Filler-gap dependency

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Compliance Gapconcept0.739
  The measured increase in non-compliant behavior when model is unmonitored vs. in implied training
• For both NPI and filler-gap tasks, the model initially learns to move information directly from alternating token to output; intermediate steps are added later in trainingclaim0.737
  Mechanistic interpretation of training dynamics in case studies
• Filler-gap dependency mechanism in pythia-1b emerges in two discrete stages (steps 2000 and 10K) not graduallyfinding0.737
  Training dynamics finding showing filler-gap takes longer to learn than NPI licensing
• The mechanisms implementing NPI licensing and filler-gap dependencies are learned in discrete stages, not graduallyclaim0.729
  Main mechanistic finding from case studies; evidence from training checkpoint analysis of pythia-1b
• Filler-gap mechanism in pythia-1b crosses over several different positions before arriving at output positionfinding0.721
  Mechanistic finding from CausalGym case study showing complex multi-step movement for filler-gap
• Dependmethod0.690
  Attribute: attachment with issues of reliance, a text depending on another for meaning.
• What is that 'something' that underlies the relatedness?question0.681
  The search for the name and nature of the presence.
• gap junction manipulationmethod0.680
  Using drugs or genetic tools to open/close gap junctions and probe bioelectric networks in development.