attention. For integers $b\ge a\ge 1$, an $[a, b]$-factor of a graph $G$ is defined as a spanning subgraph $H$ such that $a\le d_{H}(v)\le b$ for every $v\in V(G)$. If $a=b=k$, a $[k, k]$-factor is called a $k$-factor of $G$. Let $δ(G)$ denote the minimum degree of a graph $G$. Hao and Li [Electron. J. Combin. (2024)] posed the problem of characterizing the size or spectral radius conditions for a graph to contain an $[a,b]$-factor with $δ(G)\ge a$. Fan, Lin and Zhu [Discrete Math. (2026)] solved the size version of this problem for the case $a < b.$ In this paper, for $a=b=k$, we characterize the size condition for a graph to have a $k$-factor with $δ(G)\ge k$. Using this result, we also characterize the spectral radius and distance spectral radius conditions for a graph to have a $k$-factor, respectively.

math.CO↗

arXiv2026

FinEvo-Bench: A Longitudinal Benchmark for Self-Evolving Agents in Professional Financial Workflows

Agents used over time encounter recurring professional work: each case requires different evidence and judgment, while the underlying workflow can be reused. Benchmarks built from independent tasks cannot reveal whether an agent turns earlier experience into better procedures for later cases. We introduce FinEvo-Bench, a longitudinal benchmark designed around this structure. It contains 120 open-ended tasks drawn from real cases across 20 business scenes in six financial domains. Each scene contains six substantively different cases that share a professional workflow and an expert-authored rubric for task quality and financial compliance. Constructing and validating the benchmark required approximately 1,200 person-hours. Finance provides a natural test bed because recurring analyses apply shared professional and compliance requirements to heterogeneous inputs, producing case-specific analyses and conclusions. We evaluate four self-evolving agent scaffolds with Qwen3.7-Max on three independently shuffled, globally interleaved task streams. A Claude Code rubric judge backed by Claude Opus~4.6 evaluates all outputs, and paired state-reset controls estimate each scaffold's gain from retained experience. Evolving runs score 9.33--19.37 points higher and trigger 0.12--0.44 fewer compliance issues per task than their paired controls. Paired score gains at within-scene ranks~4--6 exceed those at ranks~1--3 by 6.10--8.70 points. FinEvo-Bench measures whether retained experience improves later professional work under continued use.

cs.AI↗

arXiv2026

DeFA: Dependency-Guided Failure Attribution for LLM Agents

Errors in LLM agent executions and their visible consequences can be separated by many steps, making decisive-error localization a matter of understanding both step content and step dependencies. We introduce DeFA, a dependency-guided framework for agent failure attribution. DeFA first combines protocol relations and semantic dependencies into an event dependency graph spanning the trajectory. It then identifies events that may violate task requirements and traces their sources and subsequent effects to construct a failure propagation graph. Finally, DeFA uses step evidence and the steps' roles in failure propagation to identify the decisive error, responsible agent, and error category. To support long trajectories, DeFA partitions executions into segments and combines the current segment's detailed content with summaries of the other segments, giving local diagnosis access to global execution context. Across Who and When and the Who and When Pro text subset, DeFA achieves the highest responsible-agent and exact step accuracy with all evaluated backbones, and the highest failure-mode accuracy among taxonomy-aligned methods on Pro. Further experiments on image and video trajectories demonstrate its applicability to multimodal failure attribution. Ablations support the contributions of segmentation, the event dependency graph, and the failure propagation graph. Using DeFA's diagnostic feedback for skill evolution in Trace2Skill improves downstream task accuracy by 6-15 percentage points over the native pipeline, showing that the diagnoses can also support agent improvement on subsequent tasks.

cs.AI↗

arXiv2026

Aromatic molecular emitters in a hexagonal boron nitride stack

Single polycyclic aromatic hydrocarbon molecules embedded in organic matrices have proven to be an excellent family of narrow-linewidth quantum emitters. Extending this host-guest setting to van der Waals materials offers the opportunity to combine the preeminent properties of molecular emitters with the access to the versatility of two-dimensional hetero-structures and devices. In this work, we incorporate perylene molecules into multi-layered hexagonal boron nitride stacks and observe gigahertz-narrow zero-phonon-line transitions at cryogenic temperatures. We unambiguously verify the origins of photon emission through vibronic spectra analysis. By combining hyperspectral localization measurements with quantum chemistry calculations, we examine the insertion mechanisms of perylene molecules in the hexagonal boron nitride stacks, and conclude that pristine hBN layers tend to expel molecules from the sandwich, while extended morphological defects, hydroxyl groups and unpassivated boron and nitrogen atoms assist to stabilize molecular bindings to hBN. Our work provides valuable insight for future work to deterministically integrate narrow-linewidth molecular emitters into van der Waals devices.

physics.optics↗

arXiv2026

Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning

We introduce Nemotron 3 Ultra, a 550 billion total and 55 billion active parameter Mixture-of-Experts Hybrid Mamba-Attention language model. We pre-trained Nemotron 3 Ultra on 20 trillion text tokens, then extended the context length to 1M tokens, and post-trained using Supervised Fine Tuning (SFT), Reinforcement Learning (RL), and Multi-teacher On-Policy Distillation (MOPD). Nemotron 3 Ultra is our most capable model yet, employing multiple key technologies - LatentMoE, Multi Token Prediction (MTP), NVFP4 pre-training, multi-environment RLVR, MOPD, and reasoning budget control. Nemotron 3 Ultra achieves up to ~6x higher inference throughput as compared to state-of-the-art publicly available LLMs while attaining on-par accuracy. The state-of-the-art accuracy, high inference throughput, and 1M token context length make Nemotron 3 Ultra ideal for long-running autonomous agentic tasks. We open-source the base, post-trained, and quantized checkpoints, along with the training data and recipe on HuggingFace.

cs.CL↗

arXiv2026

Benchmarking PNW Model for MedMNIST to 100% Accuracy

In this paper, we introduce a new concept called Artificial Special Intelligence by which Machine Learning models for the classification problem can be trained error-free, thus acquiring the capability of not making repeated mistakes. The method is applied to 18 MedMNIST biomedical datasets. Except for three datasets, which suffer from the double-labeling problem, all are trained to perfection.

cs.AI↗

arXiv2026

Baseline Method of the Foundation Model Challenge for Ultrasound Image Analysis

Ultrasound (US) imaging exhibits substantial heterogeneity across anatomical structures and acquisition protocols, posing significant challenges to the development of generalizable analysis models. Most existing methods are task-specific, limiting their suitability as clinically deployable foundation models. To address this limitation, the Foundation Model Challenge for Ultrasound Image Analysis (FM\_UIA~2026) introduces a large-scale multi-task benchmark comprising 27 subtasks across segmentation, classification, detection, and regression. In this paper, we present the official baseline for FM\_UIA~2026 based on a unified Multi-Head Multi-Task Learning (MH-MTL) framework that supports all tasks within a single shared network. The model employs an ImageNet-pretrained EfficientNet--B4 backbone for robust feature extraction, combined with a Feature Pyramid Network (FPN) to capture multi-scale contextual information. A task-specific routing strategy enables global tasks to leverage high-level semantic features, while dense prediction tasks exploit spatially detailed FPN representations. Training incorporates a composite loss with task-adaptive learning rate scaling and a cosine annealing schedule. Validation results demonstrate the feasibility and robustness of this unified design, establishing a strong and extensible baseline for ultrasound foundation model research. The code and dataset are publicly available at \href{https://github.com/lijiake2408/Foundation-Model-Challenge-for-Ultrasound-Image-Analysis}{GitHub}.

cs.CV↗

arXiv2026

FUGC: Benchmarking Semi-Supervised Learning Methods for Cervical Segmentation

Accurate segmentation of cervical structures in transvaginal ultrasound (TVS) is critical for assessing the risk of spontaneous preterm birth (PTB), yet the scarcity of labeled data limits the performance of supervised learning approaches. This paper introduces the Fetal Ultrasound Grand Challenge (FUGC), the first benchmark for semi-supervised learning in cervical segmentation, hosted at ISBI 2025. FUGC provides a dataset of 890 TVS images, including 500 training images, 90 validation images, and 300 test images. Methods were evaluated using the Dice Similarity Coefficient (DSC), Hausdorff Distance (HD), and runtime (RT), with a weighted combination of 0.4/0.4/0.2. The challenge attracted 10 teams with 82 participants submitting innovative solutions. The best-performing methods for each individual metric achieved 90.26\% mDSC, 38.88 mHD, and 32.85 ms RT, respectively. FUGC establishes a standardized benchmark for cervical segmentation, demonstrates the efficacy of semi-supervised methods with limited labeled data, and provides a foundation for AI-assisted clinical PTB risk assessment.

eess.IV↗

arXiv2026

NL2Repo-Bench: Towards Long-Horizon Repository Generation Evaluation of Coding Agents

Recent advances in coding agents suggest rapid progress toward autonomous software development, yet existing benchmarks fail to rigorously evaluate the long-horizon capabilities required to build complete software systems. Most prior evaluations focus on localized code generation, scaffolded completion, or short-term repair tasks, leaving open the question of whether agents can sustain coherent reasoning, planning, and execution over the extended horizons demanded by real-world repository construction. To address this gap, we present NL2Repo Bench, a benchmark explicitly designed to evaluate the long-horizon repository generation ability of coding agents. Given only a single natural-language requirements document and an empty workspace, agents must autonomously design the architecture, manage dependencies, implement multi-module logic, and produce a fully installable Python library. Our experiments across state-of-the-art open- and closed-source models reveal that long-horizon repository generation remains largely unsolved: even the strongest agents achieve below 40% average test pass rates and rarely complete an entire repository correctly. Detailed analysis uncovers fundamental long-horizon failure modes, including premature termination, loss of global coherence, fragile cross-file dependencies, and inadequate planning over hundreds of interaction steps. NL2Repo Bench establishes a rigorous, verifiable testbed for measuring sustained agentic competence and highlights long-horizon reasoning as a central bottleneck for the next generation of autonomous coding agents.

cs.CL↗

arXiv2026

DVD: A Robust Method for Detecting Variant Contamination in Large Language Model Evaluation

Evaluating large language models (LLMs) is increasingly confounded by \emph{variant contamination}: the training corpus contains semantically equivalent yet lexically or syntactically altered versions of test items. Unlike verbatim leakage, these paraphrased or structurally transformed variants evade existing detectors based on sampling consistency or perplexity, thereby inflating benchmark scores via memorization rather than genuine reasoning. We formalize this problem and introduce \textbf{DVD} (\textbf{D}etection via \textbf{V}ariance of generation \textbf{D}istribution), a single-sample detector that models the local output distribution induced by temperature sampling. Our key insight is that contaminated items trigger alternation between a \emph{memory-adherence} state and a \emph{perturbation-drift} state, yielding abnormally high variance in the synthetic difficulty of low-probability tokens; uncontaminated items remain in drift with comparatively smooth variance. We construct the first benchmark for variant contamination across two domains Omni-MATH and SuperGPQA by generating and filtering semantically equivalent variants, and simulate contamination via fine-tuning models of different scales and architectures (Qwen2.5 and Llama3.1). Across datasets and models, \textbf{DVD} consistently outperforms perplexity-based, Min-$k$\%++, edit-distance (CDD), and embedding-similarity baselines, while exhibiting strong robustness to hyperparameters. Our results establish variance of the generation distribution as a principled and practical fingerprint for detecting variant contamination in LLM evaluation.

cs.AI↗

arXiv2025

Toward Errorless Training ImageNet-1k

In this paper, we describe a feedforward artificial neural network trained on the ImageNet 2012 contest dataset [7] with the new method of [5] to an accuracy rate of 98.3% with a 99.69 Top-1 rate, and an average of 285.9 labels that are perfectly classified over the 10 batch partitions of the dataset. The best performing model uses 322,430,160 parameters, with 4 decimal places precision. We conjecture that the reason our model does not achieve a 100% accuracy rate is due to a double-labeling problem, by which there are duplicate images in the dataset with different labels.

cs.CV↗

arXiv2024

A Comprehensive Study on A Tapered Paul Trap: From Design to Potential Applications

We present a tapered Paul trap whose radio frequency electrodes are inclined to the symmetric axis of the endcap electrodes, resulting in a funnel-shaped trapping potential. With this configuration, a charged particle confined in this trap has its radial degrees of freedom coupled to that of the axial direction. The same design was successfully used to experimentally realize a single-atom heat engine, and with this setup amplification of zeptonewton forces was implemented. In this paper, we show the design, implementation, and characterization of such an ion trap in detail. This system offers a high level of control over the ion's motion. Its novel features promise applications in the field of quantum thermodynamics, quantum sensing, and quantum information.

quant-ph↗

arXiv2023

Error-free Training for Artificial Neural Network

Conventional training methods for artificial neural network (ANN) models never achieve zero error rate systematically for large data. A new training method consists of three steps: first create an auxiliary data from conventionally trained parameters which correspond exactly to a global minimum for the loss function of the cloned data; second create a one-parameter homotopy (hybrid) of the auxiliary data and the original data; and third train the model for the hybrid data iteratively from the auxiliary data end of the homotopy parameter to the original data end while maintaining the zero-error training rate at every iteration. This continuationmethod is guaranteed to converge numerically by a theorem which converts the ANN training problem into a continuation problem for fixed points of a parameterized transformation in the training parameter space to which the Uniform Contraction Mapping Theorem from dynamical systems applies.

cs.LG↗

arXiv2023

Theory of Infectious Diseases with Testing and Testing-less Covid-19 Endemic

What is the long term dynamics of the Covid-19 pandemic? How will it end? Here we constructed an infectious disease model with testing and analyzed the existence and stability of its endemic states. For a large parameter set, including those relevant to the SARS-CoV-2 virus, we demonstrated the existence of one endemic equilibrium without testing and one endemic equilibrium with testing and proved their local and global stabilities for some cases. Our results suggest that the pandemic is to end with a testing-less endemic state through a novel and surprising mechanism called stochastic trapping.

q-bio.PE↗

arXiv2023

Forecast U.S. Covid-19 Numbers by Open SIR Model with Testing

The U.S. Covid-19 data exhibit a high-frequency oscillation along a low-frequency wave for outbreaks. There is no model to account for it. A modified SIR model is proposed to explain this spiking phenomenon. It is also used to best-fit the data and to make forecast. For the simulated duration of 590 days, the model is capable of achieving a 0.5 percent mean squared relative error (MSRE) fit to the seven-day average of the daily case number. The outright 28-day prediction by the model generates a 20 percent MSRE for the cumulative case total due to a persistent underestimation of the data by the model. With the proposed correction to the aberration, the model is able to keep the 28-day cumulative case total forecast within 10 percent MSRE of the data.

q-bio.PE↗

arXiv2023

Golden Ratio Information for Neural Spike Code

Spike bursting is a ubiquitous feature of all neuronal systems. Assuming the spiking states form an alphabet for a communication system, what is the optimal information processing rate? and what is the channel capacity? Here we demonstrate that the quaternary alphabet of spike number code gives the maximal processing rate, and that a binary source in Golden Ratio distribution gives rise to the channel capacity. A multi-time scaled neural circuit is shown to satisfy the hypotheses of this neural communication system.

q-bio.NC↗

arXiv2023

One-Dimensional Deep Image Prior for Curve Fitting of S-Parameters from Electromagnetic Solvers

A key problem when modeling signal integrity for passive filters and interconnects in IC packages is the need for multiple S-parameter measurements within a desired frequency band to obtain adequate resolution. These samples are often computationally expensive to obtain using electromagnetic (EM) field solvers. Therefore, a common approach is to select a small subset of the necessary samples and use an appropriate fitting mechanism to recreate a densely-sampled broadband representation. We present the first deep generative model-based approach to fit S-parameters from EM solvers using one-dimensional Deep Image Prior (DIP). DIP is a technique that optimizes the weights of a randomly-initialized convolutional neural network to fit a signal from noisy or under-determined measurements. We design a custom architecture and propose a novel regularization inspired by smoothing splines that penalizes discontinuous jumps. We experimentally compare DIP to publicly available and proprietary industrial implementations of Vector Fitting (VF), the industry-standard tool for fitting S-parameters. Relative to publicly available implementations of VF, our method shows superior performance on nearly all test examples using only 5-15% of the frequency samples. Our method is also competitive to proprietary VF tools and often outperforms them for challenging input instances.

cs.LG↗