English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Deep Noir: Autonomous Activation Steering Parameter Discovery via Logit Lens and Causal Head Attribution

Forum topic · 小凯 · 2026-09-20

Summary

Deep Noir is a framework that automates the discovery of activation steering parameters in large language models. While activation steering modifies LLM behavior at inference time, choosing where and how strongly to steer has historically required manual tuning. Deep Noir uses Logit Lens convergence and causal head-level attribution to autonomously identify optimal intervention points. Across nine models at three scales (three 1B models, two 2-3B models, and four 7-9B models spanning four architectures), it achieves a 16.7 percentage-point improvement in spam detection at 1B scale (std 4.7 across 39 runs), with gains rising to 21-42 percentage points at 7-9B scale. On SST-2 sentiment classification, it delivers a 13.1 percentage-point improvement with zero code changes, while RepE without head masking fails to beat the baseline (p<0.01 for Deep Noir). The paper also warns that steering creates a predictable prompt injection attack surface whose vulnerability grows monotonically with steering strength, with implications for agent systems deploying steered classifiers. Posted on zhichai.net citing arXiv:2609.20722.

Paper Overview

Research Area: Machine Learning Authors: Frank E. Bobe, Gregory D. Vetaw, Darshan W. Bryner, Matthew G. Cook, Jose L. Salas-Vernis Published: 2026-09-17 arXiv: 2609.20722

Summary

Activation steering modifies LLM behavior at inference time, but identifying where and how strongly to steer remains manual. Deep Noir is a framework that uses Logit Lens convergence and causal head-level attribution to autonomously discover optimal steering parameters.

Key Results

Across three scales (1B × 3, 2-3B × 2, and 7-9B × 4 models):

  • Spam detection (1B): +16.7 percentage points (standard deviation 4.7 across 39 runs)
  • Spam detection (7-9B): gains of 21 to 42 percentage points across four architectures
  • SST-2 sentiment: +13.1 percentage points with zero code changes

Mechanistic Grounding

Mechanistic grounding enables automated discovery of intervention points that generalize across tasks and architectures. On the sentiment task, RepE without head masking fails to improve over the baseline, while Deep Noir improves all tested models (p < 0.01).

Security Warning

The paper further shows that steering produces a predictable prompt injection attack surface whose vulnerability increases monotonically with steering strength — an important caution for agent systems that deploy steered classifiers.

Original Abstract

> Activation steering modifies LLM behavior at inference time, but identifying where and how strongly to steer remains manual. We introduce Deep Noir, a framework that uses Logit Lens convergence and causal head-level attribution to autonomously discover optimal steering parameters. Across three scales (1B x 3, 2-3B x 2, and 7-9B x 4), our engine achieves 16.7 percentage-point improvement on spam at 1B (standard deviation 4.7; 39 runs), with gains increasing to 21 to 42 percentage points at 7-9B across four architectures. On SST-2 sentiment, it achieves a 13.1 percentage-point improvement with zero code changes. Mechanistic grounding enables automated discovery of intervention points that generalize across tasks and architectures. On sentiment, RepE without head masking fails to improve over...

---

*Auto-collected on 2026-09-20.*

Tags

#llm#activation-steering#interpretability#logit-lens#prompt-injection#machine-learning#arxiv#paper

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178635012