arXiv subjects

Shupeng Li

Publications and source records attributed to Shupeng Li.

10 recordsLinked to original sources

arXiv2026

LOGIGEN: Logic-Driven Generation of Verifiable Agentic Tasks

The evolution of Large Language Models (LLMs) from static instruction-followers to autonomous agents necessitates operating within complex, stateful environments to achieve precise state-transition objectives. However, this paradigm is bottlenecked by data scarcity, as existing tool-centric reverse-synthesis pipelines fail to capture the rigorous logic of real-world applications. We introduce \textbf{LOGIGEN}, a logic-driven framework that synthesizes verifiable training data based on three core pillars: \textbf{Hard-Compiled Policy Grounding}, \textbf{Logic-Driven Forward Synthesis}, and \textbf{Deterministic State Verification}. Specifically, a Triple-Agent Orchestration is employed: the \textbf{Architect} compiles natural-language policy into database constraints to enforce hard rules; the \textbf{Set Designer} initializes boundary-adjacent states to trigger critical policy conflicts; and the \textbf{Explorer} searches this environment to discover causal solution paths. This framework yields a dataset of 20,000 complex tasks across 8 domains, where validity is strictly guaranteed by checking exact state equivalence. Furthermore, we propose a verification-based training protocol where Supervised Fine-Tuning (SFT) on verifiable trajectories establishes compliance with hard-compiled policy, while Reinforcement Learning (RL) guided by dense state-rewards refines long-horizon goal achievement. On $\tau^2$-Bench, LOGIGEN-32B(RL) achieves a \textbf{79.5\% success rate}, substantially outperforming the base model (40.7\%). These results demonstrate that logic-driven synthesis combined with verification-based training effectively constructs the causally valid trajectories needed for next-generation agents.

cs.AI↗

arXiv2026

SWE-Hub: A Unified Production System for Scalable, Executable Software Engineering Tasks

Progress in software-engineering agents is increasingly constrained by the scarcity of executable, scalable, and realistic data for training and evaluation. This scarcity stems from three fundamental challenges in existing pipelines: environments are brittle and difficult to reproduce across languages; synthesizing realistic, system-level bugs at scale is computationally expensive; and existing data predominantly consists of short-horizon repairs, failing to capture long-horizon competencies like architectural consistency. We introduce \textbf{SWE-Hub}, an end-to-end system that operationalizes the data factory abstraction by unifying environment automation, scalable synthesis, and diverse task generation into a coherent production stack. At its foundation, the \textbf{Env Agent} establishes a shared execution substrate by automatically converting raw repository snapshots into reproducible, multi-language container environments with standardized interfaces. Built upon this substrate, \textbf{SWE-Scale} engine addresses the need for high-throughput generation, combining cross-language code analysis with cluster-scale validation to synthesize massive volumes of localized bug-fix instances. \textbf{Bug Agent} generates high-fidelity repair tasks by synthesizing system-level regressions involving cross-module dependencies, paired with user-like issue reports that describe observable symptoms rather than root causes. Finally, \textbf{SWE-Architect} expands the task scope from repair to creation by translating natural-language requirements into repository-scale build-a-repo tasks. By integrating these components, SWE-Hub establishes a unified production pipeline capable of continuously delivering executable tasks across the entire software engineering lifecycle.

cs.AI↗

arXiv2025

QianfanHuijin Technical Report: A Novel Multi-Stage Training Paradigm for Finance Industrial LLMs

Domain-specific enhancement of Large Language Models (LLMs) within the financial context has long been a focal point of industrial application. While previous models such as BloombergGPT and Baichuan-Finance primarily focused on knowledge enhancement, the deepening complexity of financial services has driven a growing demand for models that possess not only domain knowledge but also robust financial reasoning and agentic capabilities. In this paper, we present QianfanHuijin, a financial domain LLM, and propose a generalizable multi-stage training paradigm for industrial model enhancement. Our approach begins with Continual Pre-training (CPT) on financial corpora to consolidate the knowledge base. This is followed by a fine-grained Post-training pipeline designed with increasing specificity: starting with Financial SFT, progressing to Finance Reasoning RL and Finance Agentic RL, and culminating in General RL aligned with real-world business scenarios. Empirical results demonstrate that QianfanHuijin achieves superior performance across various authoritative financial benchmarks. Furthermore, ablation studies confirm that the targeted Reasoning RL and Agentic RL stages yield significant gains in their respective capabilities. These findings validate our motivation and suggest that this fine-grained, progressive post-training methodology is poised to become a mainstream paradigm for various industrial-enhanced LLMs.

cs.CL↗

arXiv2025

Qianfan-VL: Domain-Enhanced Universal Vision-Language Models

We present Qianfan-VL, a series of multimodal large language models ranging from 3B to 70B parameters, achieving state-of-the-art performance through innovative domain enhancement techniques. Our approach employs multi-stage progressive training and high-precision data synthesis pipelines, which prove to be critical technologies for enhancing domain-specific capabilities while maintaining strong general performance. Qianfan-VL achieves comparable results to leading open-source models on general benchmarks, with state-of-the-art performance on benchmarks such as CCBench, SEEDBench IMG, ScienceQA, and MMStar. The domain enhancement strategy delivers significant advantages in OCR and document understanding, validated on both public benchmarks (OCRBench 873, DocVQA 94.75%) and in-house evaluations. Notably, Qianfan-VL-8B and 70B variants incorporate long chain-of-thought capabilities, demonstrating superior performance on mathematical reasoning (MathVista 78.6%) and logical inference tasks. All models are trained entirely on Baidu's Kunlun P800 chips, validating the capability of large-scale AI infrastructure to train SOTA-level multimodal models with over 90% scaling efficiency on 5000 chips for a single task. This work establishes an effective methodology for developing domain-enhanced multimodal models suitable for diverse enterprise deployment scenarios.

cs.CV↗

arXiv2024

Vector spectrometer with Hertz-level resolution and super-recognition capability

High-resolution optical spectrometers are crucial in revealing intricate characteristics of signals, determining laser frequencies, measuring physical constants, identifying substances, and advancing biosensing applications. Conventional spectrometers, however, often grapple with inherent trade-offs among spectral resolution, wavelength range, and accuracy. Furthermore, even at high resolution, resolving overlapping spectral lines during spectroscopic analyses remains a huge challenge. Here, we propose a vector spectrometer with ultrahigh resolution, combining broadband optical frequency hopping, ultrafine microwave-photonic scanning, and vector detection. A programmable frequency-hopping laser was developed, facilitating a sub-Hz linewidth and Hz-level frequency stability, an improvement of four and six orders of magnitude, respectively, compared to those of state-of-the-art tunable lasers. We also designed an asymmetric optical transmitter and receiver to eliminate measurement errors arising from modulation nonlinearity and multi-channel crosstalk. The resultant vector spectrometer exhibits an unprecedented frequency resolution of 2 Hz, surpassing the state-of-the-art by four orders of magnitude, over a 33-nm range. Through high-resolution vector analysis, we observed that group delay information enhances the separation capability of overlapping spectral lines by over 47%, significantly streamlining the real-time identification of diverse substances. Our technique fills the gap in optical spectrometers with resolutions below 10 kHz and enables vector measurement to embrace revolution in functionality.

physics.optics↗

arXiv2023

Advancing Transformer Architecture in Long-Context Large Language Models: A Comprehensive Survey

Transformer-based Large Language Models (LLMs) have been applied in diverse areas such as knowledge bases, human interfaces, and dynamic agents, and marking a stride towards achieving Artificial General Intelligence (AGI). However, current LLMs are predominantly pretrained on short text snippets, which compromises their effectiveness in processing the long-context prompts that are frequently encountered in practical scenarios. This article offers a comprehensive survey of the recent advancement in Transformer-based LLM architectures aimed at enhancing the long-context capabilities of LLMs throughout the entire model lifecycle, from pre-training through to inference. We first delineate and analyze the problems of handling long-context input and output with the current Transformer-based models. We then provide a taxonomy and the landscape of upgrades on Transformer architecture to solve these problems. Afterwards, we provide an investigation on wildly used evaluation necessities tailored for long-context LLMs, including datasets, metrics, and baseline models, as well as optimization toolkits such as libraries, frameworks, and compilers to boost the efficacy of LLMs across different stages in runtime. Finally, we discuss the challenges and potential avenues for future research. A curated repository of relevant literature, continuously updated, is available at https://github.com/Strivin0311/long-llms-learning.

cs.CL↗

arXiv2022

Crop mapping in the small sample/no sample case: an approach using a two-level cascade classifier and integrating domain knowledge

Mapping crops using remote sensing technology is important for food security and land management. Machine learning-based methods has become a popular approach for crop mapping in recent years. However, the key to machine learning, acquiring ample and accurate samples, is usually time-consuming and laborious. To solve this problem, a crop mapping method in the small sample/no sample case that integrating domain knowledge and using a cascaded classification framework that combine a weak classifier learned from samples with strong features and a strong classifier trained by samples with weak feature was proposed. First, based on the domain knowledge of various crops, a low-capacity classifier such as decision tree was applied to acquire those pixels with distinctive features and complete observation sequences as "strong feature" samples. Then, to improve the representativeness of these samples, sample augmentation strategy that artificially remove the observations of "strong feature" samples according to the average valid observation proportion in target area was applied. Finally, based on the original samples and augmented samples, a large-capacity classifier such as random forest was trained for crop mapping. The method achieved an overall accuracy of 82% in the MAP crop recognition competition held by Syngenta Group, China in 2021 (third prize, ranked fourth). This method integrates domain knowledge to overcome the difficulties of sample acquisition, providing a convenient, fast and accurate solution for crop mapping.

cs.CV↗

arXiv2022

Learning from Training Dynamics: Identifying Mislabeled Data Beyond Manually Designed Features

While mislabeled or ambiguously-labeled samples in the training set could negatively affect the performance of deep models, diagnosing the dataset and identifying mislabeled samples helps to improve the generalization power. Training dynamics, i.e., the traces left by iterations of optimization algorithms, have recently been proved to be effective to localize mislabeled samples with hand-crafted features. In this paper, beyond manually designed features, we introduce a novel learning-based solution, leveraging a noise detector, instanced by an LSTM network, which learns to predict whether a sample was mislabeled using the raw training dynamics as input. Specifically, the proposed method trains the noise detector in a supervised manner using the dataset with synthesized label noises and can adapt to various datasets (either naturally or synthesized label-noised) without retraining. We conduct extensive experiments to evaluate the proposed method. We train the noise detector based on the synthesized label-noised CIFAR dataset and test such noise detector on Tiny ImageNet, CUB-200, Caltech-256, WebVision and Clothing1M. Results show that the proposed method precisely detects mislabeled samples on various datasets without further adaptation, and outperforms state-of-the-art methods. Besides, more experiments demonstrate that the mislabel identification can guide a label correction, namely data debugging, providing orthogonal improvements of algorithm-centric state-of-the-art techniques from the data aspect.

cs.CV↗

arXiv2020

Dual-comb delay spectroscopy with attometer resolution

Spectroscopy has attracted much attention in molecular detection, biomolecular identification, and chemical analysis for providing accurate measurement. However, it is almost unable to distinguish different sources with overlapped resonances in mixed analytes. Here, we present dual-comb delay spectroscopy to overcome this problem. The introduction of group delay spectroscopy provides a new tool to identify sources that would lead to overlapped resonances in intensity or phase spectroscopy. To obtain sufficiently high spectral resolution and signal-to-noise ratio for achieving reliable group delay spectrum, a probe comb with the wavelengths precisely scaned by a microwave source is applied, leading to attometer-level resolution and million-level signal-to-noise ratio. In an experiment, spectroscopy with an optional resolution up to 1 kHz (8 attometer), an average signal-to-noise ratio surpassing 2,000,000, and a span exceeding 33 nm is demonstrated. Two overlapped resonances from two different sources are clearly differentiated. Our work offers a new perspective for exploring the interaction between matter and light.

physics.ins-det↗

arXiv2019

Optical vector analysis with attometer resolution, 90-dB dynamic range and THz bandwidth

Optical vector analysis (OVA) having the capability to achieve magnitude and phase responses is essential for fabrication and application of emerging optical devices. However, the conventional OVA often have to make compromises among resolution, dynamic range and bandwidth. Modulation-based OVA promises ultra-high resolution but suffers from measurement errors due to the modulation nonlinearity, narrow bandwidth restricted by the employment of microwave and optoelectronic devices, and difficulties in measuring optical devices with deep notch restricted by small signal modulation or the high-order sidebands. This paper proposes an original method to meet the measurement requirements for ultra-wide bandwidth, ultra-high resolution and ultra-large dynamic range simultaneously, based on asymmetric optical probe signal generator (ASG) and receiver (ASR). The use of the ASG and ASR not only doubles the measurement range without spectral aliasing, but also removes the measurement errors introduced by the modulation nonlinearity. The optimal signal modulation and enormous sideband suppression ratio in the ASG enables an ultra-large dynamic range. Thanks to the wavelength-independence of the ASG and ASR, the measurement range can boost to 2N times by applying an N-tone optical frequency comb (OFC) without complicated operation. In an experiment, OVA with a record resolution of 334 Hz (2.67 attometer in the 1550-nm band), a dynamic range of >90 dB and a measurement range of 1.025 THz is demonstrated.

physics.ins-det↗