paper-with-me

Papers

Soteria: Language-Specific Functional Parameter Steering for Multilingual Safety Alignment

2025-02-16 · Somnath Banerjee, Sayan Layek, Pratyush Chatterjee, Animesh Mukherjee, Rima Hazra

Ensuring consistent safety across multiple languages remains a significant challenge for large language models (LLMs). We introduce Soteria, a lightweight yet powerful strategy that locates and minimally adjusts the "functional heads" most responsible for harmful content generation in each language. By altering only a fraction of parameters, Soteria drastically reduces policy violations without sacrificing overall model performance, even in low-resource settings. To rigorously evaluate our approach, we also present XThreatBench, a specialized multilingual dataset capturing fine-grained harmful behaviors drawn from real policy guidelines. Experiments with leading open-source LLMs (e.g., Llama, Qwen, Mistral) show that Soteria consistently improves safety metrics across high-, mid-, and low-resource languages. These findings highlight a promising path toward scalable, linguistically attuned, and ethically aligned LLMs worldwide.

📄 PDF Abstract BibTeX arXiv:2502.11244

Code (1)

neuralsentinel/soteria 공식 구현

Tasks

Safety Alignment

Similar Papers 제목 키워드 기반

SOTERIA: In Search of Efficient Neural Networks for Private Inference

2020-07-25 · Anshul Aggarwal, Trevor E. Carlson, Reza Shokri, Shruti Tople

ML-as-a-service is gaining popularity where a cloud server hosts a trained model and offers prediction (inference) service to users. In this setting, our objective is to protect the confidentiality of both the users' inp…

Neural Architecture Search

From Weights to Activations: Is Steering the Next Frontier of Adaptation?

2026-04-15 · Simon Ostermann, Daniil Gurgurov, Tanja Baeumel, Michael A. Hedderich 외 arxiv

Post-training adaptation of language models is commonly achieved through parameter updates or input-based methods such as fine-tuning, parameter-efficient adaptation, and prompting. In parallel, a growing body of work mo…

Mechanistic Finetuning of Vision-Language-Action Models via Few-Shot Demonstrations

2025-11-27 · Chancharik Mitra, Yusen Luo, Raj Saravanan, Dantong Niu 외 arxiv

Vision-Language Action (VLAs) models promise to extend the remarkable success of vision-language models (VLMs) to robotics. Yet, unlike VLMs in the vision-language domain, VLAs for robotics require finetuning to contend …

Soteria: Provable Defense Against Privacy Leakage in Federated Learning From Representation Perspective

2021-06-19 · CVPR 2021 1 · Jingwei Sun, Ang Li, Binghui Wang, Huanrui Yang 외

Federated learning (FL) is a popular distributed learning framework that can reduce privacy risks by not explicitly sharing private data. However, recent works have demonstrated that sharing model updates makes FL vu…

Federated LearningInference Attack

FBHM: Functional Benchmarking and Steering of VLMs for Hateful Meme Detection

2026-05-29 · Paramananda Bhaskar, Naquee Rizwan, Daksh Jogchand, Saurabh Kumar Pandey 외 arxiv

Hateful meme detection remains a formidable challenge for vision-language models, as existing benchmarks are structurally observational - confounding rhetorical hate mechanisms with target community features and preventi…

Multimodal Reasoning