Entity · paper

Listening with Attention: Entropy-Guided Explainability for Transformer-Based Audio Models

paperactivelistening-with-attention-entropy-guided-explainability-for-transformer-based-audio-models-624892ec·1 events·first seen Jun 15, 2026

Aliases: Listening with Attention: Entropy-Guided Explainability for Transformer-Based Audio Models

Co-occurring entities

LEAF-X Whisper

More like this (12)

Train the Model, Not the Reader: Decodability Supervision for Verifiable Activation Explanations First-Token Broadcasters: Mechanistic Origins of Language Identity and Distributed Robustness in Transformers Localized Adaptation Reveals Distinct Learning Signatures in Transformers Beyond Fully Random Masking: Attention-Guided Denoising and Optimization for Diffusion Language Models Generative Explainability for Next-Generation Networks: LLM-Augmented XAI with Mutual Feature Interactions Cross-seed explainability using Procrustes-conditioned Joint End-to-end Top-K Sparse Autoencoders Sound Probabilistic Safety Bounds for Large Language Models Leveraging Audio-LLMs to Filter Speech-to-Speech Training Data Towards Explainable Adjudicative Variance: Quantifying Judicial Discretion via Gated Multi-Task Learning From Black-Box to Clinical Insight: A Multi-Stage Explainable Framework for Speech-Based Cognitive Impairment Detection Evaluating RE Practices for Explainability: Synthesizing Insights from Daimler Truck into an Explainable RE Framework Proposal Bridging the Gap Between Latent and Explicit Reasoning with Looped Transformers

Recent events (1)

4arXiv · cs.AI·Jun 15, 2026·source ↗

LEAF-X: Entropy-guided explainability framework for transformer-based ASR models

Researchers introduce LEAF-X (Listening with Entropy-guided Attention for Faithful explainability), a model-intrinsic XAI framework for transformer-based automatic speech recognition systems like Whisper. The method combines entropy-guided attention weighting, multi-layer attention rollout, and optional causal ablations to produce sparse token-to-frame attributions. Evaluations show 32% improved faithfulness and 35-39% stronger locality/sparsity compared to perturbation-based explainers and raw attention maps, enabling more auditable ASR.

AI Safety Research Listening with Attention: Entropy-Guided Explainability for Transformer-Based Audio Models LEAF-X Whisper