<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Code Generation on Software Engineering Group</title><link>https://seg.inf.unibe.ch/keywords/code-generation/</link><description>Recent content in Code Generation on Software Engineering Group</description><generator>Hugo</generator><language>en-US</language><copyright>Software Engineering Group (SEG), [Institute of Computer Science](https://www.inf.unibe.ch/), [University of Bern](https://www.unibe.ch/). All rights reserved.</copyright><atom:link href="https://seg.inf.unibe.ch/keywords/code-generation/index.xml" rel="self" type="application/rss+xml"/><item><title>Predicting Code Generation Failures from Language Model Attention</title><link>https://seg.inf.unibe.ch/teaching/student-projects/current/t0098-predicting-code-generation-through-probing/</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://seg.inf.unibe.ch/teaching/student-projects/current/t0098-predicting-code-generation-through-probing/</guid><description>Language models can generate plausible code that fails its tests. This thesis investigates whether internal attention patterns predict failure before generation finishes, allowing computation to shift toward more promising candidates. The student will build on the Prober repository to train and evaluate lightweight prediction models.</description></item></channel></rss>