Optimal temporal context length

Determine the optimal fMRI temporal-context length for the NOEMA decoder rather than restricting the receptive field to the evaluated seven-frame window.

Background

NOEMA is evaluated with a seven-frame receptive field, and the temporal-context ablations compare window lengths up to seven frames. Although performance improves over the tested range for NOEMA, these experiments do not identify whether a longer or otherwise different context would produce better decoding.

The paper therefore leaves unresolved the optimal amount of temporal context for aligning fMRI activity with multimodal speech representations.

References

Moreover, the temporal window is bounded by the architecture's seven-frame receptive field, and the current results do not establish an optimal context length.

— Decoding Naturalistic Speech from fMRI with Temporal Modeling and Multimodal Alignment  (2610.22049 - Caris et al., 15 Oct 2026) in Section Conclusion, subsection “Limitations and future work”