finding
active
finding:spd-splits-w-out-into-50-subcomponents-forming-a-single-rank-50-component-in-compressed-computation-modelSPD splits W_out into 50 subcomponents forming a single rank-50 component in compressed computation model
Contradicts APD's 100 components for W_out.
Source paper
extracted_from(2025) · Lucius Bushnaq · Dan Braun · Lee Sharkey
Neighborhood — ranked by edge-count
Papers (1)
paper
- Stochastic Parameter Decompositioncontradictsmentions
Claims (1)
claim
- Author interpretation.
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.
- SPD decomposes W_in of compressed computation model into 100 subcomponents, one per input featurefinding0.920Neuron contributions match target.
- SPD finds 100 W_in components and a single rank-50 W_out component spanning both layers in the 2-layer modelfinding0.882Multi-layer mechanisms.
- SPD finds 102 W_in components and a rank-51 W_out component spanning all three layers in the 3-layer modelfinding0.858APD struggled beyond two layers.
- Neuron contributions of SPD subcomponents closely match target model across all 100 featuresfinding0.800Figure 7.
- Figure 1.
- Ground truth mechanisms identified.
- Assertion about the qualitative advantages of VPD's rank-one decomposition.
- Source of robustness.