AI Paper Atlas

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

Beyond Transcription: Mechanistic Interpretability in ASR

Neta Glazer, Yael Segal-Feldman, Hilit Segev, Aviv Shamsian, Asaf Buchnick, Gill Hetz, Ethan Fetaya, Joseph Keshet, Aviv Navon

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Task vectors and steering · 2 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
probe internal encoder representation instead of reading the text outputQwen2-Audio · speaker gender, LibriSpeechaccuracy (%): 87.8 -> 94.6Improved
patch cross-attention at decoder layer 23 (ablation)Whisper · repetition hallucinationscases resolved (%): 76%Improved
patch decoder self-attention or feed-forward layers (ablation)Whisper · repetition hallucinationsrepetition rateNo clear effect
patch cross-attention head 13 in layer 18 alone (ablation)Whisper · repetition hallucinationsrepetition suppressed (%): 78.1%Improved
patch encoder components with white noise (ablation)Whisper · acoustic-contextual errorsrestored accuracy: 130/153 (85.0%) restoredImproved
patch decoder components with white noise (ablation)Qwen2-Audio · acoustic-contextual errorsrestored accuracy: 147/251 (58.6%) restored vs 171/251 via encoderImproved
apply final encoder layer normalization before passing intermediate layers to decoder (Encoder Lens) (ablation)Whisper · encoder lensoutput coherenceImproved
probe decoder residual stream at later layers for hallucination predictionWhisper · LibriSpeech test-cleanprobe test accuracy: 0.622 (L5) -> 0.934 (L22)Improved
probe decoder residual stream for hallucination on LibriSpeechQwen2-Audio vs Whisper · LibriSpeechprobe accuracy (%): 93.4 -> 70.2Hurt

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