framework
active
framework:distill-circuits-thread

Distill Circuits Thread

Prior mechanistic interpretability work reverse-engineering vision models (InceptionV1); the direct predecessor this paper extends to language models

Neighborhood — ranked by edge-count

Frameworks (2)

framework
  • Circuits Thread
    related_to
    An open scientific collaboration hosted on Distill slack studying the inner workings of neural networks via zoomed-in mechanistic investigation
  • Prior Anthropic paper enabling circuit-level analysis of attention-only transformers; motivates current MLP decomposition

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.

  • Circuits Frameworkframework0.752
    Mechanistic interpretability framework for understanding neural network computation as circuits of features
  • Default circuitsconcept0.736
    Circuits active by default, e.g. can't-answer, suppressed by contrary evidence.
  • Circuit Motifconcept0.725
    A recurring, abstract pattern found in circuits (e.g., equivariance, unioning over cases), inspired by circuit motifs in systems biology
  • Circuit Findingmethod0.721
    Interpretability technique for identifying functional sub-circuits in neural networks, supported by pyvene
  • Stage 2 of character training: DPO from teacher model to student model to transfer desired behavioral expressions
  • Sparse circuitsconcept0.717
    A goal in mechanistic interpretability to identify sparse computational subgraphs; VPD promotes sparse parameter circuits.
  • Circuit tracingmethod0.715
    Anthropic technique revealing computational pathways in Claude.
  • The key novel property of DiffLogic CA — logic gate networks that are recurrent both spatially and temporally