READ: Reading the Selective Scan for Faithful Video Attribution
Selective state-space models are an efficient approach to long-sequence video modeling.
ProofPaper ↗
Key points
- In this work, we introduce READ, a parameter-free attribution method for selective state-space models that unrolls the recurrence at the position read by the classifier.
- READ traces the contribution accumulated in the recurrent state back to the video tokens that contribute to the final prediction.
- Experimental results show that READ avoids the systematic relevance biases of existing methods while recovering meaningful temporal and class-specific structure across datasets and backbones.
- More broadly, our results show that the recurrence underlying selective state-space models is not only an efficient mechanism for inference, but also a direct source of interpretable evidence for their predictions.
Sources (1)
- [1]READ: Reading the Selective Scan for Faithful Video AttributionarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 03:59 PM
Selective state-space models are an efficient approach to long-sequence video modeling.
In this work, we introduce READ, a parameter-free attribution method for selective state-space models that unrolls the recurrence at the position read by the classifier.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 9, 2026State-Isolated Vision Mamba via Update-Tendency Routing for Underwater Image Enhancement
- Oct 9, 2026Understanding Mamba-3 Through Oscillatory Dynamics
- Oct 9, 2026Scan-Parallel Variational Inference for event-driven Latent Jump Models
- Oct 9, 2026Selective Spatio-Temporal Fruit Inspector (S$^2$FI): A Mamba-based Framework for Rotational Fruit Quality Assessment
- Oct 9, 2026Time Delay System for Sequence Modeling
- Oct 8, 2026Bridging Human and Model Attention: Explainability Analysis of CNN, Mamba, and ViT Architectures with Gaze-Based Validation