Papers with Llama-3.3-70B-Instruct
In-Context Representation Hijacking (2026.acl-long)
Copied to clipboard
| Challenge: | In-context representation hijacking attacks on large language models are insufficient to prevent harm, yet the mechanisms underlying this behavior remain unclear. |
| Approach: | They propose a simple in-context representation hijacking attack that replaces a harmful keyword with a benign token, and embeds the harmful semantics under a euphemism. |
| Outcome: | The proposed attack is optimized for open-source and achieves strong success rates on closed-source systems, reaching 74% on Llama-3.3-70B-Instruct with a single-sentence context override. |
Learning to Retrieve User History and Generate User Profiles for Personalized Persuasiveness Prediction (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) are increasingly used in decision-support applications that aim to influence human behavior or beliefs, such as health coaching, tutoring, and targeted marketing. |
| Approach: | They propose a context-aware user profiling framework with two trainable components that generate optimal queries to retrieve persuasion-relevant records from a user’s history and a profiler that summarizes these records into a model. |
| Outcome: | The proposed framework raises F1 from 33% to 47% on Llama-3.3-70B-Instruct. |