15 citations · 16 across the 20 of their papers we have counts for
5 papers · 1 filter
Self-Improving In-Context Learning
Baturay Saglam, Dionysis Kalogerias
We propose to improve in-context learning (ICL) by optimizing the continuous embeddings of a fixed few-shot prompt at test time. The key observation is that the log-probabilities a…
Test-Time Safety Alignment
Baturay Saglam, Dionysis Kalogerias
Recent work has shown that a model's input word embeddings can serve as effective control variables for steering its behavior toward outputs that satisfy desired properties. Howeve…
Test-Time Detoxification without Training or Learning Anything
Baturay Saglam, Dionysis Kalogerias
Large language models can produce toxic or inappropriate text even for benign inputs, creating risks when deployed at scale. Detoxification is therefore important for safety and us…
Large Language Models Encode Semantics and Alignment in Linearly Separable Representations
Baturay Saglam, Paul Kassianik, Blaine Nelson +3
Understanding the latent space geometry of large language models (LLMs) is key to interpreting their behavior and improving alignment. Yet it remains unclear to what extent LLMs li…
Learning Task Representations from In-Context Learning
Baturay Saglam, Xinyang Hu, Zhuoran Yang +2
Large language models (LLMs) have demonstrated remarkable proficiency in in-context learning (ICL), where models adapt to new tasks through example-based prompts without requiring…