Refusal Mechanisms in Language Models

Latest papers

39 papers2 papers in the last four weeks, down 78% on the four weeks before. 0.0% of all new papers.

How Alignment Routes: Localizing, Scaling, and Controlling Policy Circuits in Language Models
Apr 6, 2026Gregory N. FrankLLM Interpretability LLM Safety
Chief Scientist, Molt AI Corp.
Blind Refusal: Language Models Refuse to Help Users Evade Unjust, Absurd, and Illegitimate Rules
Apr 3, 2026Cameron Pattison, Lorenzo Manuali, Seth LazarRefusal Behavior in Language Models Refusal Mechanisms in Language Models
Department of Philosophy, Vanderbilt University · Department of Philosophy, University of Michigan · School of Government and Policy, Johns Hopkins University
Do Language Models Know When They'll Refuse? Probing Introspective Awareness of Safety Boundaries
Mar 31, 2026Tanay GondilLanguage Model Introspection Refusal Behavior in Language Models
Purdue University
Over-Refusal and Representation Subspaces: A Mechanistic Analysis of Task-Conditioned Refusal in Aligned LLMs
Mar 29, 2026Utsav Maskey, Mark Dras, Usman NaseemLLM Interpretability LLM Refusal Behavior
Macquarie University
Hallucination as output-boundary misclassification: a composite abstention architecture for language models
Mar 12, 2026Angelina HintsanenLLM Hallucination Mitigation Hallucination in Language Models
NEXUS Laboratory, Helsinki, Finland
From Refusal Tokens to Refusal Control: Discovering and Steering Category-Specific Refusal Directions
Mar 9, 2026Rishab Alagharu, Ishneet Sukhvinder Singh, Shaibi Shamsudeen +2Language Model Steering LLM Refusal Behavior
Algoverse AI Research · Carnegie Mellon University · University of Maryland
Knowing without Acting: The Disentangled Geometry of Safety Mechanisms in Large Language Models
Mar 6, 2026Jinman Wu, Yi Xie, Shen Lin +2Refusal Mechanisms in Language Models LLM Safety Alignment
Xidian University, Xi’an, China · Tsinghua University, Shenzhen, China · Fujian Normal University, Fuzhou, China +1
There Is More to Refusal in Large Language Models than a Single Direction
Feb 2, 2026Faaiz Joad, Majd Hawasly, Sabri Boughorbel +2LLM Interpretability LLM Refusal Behavior
Qatar Computing Research Institute, HBKU, Doha, Qatar
Honesty over Accuracy: Trustworthy Language Models through Reinforced Hesitation
Nov 14, 2025Mohamad Amin Mohamadi, Tianhao Wang, Zhiyuan LiLLM Abstention LLM Hallucination Mitigation
Toyota Technological Institute at Chicago · University of California, San Diego

Apr 6, 2026J/K move · Enter open · S save

How Alignment Routes: Localizing, Scaling, and Controlling Policy Circuits in Language Models

Gregory N. Frank

Chief Scientist, Molt AI Corp.

LLM Interpretability LLM Safety LLM Jailbreak Attacks Circuits in Language Models Causal Interventions in Language Models Refusal Behavior in Language Models+1

Open paper AskPDF arXiv

We localize the policy routing mechanism in alignment-trained language models. An intermediate-layer attention gate reads detected content and triggers deeper amplifier heads that boost the signal toward refusal. In smaller models the gate and amplifier are single heads; at larger scale they become bands of heads across adjacent layers. The gate contributes under 1% of output DLA, yet interchange testing (p < 0.001) and knockout cascade confirm it is causally necessary. Interchange screening at n >= 120 detects the same motif in twelve models from six labs (2B to 72B), though specific heads differ by lab. Per-head ablation weakens up to 58x at 72B and misses gates that interchange identifies; at scale, interchange is the only reliable audit. Modulating the detection-layer signal continuously controls policy from hard refusal through evasion to factual answering. On safety prompts the same intervention turns refusal into harmful guidance, showing that the safety-trained capability is gated by routing, not removed. Thresholds vary by topic and by input language, and the circuit relocates across generations within a family even while behavioral benchmarks register no change. Routing is early-commitment: the gate fires at its own layer before deeper layers finish processing the input. An in-context substitution cipher collapses gate interchange necessity by 70 to 99% across three models, and the model switches to puzzle-solving rather than refusal. Injecting the plaintext gate activation into the cipher forward pass restores 48% of refusals in Phi-4-mini, localizing the bypass to the routing interface. A second method, cipher contrast analysis, uses plain/cipher DLA differences to map the full cipher-sensitive routing circuit in O(3n) forward passes. Any encoding that defeats detection-layer pattern matching bypasses the policy regardless of whether deeper layers reconstruct the content.