finding

active

finding:single-dendritic-layer-solves-xor-like-problems-with-capacity-matching-8-layer-deep-networks

Single dendritic layer solves XOR-like problems with capacity matching 8-layer deep networks.

Evidence from Beniaguev et al. (2021) that individual biological neurons vastly outperform McCulloch-Pitts model; supports hybrid computation claim.

Neighborhood — ranked by edge-count

Thinkers (1)

thinker

Beniaguev et al. (2021)
authored
Authors of dendritic computation study cited as evidence for individual neuron computational power.

Communities (2)

community

Few-shot anchoring & latent structure
members_of
How minimal examples disambiguate and recruit latent arithmetic/reasoning interpretations in LLMs
Mid-layer representation geometry in neural networks
members_of
Studies how internal layer-wise geometric properties (anchoring, clustering trajectories, geometry summaries) peak in middle layers and predict downstream task performance across LLMs and shallow networks.

Concepts (1)

concept

Hybrid Discrete-Continuous Computation
supports
Brains perform simultaneous discrete operations (spikes) and continuous operations (graded potentials, fields) that co-determine each other; distinguishes biological from digital substrates.

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.

Cells and cellular networks have problem-solving capacities that allow them to navigate entirely novel stressors (e.g., barium) without prior selection, implying a general competency beyond hardwired responses.claim0.760
Derived from the planarian barium adaptation finding.
Cross-layer superposition is a fundamental challenge for dictionary learning.claim0.755
Features smeared across layers cannot be fully disentangled by SAE on a single residual stream.
2D projections of activations show clearly separable clusters for F0-F2 and A1 at layer 25, but increasingly entangled activations for F4-F5 and A2-A3.finding0.754
Visual geometric evidence for the fundamental entanglement of true/false activations in harder tasks.
Representing non-linearly separable functions requires a network with multiple layers.claim0.748
Architectural requirement from machine learning.
InceptionV1 implements a four-layer circuit for pose-invariant dog head detection with mirrored left/right pathways that inhibit each other then unite, exhibiting XOR-like propertiesfinding0.746
Evidence that neural networks learn sophisticated invariance mechanisms through structured circuits rather than loose feature aggregation
Layer 27 (last layer) has largest projection magnitude on the reflection direction among all attention head layers in DeepSeek-R1-Qwen-1.5Bfinding0.742
Attribution finding suggesting the last layer directly controls reflection keyword generation
All cases satisfying Criteria 1 and 2 (two out of three) originate from deeper transformer layers and/or the 2/3 layer of LLMs.finding0.741
Consistent with literature that deeper layers encode semantic information and align with human brain activity.
No single layer is universally optimal for probing truth directions; different tasks peak at different layers.claim0.740
Argues against the single-layer analysis approach of prior work.