claim
active
claim:sdl-has-substantial-practical-and-conceptual-limitations-despite-being-the-leading-decomposition-methodSDL has substantial practical and conceptual limitations despite being the leading decomposition method
Core assessment of SDL.
Source paper
extracted_from(2025) · Lee Sharkey · Bilal I. Chughtai · Joshua Batson · Jack W. Lindsey +25
Neighborhood — ranked by edge-count
Papers (1)
paper
- Open Problems in Mechanistic Interpretabilitymentionssupports
Findings (4)
finding
- Example where SAE yields useful latent.
- Shows SDL reconstruction error is too high.
- Quantifies SAE reconstruction damage.
- SAE latents depend on training data.
Questions (1)
question
- Open question A.1.1a.4.
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.
- SDL reconstruction error might be reduced by improved SDL training or new decomposition methodshypothesis0.812Structured errors imply potential solutions.
- Calls for conceptual clarity.
- Author interpretation.
- Claimed by Braun et al.; demonstrated empirically here.
- Alexander's claim that the limiting factor in creating living structure is not method but the maker's persistence.
- Asserts that only piecemeal injection of morphogenetic sequences can realistically transform the system.
- Applied contribution.
- We hypothesize that degraded generalization on benchmarks like MMLU may reflect the computational demands of the tasks.hypothesis0.728Connecting the paper's task-difficulty findings to prior observations of weak generalization on complex QA benchmarks.