concept
active
concept:high-low-frequency-detectorHigh-Low Frequency Detector
A less intuitive feature family detecting low-frequency patterns on one side of the receptive field and high-frequency on the other; used as example of non-obvious but understandable features
Neighborhood — ranked by edge-count
Thinkers (1)
thinker
- Ludwig SchubertstudiesCo-author; performed detailed analysis of high-low frequency detectors
Concepts (1)
concept
- Curve Detectoranalogous_toA family of neurons found in every non-trivial vision model that detect curved lines in different orientations; primary example used to demonstrate feature understanding
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Second low-level feature type demonstrating cross-architecture universality
- Specific cross-domain prediction mentioned by neuroscientists in conversation with the authors
- 26 SAE latents identified as differentially activated during off-topic content and causally linked to ESR
- Functional role of SOHMs, detecting symmetries in input to compress information.
- Edgerton and Killian's technique for capturing microsecond-scale processes (milk drop splash, glass shattering) revealing smooth structural transitions invisible at normal timescales
- Causal intervention clamping 26 identified OTD latents to zero during steered inference to test ESR contribution
- Epistemic caution about over-interpreting the OTD label given the heterogeneity of identified latents
- Simple stimuli that evoke complex, intelligent responses because the receiver interprets them.