Papers by San Kim
Merging Triggers, Breaking Backdoors: Defensive Poisoning for Instruction-Tuned Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are vulnerable to backdoor attacks, where adversaries poison a small subset of data to implant hidden behaviors. |
| Approach: | They propose a training pipeline that immunizes instruction-tuned LLMs against backdoor attacks. |
| Outcome: | The proposed defenses lower attack success rates while preserving instruction-following ability. |
PanicToCalm: A Proactive Counseling Agent for Panic Attacks (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing models for training such models are limited due to ethical and logistical issues. |
| Approach: | They propose a dataset that includes high-distress episodes constructed from first-person narratives and structured around the principles of Psychological First Aid. |
| Outcome: | The proposed model outperforms baseline models in counselor-side metrics and client affect improvement. |
Safeguarding RAG Pipelines with GMTP: A Gradient-based Masked Token Probability Method for Poisoned Document Detection (2025.findings-acl)
Copied to clipboard
| Challenge: | Retrieval-Augmented Generation (RAG) provides external knowledge for accurate and up-to-date responses, but external knowledge is vulnerable to poisoning and unauthorized injections. |
| Approach: | They propose a Gradient-based Masked Token Probability defense method to detect and filter out adversarially crafted documents by examining gradients of the retriever’s similarity function. |
| Outcome: | Experiments show that the proposed method eliminates over 90% of poisoned content while retaining relevant documents. |
Adversarial DPO: Harnessing Harmful Data for Reducing Toxicity with Minimal Impact on Coherence and Evasiveness in Dialogue Agents (2024.findings-naacl)
Copied to clipboard
| Challenge: | Existing toxicity within large language models can negatively impact the user experience, causing performance degradation. |
| Approach: | They propose an adversarial DPO algorithm that improves direct preference optimization (DPO) by incorporating harmful data into the generative model. |
| Outcome: | The proposed training algorithm improves the model’s resilience against harmful conversations while minimizing performance degradation. |
BPM_MT: Enhanced Backchannel Prediction Model using Multi-Task Learning (2021.emnlp-main)
Copied to clipboard
| Challenge: | Backchannel (BC) is a short and quick reaction signal of a listener to a speaker's utterances. |
| Approach: | They propose a model that utilizes lexical information in utterances to enhance backchannel (BC) prediction. |
| Outcome: | The proposed model showed 14.24% performance improvement compared to baseline in the four BC categories: continuer, understanding, empathic response, and No BC. |
A Model of Cross-Lingual Knowledge-Grounded Response Generation for Open-Domain Dialogue Systems (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Existing studies on open-domain dialogue systems that allow free topics are challenging . however, non-English dialogue systems suffer from reproducing the performance of English dialogue systems . |
| Approach: | They propose to use English knowledge to improve the performance of open-domain dialogue systems . they construct a Korean-English T5 language model and develop a knowledge-grounded Korean dialogue model . |
| Outcome: | The proposed model improves even when only English knowledge is given . the model is built with a pre-trained language model and a knowledge-grounded Korean dialogue model . |