# Apple — Dynamically Scaled Activation Steering

- Company: Apple (apple.com)
- Announced: 2026-09-18
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://machinelearning.apple.com/research/dynamically-scaled-activation-steering
- Record: https://forck.live/items/12046-dynamically-scaled-activation-steering
- Subject: Machine Learning Research

Apple researchers introduced Dynamically Scaled Activation Steering (DSAS), a method-agnostic framework that adaptively modulates the strength of steering interventions across layers and inputs to guide generative model behavior. The approach decouples when to steer from how to steer, intervening strongly only when undesired behavior is detected, and demonstrates improved trade-offs between toxicity mitigation and utility preservation compared to existing steering methods alone.

## Evidence

Verbatim from https://machinelearning.apple.com/research/dynamically-scaled-activation-steering:

> DSAS adaptively modulates the strength of existing steering transformations across layers and inputs, intervening strongly only when undesired behavior is detected. At generation time, DSAS computes context-dependent scaling factors that selectively adjust the strength of any steering method.

---

Record: https://forck.live/items/12046-dynamically-scaled-activation-steering
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md