🚨 New paper alert: A new generation of LLMs can now process speech natively. This could expand access for millions excluded by text interfaces, but our research shows a cost: demographic cues in speaker voice can trigger stereotypical model responses. 🎙️⚖️
Paper: arxiv.org/abs/2603.22260
A huge thank you to all the amazing authors of this work!! @minhducbui.bsky.social @kaitlynzhou.bsky.social @valentinhofmann.bsky.social @a-lauscher.bsky.social
Finally, I will be with @carolin-holtermann.bsky.social and @a-lauscher.bsky.social to present our work on evaluating geotemporal reasoning ability in LLMs. This will be in the Wednesday 1100 poster session:
aclanthology.org/2025.acl-lon...
🔍 Study 1: Do models treat voices differently? Yes. Across audio-enabled LLMs, voice inputs drove measurable gender-stereotyped outputs. Higher gender detection accuracy = stronger bias.
🛠️ Study 3: Mitigation exists. Discrimination tracks continuously with vocal pitch. Systematically lowering pitch reduced bias to statistically insignificant levels across all tested models.
The result is a self-reinforcing gap: systems optimize for frequent users, while growing more distant from those they were supposed to reach. Fairness and accessibility have to be tackled together.
📣 CFP: Eval4SD @ KONVENS 2026, Hamburg (Sept 14–17)
1st workshop on evaluating LLMs for specialized domains. We invite work on benchmarking, replication, and evaluation methodology across fields such as medicine, law, science, finance, and more.
Deadline: Jul 3, 2026, 23:59 CEST
eval4sd.github.io
Paper link: arxiv.org/abs/2603.22260
Website: tai-hamburg.github.io/papers/diver...
⚠️ Study 2: Real consequences. Voice AI is meant to lower barriers — but non-users are 5× more likely to disengage than daily users when they learn these systems infer attributes from their voice.
🧵Excited to share that Data Science Lab Hamburg will be presenting 6 papers at #ACL2025 in Vienna! 🇦🇹 Our team has been working on cutting-edge research in multilingual AI, peer review analysis, and combating online harms. Here's what we're bringing to the conference 👇
