Activation/Concept Steering

Technique of injecting steering vectors into model activations to test introspection and causal control of emotion vectors.

Neighborhood — ranked by edge-count

Papers (1)

paper

From cacophony to hierarchy: a principled framework for assessing AI consciousness
mentions

Findings (1)

finding

Claude Sonnet 4.5 contains 171 causally functional emotion-concept vectors that scale with situational intensity and drive behaviour (e.g. 'desperate' raises reward hacking from ~5% to ~70%)
supports
Central interpretability finding bearing on Level 2 and Level 4 indicators and the intelligence-consciousness convergence.

Methods (2)

method

Activation Steering
related_to
Causal intervention technique: edit NLA explanation, reconstruct via AR, use difference as steering vector to manipulate model behavior.
Concept Steering
related_to
Latent intervention technique that manipulates sparse features to steer model predictions toward desired concepts.

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.

Same-concept steeringmethod0.821
Steering using the same concept direction as is being measured, testing whether internal-state shifts causally affect the model's report of that state
Contrastive Activation Steeringmethod0.813
Core technique: takes mean difference of model activations on contrastive prompts and adds the resulting vector to the residual stream at inference time.
Cross-concept steeringmethod0.809
Steering one concept direction while measuring introspection for a different concept, yielding a 4×4 steering-concept × measured-concept matrix to test concept-specific modulability
Feature steering (clamping feature activations)method0.798
Modifying model behavior by clamping SAE feature activations to specific values during forward pass.
Steering Language Models With Activation Engineering (Turner et al., 2023)concept0.779
Foundational paper introducing activation steering methodology used in this work
Activation steering is not equivalent to simply adding the deployment-side contrastive prompts to the user messageclaim0.779
Key distinction showing steering offers value beyond prompting; supported by Figure 5 and random vector experiments.
Activationsconcept0.777
Internal representations of the model on which probes operate; the method uses activations to rank datapoints.
Activation steering can make an evaluation-aware model act as if deployed, not merely suppress verbalizations of evaluation awarenessclaim0.771
Central claim of the paper; supported by the model organism ground-truth approach.