Benign fine-tuning on audio data breaks safety alignment in Audio LLMs by raising jailbreak success rates up to 87%, with the dominant risk axis depending on model architecture and embedding proximity to harmful content.
arXiv:2309.07045 (2023), https://arxiv.org/abs/2309.07045
8 Pith papers cite this work. Polarity classification is still indexing.
8
Pith papers citing it
representative citing papers
VoxSafeBench reveals that speech language models recognize social norms from text but fail to apply them when acoustic cues like speaker or scene determine the appropriate response.
Toxicity benchmarks for LLMs produce inconsistent results when task type, input domain, or model changes, revealing intrinsic evaluation biases.
LLM judge prompt variations alone shift HarmBench harmful-response rates by up to 24.2 percentage points and produce moderate instability in model safety rankings.
A formalization of benchmarkless LLM safety scoring validated via an instrumental-validity chain of contrast separation, target variance dominance, and rerun stability, demonstrated on Norwegian scenarios.
Mainstream conversational models show escalating affective misalignments and ethical guidance failures during staged emotional trajectories, organized into a taxonomy of interactional breakdowns.
GLM-4.5, a 355B-parameter MoE model with hybrid reasoning, scores 70.1% on TAU-Bench, 91.0% on AIME 24, and 64.2% on SWE-bench Verified while ranking 3rd overall and 2nd on agentic benchmarks.
GLM-4 models rival or exceed GPT-4 on MMLU, GSM8K, MATH, BBH, GPQA, HumanEval, IFEval, long-context tasks, and Chinese alignment while adding autonomous tool use for web, code, and image generation.
citing papers explorer
-
VoxSafeBench: Not Just What Is Said, but Who, How, and Where
VoxSafeBench reveals that speech language models recognize social norms from text but fail to apply them when acoustic cues like speaker or scene determine the appropriate response.