Minglai Yang

AlignSAE: Concept-Aligned Sparse Autoencoders

Minglai Yang, Xinyu Guo

Large Language Models (LLMs) encode factual knowledge within hidden parametric spaces that are difficult to inspect or control. While Sparse Autoencoders (SAEs) can decompose hidden activations into more fine-grained, interpretable features, they often struggle to reliably align these features with human-defined concepts, …

How Is LLM Reasoning Distracted by Irrelevant Context? An Analysis Using a Controlled Benchmark

Minglai Yang, Ethan I. Huang, Liang Zhang, Mihai Surdeanu et autres

We introduce Grade School Math with Distracting Context (GSM-DC), a synthetic benchmark to evaluate Large Language Models' (LLMs) reasoning robustness against systematically controlled irrelevant context (IC). GSM-DC constructs symbolic reasoning graphs with precise distractor injections, enabling rigorous, reproducible evaluation. Our experiments demonstrate …

CopySpec: Accelerating LLMs with Speculative Copy-and-Paste Without Compromising Quality

Razvan-Gabriel Dumitru, Minglai Yang, Vikas Yadav, Mihai Surdeanu

We introduce CopySpec, a simple yet effective technique to tackle the inefficiencies LLMs face when generating responses that closely resemble previous outputs or responses that can be verbatim extracted from context. CopySpec identifies repeated sequences in the model's chat history or context …

CopySpec: Accelerating LLMs with Speculative Copy-and-Paste

Razvan-Gabriel Dumitru, Minglai Yang, Vikas Yadav, Mihai Surdeanu

We introduce CopySpec, a simple yet effective technique to tackle the inefficiencies LLMs face when generating responses that closely resemble previous outputs or responses that can be verbatim extracted from context.CopySpec identifies repeated sequences in the model's chat history or context and …

us (code pays fourni par la source)

How Is LLM Reasoning Distracted by Irrelevant Context? An Analysis Using a Controlled Benchmark

Minglai Yang, Ethan I. Huang, Liang Zhang, Mihai Surdeanu et autres

We introduce Grade School Math with Distracting Context (GSM-DC 1 ), a synthetic benchmark to evaluate Large Language Models' (LLMs) reasoning robustness against systematically controlled irrelevant context (IC).GSM-DC constructs symbolic reasoning graphs with precise distractor injections, enabling rigorous, reproducible evaluation.Our experiments demonstrate …

us, cn (code pays fourni par la source)

Intelligent Control and Data Analysis System Based on Internet of Things

Guangming Luo, Minglai Yang, Zhenfeng Cao

This paper designs a set of intelligent regulation and growth analysis system based on the Internet of things, which aims to summarize a set of application scheme which integrates data analysis and intelligent regulation. So as to solve the problems of insufficient …

cn (code pays fourni par la source)

Design of Embedded Terminal Data Remote Transmission Control System in Internet of Things

Weiqing Fu, Minglai Yang

Abstract The underlying data exchange system of the Internet of Things is an integrated system for data collection and information transmission of the Internet of Things. The design method of the underlying data exchange system of the Internet of Things based on …

cn (code pays fourni par la source)

Analysis and Research of Communication Interrupt Fault for Shanghai Metro Data Transmission System

Jianru Liang, Xinyuan Lu, Cong Shi, Minglai Yang

A line of Shanghai metro has been put into use for nearly fifteen years. There are three times extended during this time. The existing line’s data transmission system was modified over the last decades and has adopted many kinds of data transmission …

cn (code pays fourni par la source)