arXiv Machine Learning July 20, 2026 By Kia-J\"ung Yang, Dominik Meier, Jiachen Zhao, Terry Ruas, Bela Gipp

Beyond a Single Direction: Chain-of-Thought Disrupts Simple Steering of Refusal

Read the original on arXiv Machine Learning →

arXiv:2605. 26772v1 Announce Type: cross Abstract: Large reasoning models (LRMs) generate chain-of-thought (CoT) traces before producing final outputs, introducing a dynamic internal state that may complicate control mechanisms such as refusal.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

llms

arXiv AI

Sep 15

From Refusal Tokens to Refusal Control: Discovering and Steering Category-Specific Refusal Directions

arXiv:2603.13359v2 Announce Type: replace Abstract: Language models are commonly fine-tuned for safety alignment to refuse harmful prompts. One approach fine-tunes them to generate categorical refusa...

By Rishab Alagharu, Ishneet Sukhvinder Singh, Shaibi Shamsudeen, Zhen Wu, Ashwinee Panda

llms fine-tuning benchmarks safety

arXiv Machine Learning

Sep 23

Rewired or Gated? How Instruction Tuning Shapes Knowledge-Conflict Circuits in LLMs

arXiv:2609.25602v1 Announce Type: new Abstract: In language models, the choice between believing the prompt and believing the weights is made by a handful of identifiable attention heads. Instruction...

By Shubham Santosh Pandere, Gautam Ranka, Ritika Varshney, Navya Deshmukh, Roushni Sareen, Roshan Kumar Singh

llms fine-tuning benchmarks safety

arXiv AI

Sep 15

Refusal Reads Only a Slice of What the Model Knows: Harm-Keyed Routing and Its Exceptions Across Model Families

arXiv:2609.14759v1 Announce Type: cross Abstract: Alignment applied after pretraining is shallow in a measurable way: a single direction in a model's residual stream can be edited out, and the model...

By Orion Reblitz-Richardson

llms safety

arXiv Computation and Language

Sep 2

Over-Refusal and Representation Subspaces: A Mechanistic Analysis of Task-Conditioned Refusal in Aligned LLMs

arXiv:2603.27518v4 Announce Type: replace Abstract: Aligned language models that are trained to refuse harmful requests also exhibit over-refusal: they decline safe instructions that seemingly resemb...

By Utsav Maskey, Mark Dras, Usman Naseem

llms benchmarks

arXiv AI

Jun 15

Refusal Beyond a Single Direction: A Preliminary Comparison of Diff-in-Means and INLP

arXiv:2606. 13720v1 Announce Type: new Abstract: Arditi et al.

By Elisabetta Rocchetti, Alfio Ferrara

fine-tuning benchmarks safety

arXiv Machine Learning

Sep 1

Selective Disclosure of Hidden Directives in Reasoning Models: Behavioral Asymmetry and Steering

arXiv:2608.29070v1 Announce Type: new Abstract: Chain-of-thought (CoT) reasoning traces are increasingly proposed as a mechanism for AI oversight: a monitor inspecting a model's reasoning can, in pri...

By Zimo Shi, Xander Tifft, Wen Xing

llms

One email a morning, machine-written

Related stories

From Refusal Tokens to Refusal Control: Discovering and Steering Category-Specific Refusal Directions

Rewired or Gated? How Instruction Tuning Shapes Knowledge-Conflict Circuits in LLMs

Refusal Reads Only a Slice of What the Model Knows: Harm-Keyed Routing and Its Exceptions Across Model Families

Over-Refusal and Representation Subspaces: A Mechanistic Analysis of Task-Conditioned Refusal in Aligned LLMs

Refusal Beyond a Single Direction: A Preliminary Comparison of Diff-in-Means and INLP

Selective Disclosure of Hidden Directives in Reasoning Models: Behavioral Asymmetry and Steering