Addition module

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

• Modular additionconcept0.853
  The mathematically natural computation for cyclic concepts (e.g., addition mod 12 for months), which the paper shows Llama does NOT directly implement
• Activation Additionmethod0.823
  Intervention method that adds a learned direction vector to residual stream activations to steer model behavior
• Base-10 additionconcept0.788
  The generic addition mechanism that Llama-3.1-8B actually uses to compute sums before mapping back to cyclic concept space
• Activation Addition (ActAdd)framework0.761
  Steering method deriving vectors from contrastive prompt pairs and adding to first-token activations.
• Generic addition mechanismconcept0.759
  The core finding: Llama-3.1-8B reuses one addition mechanism across all cyclic tasks rather than learning task-specific modular arithmetic
• Specialised modulesconcept0.745
  Independent, parallel cognitive subsystems that share information via a global workspace.
• Contrastive Activation Additionframework0.732
  Method formalized by Rimsky et al. (2024) that this paper adopts for extracting persona vectors
• Reflection Enhancement via Activation Additionmethod0.731
  Adding steering vector in forward direction to push model activations toward stronger reflective behavior.