BWhat changed
Constructed steering directions by averaging activation differences from contrasting response pairs and applied them during inference to shift selected behaviors of Llama 2 models without updating the base model's weights.
Scientific scope
Constructed steering directions by averaging activation differences from contrasting response pairs and applied them during inference to shift selected behaviors of Llama 2 models without updating the base model's weights. This uses contrastive examples, activation access and selected intervention parameters. Behavioral changes and side effects depend on the model, task and steering strength; pair-averaged construction is distinct from a single-prompt activation difference.
Evidence
The linked original works support the scoped statement and contribution attribution. WRS has not independently replayed experiments or reconstructed a full mathematical proof. This uses contrastive examples, activation access and selected intervention parameters. Behavioral changes and side effects depend on the model, task and steering strength; pair-averaged construction is distinct from a single-prompt activation difference.
Contributors
- The named authors in the research record · SCIENCE: Source-listed authors for the exact contribution; experimental subjects are not autonomous scientific authors. Source support and predecessor work retain separate roles. Individual effort shares and unresolved identities are not inferred.
Provisional WRS assessment, edition 2026-10-23.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction