New preprint on LLM ownership protection
In InstructionalFingerprint, we present a pilot study on LLM fingerprinting as a form of very lightweight instruction tuning. Model publisher specifies a confidential private key and implants it as an instruction backdoor that causes the LLM to generate specific text when the key is present. Results on 11 popularly-used LLMs showed that this approach is lightweight and does not affect the normal behavior of the model. It also prevents publisher overclaim, maintains robustness against fingerprint guessing and parameter-efficient training, and supports multi-stage fingerprinting akin to MIT License.
New preprint on bias mitigation
In BMBI, we propose to mitigate bias exhibited in QA models by observing the query instance’s influence on another instance, enabling bias mitigation with extremely low resources. With our method, bias levels in multiple bias categories can be reduced without using category-specific instance-level annotation.
New preprints on data generation with LLMs and LLM's backdoor attack
In STAR, we propose to synthesize training data by structure-to-text generation using Large Language Models and we show that the generated data is even more effective than human-curated data instances to boost the low-resource event extraction performance.
In the new study about LLM’s backdoor attack, we demonstrate that an attacker can inject backdoors by issuing very few malicious instructions and control model behavior through data poisoning.
New preprint on clinical event extraction
In DICE, we introduce a data-efficient generative model for clinical event extraction and propose the first benchmark in the domain.