CodecAttack perturbs audio in codec latent space with multi-bitrate EoT to achieve 85.5% average ASR on Opus-compressed Audio LLMs versus under 26% for waveform baselines, with transfer to MP3 and AAC.
Efficient adversarial training in llms with continuous attacks
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
representative citing papers
LLM safety evaluations are hindered by noise in dataset curation, automated red-teaming, response generation, and LLM-judge evaluation, making fair comparisons difficult and slowing progress.
A survey categorizing LLM-powered agent systems into software-based, physical, and hybrid types, covering industrial applications and challenges such as latency and security.
citing papers explorer
-
Codec-Robust Attacks on Audio LLMs
CodecAttack perturbs audio in codec latent space with multi-bitrate EoT to achieve 85.5% average ASR on Opus-compressed Audio LLMs versus under 26% for waveform baselines, with transfer to MP3 and AAC.
-
LLM-Safety Evaluations Lack Robustness
LLM safety evaluations are hindered by noise in dataset curation, automated red-teaming, response generation, and LLM-judge evaluation, making fair comparisons difficult and slowing progress.
-
LLM-Powered AI Agent Systems and Their Applications in Industry
A survey categorizing LLM-powered agent systems into software-based, physical, and hybrid types, covering industrial applications and challenges such as latency and security.
- REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations
- Unlearning with Asymmetric Sources: Improved Unlearning-Utility Trade-off with Public Data