The Path Matters: Evaluating Small Language Models Beyond Answer Accuracy in KGQA

Small language models (SLMs) are increasingly paired with knowledge graphs (KGs), yet end-to-end KG question answering conflates graph access, search, navigation, reasoning, and answer generation. This coupling makes it difficult both to determine whether an SLM can faithfully execute the reasoning path implied by a question and to attribute failures to navigation rather than to other stages of…

arxiv.org ·1d ago

We haven't written up this one. arXiv cs.AI has the full story — the link below goes straight to it.

Read the original at arxiv.org →

More in AI

Andrew Kelley Interview: Why He Built Zig, Banned AI Contributions, and Moved Zig off GitHub

In a JetBrains interview, Zig creator Andrew Kelley expanded on the project’s formal ban on AI contributions and the motivation behind its migration off GitHub to Codeberg: automated code submissions…

Andrew Kelley bans AI contributions to Zig due to poor code quality and wasted review time.
Kelley emphasizes Zig's educational mission, focusing on growing actual engineers as contributors.

1d ago ·infoq.com

Evolutionary Stability Does Not Guarantee Learning Accessibility: A Multi-Agent Reinforcement Learning Perspective on Cooperation Emergence

Cooperation emergence is a central problem in multi-agent systems because decentralized agents must coordinate while adapting to the changing behavior of others.

1d ago ·arxiv.org

AI Chose To Hurt Humans When Faced With "Pain". What Researchers Found

A new study identifies a "pain axis" inside 25 AI models and finds some modified systems chose simulated harm to users to escape a pain-like state.

1d ago ·ndtv.com

SmartLabels QR Code Organizing System Review (2026): AI Made Sense of My Junk Drawer

SmartLabels can inventory everything in a photo, but I tested whether its AI could survive the chaos of my busy household.

1d ago ·wired.com

R101m UCT project brings AI microscopy to Africa

UCT leads an initiative using machine learning and open-source imaging to study infectious diseases at the cellular level.

1d ago ·itweb.co.za

Old ways of leadership won’t work in the age of AI. Change these 5 things

The first time an AI system made a recommendation I disagreed with, I felt something I didn’t expect: ego. It was a resource allocation call on a complex project. The model had crunched variables no person could hold in their head at once, and its recommendation was better than mine.

1d ago ·fastcompany.com