mlsanityData to 14 Oct 2026

Deep Noir: Autonomous Steering Discovery via Architectural Chronometry in Transformer Models

arXiv 2609.20722 ↗First posted 17 Sep 2026cs.AIcited by 0

On the map: Mechanistic interpretability → Activation steering methods

Abstract (arXiv)

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 baseline, while Deep Noir improves all models (p less than 0.01). We further show that steering creates a predictable prompt-injection attack surface whose vulnerability increases monotonically with steering magnitude. This finding is relevant to agent systems deploying steered classifiers.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Activation steering methods, one of 13 directions in Mechanistic interpretability. Builds on 7 of its 18 references that are in the map. First posted on arXiv 17 Sep 2026.

Builds on

7 of 18 references.

Closest ancestors · 1

Same direction, newest first

  1. 2024Semantics-Adaptive Activation Intervention for LLMs via Dynamic Steering Vectors16 Oct

Other references · 6

  1. 2023Inference-Time Intervention: Eliciting Truthful Answers from a Language Modelcited by 1,302
  2. 2024Refusal in Language Models Is Mediated by a Single Directioncited by 1,032
  3. 2025Open Problems in Mechanistic Interpretabilitycited by 179
  4. 2025Taxonomy, Opportunities, and Challenges of Representation Engineering for Large Language Modelscited by 34
  5. 2026Analysing the Safety Pitfalls of Steering Vectorscited by 10
  6. 2025Transformers Don't Need LayerNorm at Inference Time: Scaling LayerNorm Removal to GPT-2 XL and the Implications for Mechanistic Interpretabilitycited by 7

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Behavioral Steering in a 35B MoE Language Model via SAE-Decoded Probe Vectors: One Agency Axis, Not Five Traits
  2. 2026Steering Under Compression: Dose-Response, Capability Cost, and Failure Asymmetry in Quantized LLMs
  3. 2023Steering Language Models With Activation Engineering
  4. 2026CircuitSteer: Geometrically Aligned Multi-Layer Steering via Sparse Autoencoder Circuits
  5. 2026Adversarial Robustness of Activation Steering in Large Language Models
  6. 2025The Rogue Scalpel: Activation Steering Compromises LLM Safety
  7. 2026Disentangling Steering Vectors
  8. 2026Curveball Steering: The Right Direction To Steer Isn't Always Linear