arXiv subjects

Baoyi Wang

Publications and source records attributed to Baoyi Wang.

4 recordsLinked to original sources

arXiv2026

StreamOPD: A Post-Training Recipe with Spatio-Temporal Cue Gating for Streaming Video Understanding

Streaming video understanding demands direct responses from the causally observed prefix of an unfolding video. Existing systems add inference-time memory, retrieval, and compression, yet a training-free sliding-window baseline already matches them. We therefore fix a memory-free recent-window protocol and ask how far post-training alone can go. Reinforcement learning with verifiable rewards fits this regime poorly, encouraging long ``think-then-answer'' generations, while on-policy distillation (OPD) supplies dense token-level teacher supervision on student trajectories but is stable only when both models train in thinking mode. These observations lead to \textsc{StreamOPD}, a recipe combining verifiable streaming-video data, thinking-mode OPD, and instruct-mode deployment. It raises StreamingBench from $77.9\%$ to $83.9\%$---within $0.3$ points of the 9B teacher---and improves OVO-Bench excluding its hallucination-detection subtask (HLD) by $9.1$ points under unchanged inference. As a teacher-privilege extension, \emph{Spatio-Temporal CueGate (ST-CueGate)} aggregates cue-versus-no-cue teacher likelihood ratios into a group-relative response score that reweights OPD. It reaches $71.9\%$ on OVO-Bench (excluding HLD) and $64.9\%$ on Video-MME, and is the only variant that stays above the base model on all four benchmarks. Replacing the teacher with a frozen copy of the student's initial policy---on-policy self-distillation---retains most of these gains and lifts HLD to $57.0\%$, above both the untrained student and the 9B teacher, so abstention loss is not intrinsic to the recipe. We provide a transparent and reproducible reference for open-source streaming-video research.

cs.CV↗

arXiv2026

Better Call Grep: Evaluating and Improving Grep-Like Lexical Retrieval for Repository-Level Code Completion

Repository-level code completion remains challenging for large language models (LLMs) due to cross-file dependencies and limited context windows. Prior work addresses this challenge using Retrieval-Augmented Generation (RAG) frameworks based on semantic indexing or structure-aware graph analysis, but these approaches incur substantial computational overhead for index construction and maintenance. Motivated by common developer workflows that rely on lightweight search utilities (e.g., ripgrep), we revisit a fundamental yet underexplored question: how far can simple, index-free lexical retrieval support repository-level code completion before more complex retrieval mechanisms become necessary? To answer this question, we systematically investigate lightweight, index-free, intent-aware lexical retrieval through extensive empirical analysis. We first introduce Naive GrepRAG, a baseline framework in which LLMs autonomously generate ripgrep commands to retrieve relevant context. Despite its simplicity, Naive GrepRAG achieves performance comparable to sophisticated graph-based baselines. Further analysis shows that its effectiveness stems from retrieving lexically precise code fragments that are spatially closer to the completion site. We also identify key limitations of lexical retrieval, including sensitivity to noisy matches from high-frequency ambiguous keywords and context fragmentation caused by rigid truncation boundaries. To address these issues, we propose GrepRAG, which augments lexical retrieval with a lightweight post-processing pipeline featuring identifier-weighted re-ranking and structure-aware deduplication. Extensive evaluation on CrossCodeEval and RepoEval-Updated demonstrates that GrepRAG consistently outperforms state-of-the-art (SOTA) methods, achieving 7.04-15.58 percent relative improvement in code exact match (EM) over the best baseline on CrossCodeEval.

cs.SE↗

arXiv2025

GRACE: Graph-Guided Repository-Aware Code Completion through Hierarchical Code Fusion

LLMs excel in localized code completion but struggle with repository-level tasks due to limited context windows and complex semantic and structural dependencies across codebases. While Retrieval-Augmented Generation (RAG) mitigates context scarcity by retrieving relevant code snippets, current approaches face significant limitations. They overly rely on textual similarity for retrieval, neglecting structural relationships such as call chains and inheritance hierarchies, and lose critical structural information by naively concatenating retrieved snippets into text sequences for LLM input. To address these shortcomings, GRACE constructs a multi-level, multi-semantic code graph that unifies file structures, abstract syntax trees, function call graphs, class hierarchies, and data flow graphs to capture both static and dynamic code semantics. For retrieval, GRACE employs a Hybrid Graph Retriever that integrates graph neural network-based structural similarity with textual retrieval, refined by a graph attention network-based re-ranker to prioritize topologically relevant subgraphs. To enhance context, GRACE introduces a structural fusion mechanism that merges retrieved subgraphs with the local code context and preserves essential dependencies like function calls and inheritance. Extensive experiments on public repository-level benchmarks demonstrate that GRACE significantly outperforms state-of-the-art methods across all metrics. Using DeepSeek-V3 as the backbone LLM, GRACE surpasses the strongest graph-based RAG baselines by 8.19% EM and 7.51% ES points on every dataset. The code is available at https://anonymous.4open.science/r/grace_icse-C3D5.

cs.SE↗

arXiv2014

A semi-empirical response function for Gamma-ray of Scintillation detector based on physical interaction mechanism

Scintillation detector has lower energy resolution for Gamma-ray as compared to semiconductor detector, better spectra analysis method is essential to traditional method. A model for describing the response function of scintillation detector over the range of incident Gamma-ray energies between 0.5 and 1.5 MeV has been established and applied to fitting radiation sources spectra. Each function form for describing the feature of Gamma-ray spectra are based on the analysis of fundamental interaction mechanism. These functions are combined to form a DRF model to fit experiment spectra by weighted least squares fitting method, parameters in this model are obtained simultaneously. Gaussian standard deviation can be calculated out by an individual procedure. Validity of the DRF model is demonstrated by fitting Co-60 and Cs-137 spectra measured by CsI(Tl) detector and comparing them to the normalized equivalent measured spectrum.

physics.ins-det↗