New metrics KSS and KPS are introduced to evaluate multilingual machine unlearning quality and cross-language consistency in LLMs, addressing limitations of single-language evaluation protocols.
Alignment-aware quantization for LLM safety
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
years
2026 5representative citing papers
Across 51 quantized checkpoints, quality metrics fail to predict safety drops in 36 pairings and 10 hidden-danger cases, while a new RTSI screen routes all 10 dangerous rows to testing at matched bucket size.
The paper introduces a paired testing protocol for batch-conditioned refusal robustness in LLM serving and reports low rates of genuine safety-label flips after adjudication, with a batch-invariant kernel ablation eliminating observed flips.
3-bit quantization induces new stereotypical biases in 6-21% of previously unbiased BBQ items across three LLMs, undetected by perplexity increases under 3%, with models declining in 'unknown' responses by 17.4%.
Self-evolving LLM agents introduce persistent, amplifying security threats that static defenses cannot address, as shown by analysis of 25 attack surface cells and case studies.
citing papers explorer
-
Knowledge Beyond Language: Bridging the Gap in Multilingual Machine Unlearning Evaluation
New metrics KSS and KPS are introduced to evaluate multilingual machine unlearning quality and cross-language consistency in LLMs, addressing limitations of single-language evaluation protocols.
-
Quality Is Not a Safety Proxy Under Quantization
Across 51 quantized checkpoints, quality metrics fail to predict safety drops in 36 pairings and 10 hidden-danger cases, while a new RTSI screen routes all 10 dangerous rows to testing at matched bucket size.
-
A Paired Testing Protocol for Batch-Conditioned Refusal Robustness in LLM Serving
The paper introduces a paired testing protocol for batch-conditioned refusal robustness in LLM serving and reports low rates of genuine safety-label flips after adjudication, with a batch-invariant kernel ablation eliminating observed flips.
-
Quantization Undoes Alignment: Bias Emergence in Compressed LLMs Across Models and Precision Levels
3-bit quantization induces new stereotypical biases in 6-21% of previously unbiased BBQ items across three LLMs, undetected by perplexity increases under 3%, with models declining in 'unknown' responses by 17.4%.
-
Safety in Self-Evolving LLM Agent Systems: Threats, Amplification, and Case Studies
Self-evolving LLM agents introduce persistent, amplifying security threats that static defenses cannot address, as shown by analysis of 25 attack surface cells and case studies.