Thesis: Tracing Input Causality in LLM Agents

AI Sweden; Göteborgs kommun; Hybrid; SEK 234,000 - 301,000; Part time

Benefits offered by this job

Hybrid work

Remote work from Lund/Stockholm

Job summary

AI Sweden in Gothenburg invites a curious MSc student to join a thesis project at the frontier of attribution for agent traces. You will build a test set with planted causes and implement a simple agent that searches documents, calls tools, and remembers past sessions.

The role emphasizes reproducible experiments, evaluating attribution methods, and collaborating with leading AI scientists. Hybrid work is available, with remote options for students in Lund or Stockholm.

Qualifications

• Currently enrolled in a Master’s program (MSc) in a relevant field.
• Proficient with Python and deep learning.
• Curious, independent, and able to work on research questions.

Responsibilities

• Contribute to a thesis project on attribution for agent traces.
• Build a test set with planted causes and verify leak paths.
• Develop a simple agent that searches documents, calls tools, and remembers past sessions; evaluate attribution methods.

Skills

Python

Deep learning

Research

Education

MSc student

Job description

AI Sweden in Gothenburg invites a curious MSc student to join a thesis project at the frontier of attribution for agent traces. You will build a test set with planted causes and implement a simple agent that searches documents, calls tools, and remembers past sessions.

The role emphasizes reproducible experiments, evaluating attribution methods, and collaborating with leading AI scientists. Hybrid work is available, with remote options for students in Lund or Stockholm.