claim
active
claim:transcoder-based-attribution-graphs-omit-attentional-computation-rendering-them-useless-for-many-promptsTranscoder-based attribution graphs omit attentional computation, rendering them useless for many prompts
Motivating problem.
Source paper
extracted_fromNeighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
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.
- Circuit simplification theory.
- Quantifies method limitations.
- Gradient-based technique using SAE features to estimate causal effects on completions; used to corroborate NLA findings.
- Downside of abstraction.
- Central claim of the paper: the method scales to state-of-the-art transformers.
- Sparse Autoencoders Find Highly Interpretable Features in Language Models (Cunningham et al., 2023)concept0.763Core methodology paper for SAE-based interpretable feature extraction
- Number of attention head paths between transcoder features grows exponentially with layer separationclaim0.762Why checkpointing is needed.
- Some gradient-based attribution methods give attributions independent of model and data-generating processfinding0.762Adebayo et al. sanity checks.