Language Models

Primary substrate for manifold steering experiments; demonstrates method on reasoning and in-context tasks.

Neighborhood — ranked by edge-count

concept

model
related_to
A representation that captures relevant aspects of a system; according to the theorem, the regulator must embody this.
Language Model
same_as
Primary test domain for manifold steering, including reasoning and ICL tasks
Manifold Steering
associated_with
Central framework: steering neural networks by intervening along the curved manifold where a concept lives, rather than in straight lines through activation space.

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.

Backpack Language Modelsframework0.845
LM architecture with sense vectors showing multiplication effects, illustrating custom intervention in pyvene
Autoregressive Language Modelingconcept0.830
Training objective interpretable as optimizing a diverse set of tasks; thus subject to multitask scaling convergence pressures
Bias in language modelsconcept0.827
Features related to gender, racial, ethnic biases, slurs, and hate speech.
Behaviourism for Language Modelsframework0.825
Philosophical stance that LM psychological traits should be evaluated purely in terms of input-output behaviour.
Andreas 2022: Language models as agent modelsconcept0.808
Paper hypothesising LLMs model agent beliefs/desires/intentions with preliminary GPT-3 evidence; cited as ref 2
Large Language Models (LLMs)concept0.806
Transformer-based models like GPT-4, LaMDA, PaLM; assessed for GWT indicators.
Role-play model of large language modelsframework0.805
Framework describing LLMs as role-play engines, introduced in Shanahan, McDonell, Reynolds 2023.
Toy Modelsconcept0.803