AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Closing the Gap Between Text and Speech Understanding in LLMs

Santiago Cuervo, Skyler Seto, Maureen de Seyssel, Richard He Bai, Zijin Gu, Tatiana Likhomanenko, Navdeep Jaitly, Zakaria Aldeneh

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Speech and audio language models · 15 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
train with distillation (alpha>0) instead of NLL on narrow speech data (ablation)speech-adapted Qwen2.5-3B · LibriHeavy+Emiliacross-modal misalignmentImproved
increase distillation weight alpha (ablation)speech-adapted Qwen2.5-3B · LibriHeavy+Emiliairreducible misalignment E: 0.32 -> 0.13Improved
use broad-domain FineWeb-Edu synthetic speech with NLL (alpha=0) (ablation)speech-adapted Qwen2.5-3B · FineWeb-Edu testmisalignmentMixed
add Stage II active selection after Stage I (ablation)SALAD-3B · Spoken benchmarks avgaccuracy: 68.5 -> 72.0Improved
add Stage II active selection after Stage I (ablation)SALAD-7B · Spoken benchmarks avgaccuracy: 73.9 -> 75.4Improved
active selection instead of uniform selection (ablation)SALAD-3B · OBQAaccuracy: 71.9 -> 76.7Improved
active selection instead of uniform selection (ablation)SALAD-3B · PIQAaccuracy: 78.1 -> 78.1No clear effect
use SALAD instead of Qwen2.5-Omnispeech-adapted LLM · Spoken benchmarks avgaccuracy: 76.7 vs 75.4 (7B)Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy