Generic addition mechanism

The core finding: Llama-3.1-8B reuses one addition mechanism across all cyclic tasks rather than learning task-specific modular arithmetic

Neighborhood — ranked by edge-count

concept

Base-10 addition
implements
The generic addition mechanism that Llama-3.1-8B actually uses to compute sums before mapping back to cyclic concept space
Fourier features
implements
Features identified in Llama-3.1-8B that compute sums using periods respecting base-10 addition (2, 5, 10) rather than concept-specific periods
Task-agnostic mechanism
associated_with
A computational mechanism reused across multiple tasks without specialization to any one domain

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.

Llama-3.1-8B reuses a single generic addition mechanism across all cyclic tasks independently of concept-specific geometryfinding0.767
Key mechanistic finding showing task-agnostic reuse of arithmetic circuitry
Activation Additionmethod0.763
Intervention method that adds a learned direction vector to residual stream activations to steer model behavior
Delegation mechanismconcept0.757
Base-10 Addition Mechanism in Llamaconcept0.742
Activation Addition (ActAdd)framework0.731
Steering method deriving vectors from contrastive prompt pairs and adding to first-token activations.
Modular additionconcept0.729
The mathematically natural computation for cyclic concepts (e.g., addition mod 12 for months), which the paper shows Llama does NOT directly implement
What Is The Mechanism That Enables Crosslevel Integratedquestion0.729
Reflection Enhancement via Activation Additionmethod0.728
Adding steering vector in forward direction to push model activations toward stronger reflective behavior.