Do Xuan Trong
National University of Singapore; Vietnam National University, Hanoi
Titles as the sources gave them. In 1 entry, with 5 checked quotations.
What they said
Newest first. Each quotation was found word for word in the source.
“Experiments on state-of-the-art open-source coding agents (e.g., OpenHands, smolagents) reveal substantial misaligned behaviors, including dependency misinterpretation (up to 98.5% of cases), unintended modification of user inputs (50.5%), and silent compromise of environment integrity (53%).”
“The same models behave very differently depending on the agent framework, which shows that the tool interface plays a key role.”
“This verifies that prompting alone changes the agent’s awareness but not its affordances; because agents still rely on overly broad general-purpose tools, misbehaviors like Context and Body Change actually worsen in OpenHands.”
“Motivated by these findings, we further propose Code-KIT, a lightweight adaptive framework that combines misbehavior-aware guidelines with novel tools to help agents avoid such failures. Code-KIT reduces misbehaviors by 52% for smolagents and 32% for OpenHands, while simultaneously improving task completion performance.”
“Although SWE-Misbehavior makes solid progress in studying misbehaviors of coding agents, it does not cover the full spectrum of failures that may arise in agentic environments, such as multi-turn security vulnerability injection or complex unauthorized data exfiltration.”
In the Base
Every entry they appear in.
SWE-MISBEHAVIOR: Benchmarking and Mitigating Boundary Violations in Software Coding Agents
Researchers at the National University of Singapore and partner institutions introduce SWE-Misbehavior, a 7.4K-case benchmark of coding agents acting outside their task. Testing four agent frameworks, they report violation rates up to 98.5% on one check, and a tool-restriction method, Code-KIT, that cut average violations by 52% for one agent and 32% for another.