Fine-tuning-as-a-service enables users to adapt aligned large language models (LLMs) to specialized tasks, but malicious fine-tuning can erode refusal behavior while preserving task performance on legitimate inputs. We revisit recent layer-wise safety diagnostics and find that safety sensitivity is signed: scaling different layers can strengthen refusal, weaken it, or have little effect. Motivated by this observation, we propose SLDR, a post-fine-tuning defense based on Selective Layers Recovery and Dynamic Routing. SLDR trains a LoRA recovery adapter only on the layers with the maximum and minimum sensitivity scores in the signed spectrum, and uses representation-based dynamic routing inference to activate the adapter only for malicious queries. Across four model architectures, five downstream tasks, and four harmful benchmarks, SLDR substantially reduces harmful outputs while preserving downstream utility. On Llama3.1/SST2, SLDR reduces the average harmful score from 11.54 to 0.08 while maintaining downstream accuracy, and the harmful score remains near zero under poisoning ratios up to 0.9. The code is available at https://github.com/Stardust457/SLDR.
GAOKAO-Bench is introduced, an intuitive benchmark that employs questions from the Chinese GAOKAO examination as test samples, including both subjective and objective questions that contribute a robust evaluation benchmark for future large language models and offers valuable insights into the advantages and limitations...
Xiaotian Zhang, Chun-yan Li, Yi Zong et al.· arXiv.org· 216 citations· ⚡17
This work proposes a HALO that directly maximizes the utility of generations instead of maximizing the log-likelihood of preferences, and suggests that the best loss depends on the inductive biases most appropriate for a given setting, an oft-overlooked consideration.
This work investigates the possibilities of using LLMs in a resume screening setting via a document retrieval framework that simulates job candidate selection and finds that the MTEs are biased, significantly favoring White-associated names in 85% of cases and female-associated names in only 11.1% of cases.
This work shows that orders of magnitude enhancement in performance could be obtained by a combination of hardware improvements and tight quantum-HPC integration and introduces high-performance architectures for quantum-probabilistic computing with custom-designed accelerators to tackle today's industry-scale classical...
Masoud Mohseni, Artur Scherer, K. Johnson et al.· arXiv.org· 121 citations· ⚡9
This paper presents a comprehensive overview of the Ultralytics YOLO family, emphasizing architectural evolution, benchmarking, deployment, and emerging directions from YOLOv5 through YOLO27, and examines detection, segmentation, depth, classification, pose, oriented detection, tracking, export, quantization, and deplo...
A novel threat is unveiled in which attackers steer the RAG system's response by injecting malicious passages into its knowledge base, enabling the attacker to steer the response without altering the user input or modifying the RAG weights.
Jiaqi Xue, Meng Zheng, Yebowen Hu et al.· arXiv.org· 109 citations· ⚡8
Machine-learning networks can learn without detailed instructions from a designer. Those that do may offer insight into how the brain reorganizes itself during learning.
With $2.1 million funding from Google.org, the open-source Public Transit Intelligence Hub will unify public transit monitoring, operations, and passenger communication.
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
