{"slug": "circuittree-composing-circuits-into-a-syllable-transcriber", "title": "CircuitTree: Composing Circuits into A Syllable Transcriber", "summary": "Researchers developed CircuitTree, a data structure built on a Θ(N log N) tree-compose schedule, to compose arbitrarily many neural circuits into a single model, according to an arXiv paper (2610.12520v1). Composing circuits pruned from HuBERT for classifying individual phonemes yielded a syllable classifier, a capability none of the 39 phoneme circuits has individually. Chaining CircuitTrees into a hierarchy-of-circuits pipeline transcribed open syllables at 64.8% accuracy, within 1.2 points of a 66.0% ceiling that shares no circuitry at all, with 99.5% of circuit components shared.", "body_md": "arXiv:2610.12520v1 Announce Type: new \nAbstract: The goal of circuit discovery is to decompose a monolithic neural network into tiny modular subnetworks which are interpretable in the sense that each does an identifiable subtask. We develop a practical method for composing arbitrarily many circuits into a single model. We propose a greedy, density-scaled composition algorithm that overcomes limitations of previous attempts to compose circuits by naive union, then generalize it to $N$ circuits via CircuitTree, a new data structure built on a $\\Theta(N\\log N)$ tree-compose schedule. We compose circuits pruned from HuBERT for classifying individual phonemes into a syllable classifier, a capability none of the 39 phoneme circuits has individually. Chaining CircuitTrees into a hierarchy-of-circuits pipeline, we transcribe open syllables at 64.8\\% accuracy, within 1.2 points of a 66.0\\% ceiling that shares no circuitry at all, with 99.5\\% of circuit components shared.", "url": "https://wpnews.pro/news/circuittree-composing-circuits-into-a-syllable-transcriber", "canonical_source": "https://www.machinebrief.com/news/circuittree-composing-circuits-into-a-syllable-transcriber-57q8", "published_at": "2026-10-12 04:00:00+00:00", "updated_at": "2026-10-12 04:06:23.156388+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "natural-language-processing", "ai-research", "neural-networks"], "entities": ["CircuitTree", "HuBERT", "arXiv"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/circuittree-composing-circuits-into-a-syllable-transcriber", "markdown": "https://wpnews.pro/news/circuittree-composing-circuits-into-a-syllable-transcriber.md", "text": "https://wpnews.pro/news/circuittree-composing-circuits-into-a-syllable-transcriber.txt", "jsonld": "https://wpnews.pro/news/circuittree-composing-circuits-into-a-syllable-transcriber.jsonld"}}