Audio Language Models
PulseAugur coverage of Audio Language Models — every cluster mentioning Audio Language Models across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
New Audits Reveal How Audio LLMs Use Paralinguistic Cues
A new research paper introduces counterfactual audits to evaluate whether audio language models (ALMs) truly utilize paralinguistic cues like affect and prosody, or if they rely solely on transcripts. The study found th…
-
ZEBRA framework enhances Audio-Language Model generalization
Researchers have developed ZEBRA, a novel framework designed to improve the generalization capabilities of Audio-Language Models (ALMs). ZEBRA addresses the trade-off where prompt learning, while enhancing performance o…
-
New framework enhances in-context learning for clinical audio diagnosis
Researchers have developed a new framework called Federated Self-Contextualization (FSC) designed to improve in-context learning for audio-language models in clinical settings, particularly in low-resource environments.…
-
New framework enhances audio language models with trainable audio prompts
Researchers have developed a new framework for Audio Language Models (ALMs) that introduces trainable prompts directly into the audio encoder. This approach aims to capture task-specific acoustic features, enhancing few…
-
Audio language models improve speech emotion recognition with acoustic cues
Researchers have developed a method to improve speech emotion recognition in audio language models by incorporating explicit acoustic cues. By deriving six interpretable acoustic concept tokens from paralinguistic featu…
-
Audio-language models override clear audio with conflicting text
Researchers have identified a significant issue in audio-language models where conflicting text inputs override clear audio evidence, leading to incorrect outputs. A new study reveals that in 64.1% of conflict cases acr…
-
New tools enhance audio deepfake detection and analysis
Researchers have developed new tools and methods to combat audio deepfakes. AUDDT is an open-source toolkit designed to evaluate the generalization capabilities of deepfake detectors across a wide array of audio dataset…
-
New PitchBench Benchmark Reveals Unreliable Pitch Hearing in Audio-Language Models
Researchers have developed PitchBench, a new evaluation suite designed to systematically measure the pitch perception abilities of audio-language models (ALMs). The suite includes 28 experiments that test both absolute …
-
Researchers warn AI voice assistants vulnerable to hidden audio commands
Researchers have identified a significant security vulnerability in AI voice assistants and audio-language models. These systems, increasingly used as everyday interfaces, can be manipulated through imperceptible audio …
-
New architecture boosts audio language models' attention to salient sounds
Researchers have developed NAACA, a novel architecture designed to improve how audio language models process long audio recordings. NAACA uses a training-free approach with an Oscillatory Working Memory (OWM) to filter …
-
New AI method automates coding of therapy sessions
Researchers have developed a new method for automatically coding Motivational Interviewing (MI) sessions using audio-language models (ALMs). This approach analyzes both spoken words and acoustic cues, integrating predic…