arXiv subjects

Zhiyu Guo

Publications and source records attributed to Zhiyu Guo.

8 recordsLinked to original sources

arXiv2026

Unsupervised Post-Training of Foundation Models: A Survey

Foundation-model post-training usually relies on human labels, preference data, stronger teachers, or executable verifiers. We study Unsupervised Post-Training (UPT): update-bearing adaptation on unlabeled inputs whose learning signal is derived from same-lineage model artifacts rather than an external oracle. We catalog 80 strict UPT methods and organize them by the object that supplies the update signal: a prediction statistic, a sample relation, a self-generated target, or an internal evaluator. Beyond inventory, we show how the choice of internal signal and task structure determines whether post-training improves the model or recursively amplifies error. An orthogonal Input Visibility $\times$ Update Persistence view maps deployment regimes and defines a unified framework for UPT selection and evaluation.

cs.CL↗

arXiv2026

HAM-RAG: Hierarchy-Aware Multimodal RAG for Structure-Faithful Interleaved Generation

Existing multimodal RAG methods often flatten structured documents into isolated text and image units, weakening the source organization and local text-image logic needed for faithful evidence selection and placement. We propose HAM-RAG, a Hierarchy-Aware Multimodal RAG framework for structure-faithful interleaved generation. HAM-RAG uses document hierarchy as a grounding signal across retrieval and generation, contextualizing textual and visual evidence and preserving source position and local text-image relations in the prompt. We further introduce HAM-Bench, covering Wukong, Wiki, arXiv, and Recipe across game walkthroughs, web pages, scientific papers, and step-wise recipe documents. Across multiple backbones, HAM-RAG improves the main multimodal average by 17.3% over the strongest non-hierarchical baseline. On Wukong, HAM-RAG improves Img-CBS by 24.2% over the strongest non-hierarchical baseline, demonstrating substantially better local text-image alignment. The main experiments and ablation study together demonstrate that document hierarchy is a key grounding signal for faithful image selection, placement, and local text-image alignment. These findings highlight the value of hierarchy-aware grounding for reliable multimodal assistants that generate answers faithful to the source organization, procedural structure, and local text-image evidence of structured documents, such as technical manuals, maintenance guides, and industrial SOPs. The code is available at https://github.com/MCCodeAI/HAM-RAG.git.

cs.IR↗

arXiv2026

You only need 4 extra tokens: Synergistic Test-time Adaptation for LLMs

Large language models (LLMs) are increasingly deployed in specialized domains such as finance, medicine, and agriculture, where they face significant distribution shifts from their training data. Domain-specific fine-tuning can mitigate this challenge but relies on high-quality labeled data that is expensive and slow to collect in expertise-limited settings. We study label-free test-time adaptation for language models and present SyTTA, an inference-time framework that adapts models on-the-fly without additional supervision. SyTTA couples two complementary uncertainty signals that arise under distribution shift: input-side perplexity, indicating mismatch with domain-specific terminology and patterns, and output-side predictive entropy, indicating diffuse and unstable token probabilities during generation. Across diverse model architectures and domain-specific benchmarks, SyTTA delivers consistent gains. Notably, on agricultural question answering, SyTTA improves Rouge-LSum by over 120% on Qwen-2.5-7B with only 4 extra tokens per query. These results show that effective test-time adaptation for language models is achievable without labeled examples, supporting deployment in label-scarce domains. The code will be made available upon acceptance.

cs.CL↗

arXiv2025

Beyond the Pre-Service Horizon: Infusing In-Service Behavior for Improved Financial Risk Forecasting

Typical financial risk management involves distinct phases for pre-service risk assessment and in-service default detection, often modeled separately. This paper proposes a novel framework, Multi-Granularity Knowledge Distillation (abbreviated as MGKD), aimed at improving pre-service risk prediction through the integration of in-service user behavior data. MGKD follows the idea of knowledge distillation, where the teacher model, trained on historical in-service data, guides the student model, which is trained on pre-service data. By using soft labels derived from in-service data, the teacher model helps the student model improve its risk prediction prior to service activation. Meanwhile, a multi-granularity distillation strategy is introduced, including coarse-grained, fine-grained, and self-distillation, to align the representations and predictions of the teacher and student models. This approach not only reinforces the representation of default cases but also enables the transfer of key behavioral patterns associated with defaulters from the teacher to the student model, thereby improving the overall performance of pre-service risk assessment. Moreover, we adopt a re-weighting strategy to mitigate the model's bias towards the minority class. Experimental results on large-scale real-world datasets from Tencent Mobile Payment demonstrate the effectiveness of our proposed approach in both offline and online scenarios.

cs.LG↗

arXiv2024

Dependency-Aware Semi-Structured Sparsity of GLU Variants in Large Language Models

The rapid advancement in Large Language Models (LLMs) has markedly enhanced the capabilities of language understanding and generation. However, the substantial model size poses hardware challenges, affecting both memory size for serving and inference latency for token generation. To address those challenges, we propose Dependency-aware Semi-structured Sparsity (DaSS), a novel method for the recent prevalent GLU-based LLMs pruning, which incorporates structural dependency into the weight magnitude-based unstructured pruning. We introduce an MLP-specific pruning metric that evaluates the importance of each weight by jointly considering its magnitude and its corresponding MLP intermediate activation norms. DaSS facilitates a balance between the adaptability offered by unstructured pruning and the structural consistency inherent in dependency-based structured pruning. Empirical evaluations on LLaMA2, Mistral, and Gemma model families demonstrate that DaSS not only outperforms both SparseGPT and Wanda in achieving hardware-friendly N:M sparsity patterns but also maintains the computational efficiency of Wanda.

cs.CL↗

arXiv2024

Attention Score is not All You Need for Token Importance Indicator in KV Cache Reduction: Value Also Matters

Scaling the context size of large language models (LLMs) enables them to perform various new tasks, e.g., book summarization. However, the memory cost of the Key and Value (KV) cache in attention significantly limits the practical applications of LLMs. Recent works have explored token pruning for KV cache reduction in LLMs, relying solely on attention scores as a token importance indicator. However, our investigation into value vector norms revealed a notably non-uniform pattern questioning their reliance only on attention scores. Inspired by this, we propose a new method: Value-Aware Token Pruning (VATP) which uses both attention scores and the $ \ell_{1} $ norm of value vectors to evaluate token importance. Extensive experiments on LLaMA2-7B-chat and Vicuna-v1.5-7B across 16 LongBench tasks demonstrate that VATP outperforms attention-score-only baselines in over 12 tasks, confirming the effectiveness of incorporating value vector norms into token importance evaluation of LLMs.

cs.CL↗

arXiv2010

Beam optimization of RFQ and SFRFQ combined accelerator at Peking University

The Peking University Integral Split Ring Radio Frequency Quadrupole(ISR RFQ) accelerator was constructed in 1999 with a high duty factor 16.7% and repetition frequency 166Hz, and it was able to accelerate N+, O+,O-, C+ and He+ from 1.6kev/u to 65keV/u. It was later upgraded as an injector of the Separated Function RFQ (SFRFQ). The experiments indicated that the maximum accelerated O+ beam current could exceed 3.2mA with energy 1.03MeV and an energy spread (FWHM) 3.1%. Then the beam transports through a 1m-long magnetic triplet to the entrance of SFRFQ and is finally accelerated to 1.64MeV. The beam conditioning of RFQ were carefully optimized to satisfy the requirements of the SFRFQ. The combined accelerator eventually can deliver 0.53mA O+ beam with energy 1.65MeV, which has sufficiently demonstrated the feasibility of the SFRFQ structure.

physics.acc-ph↗

arXiv1997

ASCA Observations of the Supernova Remnant VRO42.05.01

We present the results of our ASCA SIS and GIS observations of the supernova remnant VRO42.05.01. Our spectral fits indicate that the SNR is isothermal (about 8.3 X 10^6 K), consistent with our ROSAT analysis results. The absorbing column density (about 2.9 X 10^21 cm^-2) obtained from these spectral fits is much smaller than expected for the nominal distance of 5 kpc, indicating that the line of sight toward VRO42.05.01 has an unusually low gas density. The spectral resolution of ASCA allows us to determine elemental abundances for the hot X-ray emitting plasma in the bright ``wing'' component of this remnant. We find that Mg, Si, and Fe are underabundant, and attribute these low abundances to the galactic metallicity gradient and to the location of the remnant in the outer Milky Way.

astro-ph↗