finding

active

finding:the-likelihood-of-a-dedicated-feature-for-a-concept-element-city-animal-food-follows-a-sigmoid-in-log-frequency-of-the-concept-in-training-data-with-threshold-frequency-inversely-proportional-to-number-of-alive-features

The likelihood of a dedicated feature for a concept (element, city, animal, food) follows a sigmoid in log-frequency of the concept in training data, with threshold frequency inversely proportional to number of alive features.

Quantitative relationship between concept frequency and feature presence.

Source paper

extracted_from

Scaling monosemanticity: Ex-tracting interpretable features from claude 3 sonnet

Neighborhood — ranked by edge-count

Claims (1)

claim

There appears to be a systematic relationship between the frequency of concepts and the dictionary size needed to resolve features for them.
supports
Feature presence depends on concept frequency in training data, with a threshold scaling inversely with alive features.

Questions (1)

question

does the model have a feature corresponding to every major world city?
answered_by
Question explored in feature completeness study.

Related by similarity (8)

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

Features are connected by weights forming circuits, and these circuits can be rigorously studied and understood as meaningful algorithms.claim0.785
Second of three speculative claims asserting that subgraphs of neural networks are tractable and meaningful objects of study
what is the precise characterization of the perceptual representational format in which information exists as a specific, sustained, globally constitutive structure?question0.776
Open theoretical problem CIMC acknowledges: precisely characterizing the representational format of perception
Optimal number of features scales faster than optimal number of training steps with compute budget.finding0.774
Allocation result from scaling laws.
The segmentation clock exhibits functional robustness consistent with James' definition of intelligence.claim0.769
Claims that the re-emergence and adaptation of the segmentation clock qualifies as intelligent behavior.
The multiscale competency architecture (MCA) speeds evolutionary search by providing generalization, reliability, tractable search space, cryptic variation, and functional intermediates.claim0.768
Main functional claim about MCA.
Features are the fundamental unit of neural networks; they correspond to directions and can be rigorously studied and understood.claim0.768
First of three speculative claims forming the foundation of the circuits research agenda
Hypothesis 1 (Threshold Behavior): There exists a task-dependent threshold Sc such that performance exhibits sharp changes as S crosses Sc, with value and transition width depending on model, layer, and poolinghypothesis0.768
Core testable hypothesis of UCCT about the nature of performance transitions under anchoring
Four features will be fundamental: (1) all processes rethought as morphogenetic; (2) sequences link into a net; (3) continuous evolution becomes widespread; (4) an ethical obligation to heal the land emerges.claim0.768
Vision of the emerging paradigm shift in society.