arXiv subjects

Chenglin Yu

Publications and source records attributed to Chenglin Yu.

7 recordsLinked to original sources

arXiv2026

The Hidden Footprint: Making Storage a First-Class Metric for LLM Agent Evaluation

LLM agent benchmarks measure task completion, reliability, and inference cost, but not the persistent data an agent run leaves on disk, including logs, context snapshots, checkpoints, and debug traces. We introduce AgentFootprint, a cross-framework benchmark of post-run agent storage footprint. Its serialization-aware metric suite measures total retention, channel composition, duplication, growth, compressibility, and conversation-history reconstructability. It addresses a measurement trap: naive byte-level measurement understates duplication by an order of magnitude because database paging and JSON escaping obscure repeated content. A fixed-trace control separates agent-generated logical volume from persistence-layer amplification: replaying the same trajectory through seven persisting frameworks yields a 6.7x spread. Under identical models, tools, and tasks, configurations with 100% accuracy differ by 15.7x in retained bytes, although their defaults support different recovery and audit capabilities. Three full-history configurations grow superlinearly on a repeated-observation stress task. Exported trajectories from 108 instance-normalized SWE-bench Verified submissions span three orders of magnitude per instance, with no detectable correlation with resolve rate. A content-addressed store reduces retention by 4.8x-32.7x while preserving every reconstructability score. These results establish persistent storage as a resource metric to report jointly with accuracy and reconstructability.

cs.AI↗

arXiv2026

Compile, Then Page: Executable SOP Programs and a Capability-Gated Runtime for Procedural LLM Agents

Enterprise agents must follow long-horizon, conditional, safety-critical standard operating procedures (SOPs). We compile machine-readable SOP constraints into executable pseudo-code and run them with a program-guided (PG) stack machine that pages the active frame while an LLM performs semantic execution. A three-arm SOPBench study across six models separates representation from runtime: compiled text never significantly hurts and gains up to 16.0 points where official prose underperforms. Runtime guidance is capability-gated. Two strong models independently show positive seven-domain PG contrasts (58:19 and 75:31 discordant pairs), whereas weak models are harmed. A full-program cursor ablation (active frame first, complete program retained) recovers much of the strong-model refusal gain; selective visibility adds a smaller improvement. Paired probe and audit measurements track this divide to spontaneous state discipline rather than reconstruction ability. On Bank the three primary arms rise from 70.4 to 86.4 to 92.8, with 100% refusal correctness. Practical guidance: compile first; enable active-frame paging only after a model-level discipline check.

cs.AI↗

arXiv2026

InfiAgent: An Infinite-Horizon Framework for General-Purpose Autonomous Agents

LLM agents can reason and use tools, but they often break down on long-horizon tasks due to unbounded context growth and accumulated errors. Common remedies such as context compression or retrieval-augmented prompting introduce trade-offs between information fidelity and reasoning stability. We present InfiAgent, a general-purpose framework that keeps the agent's reasoning context strictly bounded regardless of task duration by externalizing persistent state into a file-centric state abstraction. At each step, the agent reconstructs context from a workspace state snapshot plus a fixed window of recent actions. Experiments on DeepResearch and an 80-paper literature review task show that, without task-specific fine-tuning, InfiAgent with a 20B open-source model is competitive with larger proprietary systems and maintains substantially higher long-horizon coverage than context-centric baselines. These results support explicit state externalization as a practical foundation for stable long-horizon agents. Github Repo:https://github.com/ChenglinPoly/infiAgent

cs.AI↗

arXiv2025

InfiAgent: Self-Evolving Pyramid Agent Framework for Infinite Scenarios

Large Language Model (LLM) agents have demonstrated remarkable capabilities in organizing and executing complex tasks, and many such agents are now widely used in various application scenarios. However, developing these agents requires carefully designed workflows, carefully crafted prompts, and iterative tuning, which requires LLM techniques and domain-specific expertise. These hand-crafted limitations hinder the scalability and cost-effectiveness of LLM agents across a wide range of industries. To address these challenges, we propose \textbf{InfiAgent}, a Pyramid-like DAG-based Multi-Agent Framework that can be applied to \textbf{infi}nite scenarios, which introduces several key innovations: a generalized "agent-as-a-tool" mechanism that automatically decomposes complex agents into hierarchical multi-agent systems; a dual-audit mechanism that ensures the quality and stability of task completion; an agent routing function that enables efficient task-agent matching; and an agent self-evolution mechanism that autonomously restructures the agent DAG based on new tasks, poor performance, or optimization opportunities. Furthermore, InfiAgent's atomic task design supports agent parallelism, significantly improving execution efficiency. This framework evolves into a versatile pyramid-like multi-agent system capable of solving a wide range of problems. Evaluations on multiple benchmarks demonstrate that InfiAgent achieves 9.9\% higher performance compared to ADAS (similar auto-generated agent framework), while a case study of the AI research assistant InfiHelper shows that it generates scientific papers that have received recognition from human reviewers at top-tier IEEE conferences.

cs.AI↗

arXiv2023

Gate-controlled neuromorphic functional transition in an electrochemical graphene transistor

Neuromorphic devices have gained significant attention as potential building blocks for the next generation of computing technologies owing to their ability to emulate the functionalities of biological nervous systems. The essential components in artificial neural network such as synapses and neurons are predominantly implemented by dedicated devices with specific functionalities. In this work, we present a gate-controlled transition of neuromorphic functions between artificial neurons and synapses in monolayer graphene transistors that can be employed as memtransistors or synaptic transistors as required. By harnessing the reliability of reversible electrochemical reactions between C atoms and hydrogen ions, the electric conductivity of graphene transistors can be effectively manipulated, resulting in high on/off resistance ratio, well-defined set/reset voltage, and prolonged retention time. Overall, the on-demand switching of neuromorphic functions in a single graphene transistor provides a promising opportunity to develop adaptive neural networks for the upcoming era of artificial intelligence and machine learning.

physics.app-ph↗

arXiv2021

A Multi-task Deep Feature Selection Method for Brain Imaging Genetics

Using brain imaging quantitative traits (QTs) to identify the genetic risk factors is an important research topic in imaging genetics. Many efforts have been made via building linear models, e.g. linear regression (LR), to extract the association between imaging QTs and genetic factors such as single nucleotide polymorphisms (SNPs). However, to the best of our knowledge, these linear models could not fully uncover the complicated relationship due to the loci's elusive and diverse impacts on imaging QTs. Though deep learning models can extract the nonlinear relationship, they could not select relevant genetic factors. In this paper, we proposed a novel multi-task deep feature selection (MTDFS) method for brain imaging genetics. MTDFS first adds a multi-task one-to-one layer and imposes a hybrid sparsity-inducing penalty to select relevant SNPs making significant contributions to abnormal imaging QTs. It then builds a multi-task deep neural network to model the complicated associations between imaging QTs and SNPs. MTDFS can not only extract the nonlinear relationship but also arms the deep neural network with the feature selection capability. We compared MTDFS to both LR and single-task DFS (DFS) methods on the real neuroimaging genetic data. The experimental results showed that MTDFS performed better than both LR and DFS in terms of the QT-SNP relationship identification and feature selection. In a word, MTDFS is powerful for identifying risk loci and could be a great supplement to the method library for brain imaging genetics.

q-bio.GN↗

arXiv2020

Large Transport Gap Modulation in Graphene via Electric Field Controlled Reversible Hydrogenation

Graphene is of interest in the development of next-generation electronics due to its high electron mobility, flexibility and stability. However, graphene transistors have poor on/off current ratios because of the absence of a bandgap. One approach to introduce an energy gap is to use hydrogenation reaction, which changes graphene into insulating graphane with sp3 bonding. Here we show that an electric field can be used to control conductor-to-insulator transitions in microscale graphene via a reversible electrochemical hydrogenation in an organic liquid electrolyte containing dissociative hydrogen ions. The fully hydrogenated graphene exhibits a lower limit sheet resistance of 200 Gohm/sq, resulting in graphene field-effect transistors with on/off current ratios of 10^8 at room temperature. The devices also exhibit high endurance, with up to one million switching cycles. Similar insulating behaviours are also observed in bilayer graphene, while trilayer graphene remains highly conductive after the hydrogenation. Changes in the graphene lattice, and the transformation from sp2 to sp3 hybridization, is confirmed by in-situ Raman spectroscopy, supported by first-principles calculations.

cond-mat.mtrl-sci↗