Refusal in Language Models Is Mediated by a Single Direction
The DOI matches its Crossref record; 18 of the 18 references it cites were found in a registry or an index.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“ref. 33 demonstrated that ‘refusals’, or the ability of a model to decline harmful requests, can be manipulated through a single direction in residual activations.”
From Training large language models on narrow tasks can lead to broad misalignment (2026) · cites this paper as [33]
In the citing paper
“Importantly, this suggests that harmlessness training (which trains models to refuse to answer harmful requests) may be cheaply undone.”
From Open Problems in Mechanistic Interpretability (2025) · cites this paper as Arditi et al., 2024
In the citing paper
“By understanding the mechanisms of jailbreaking, we can anticipate the means through which a user might bypass existing safeguards (Lee et al., 2025;Arditi et al., 2024).”
From Open Problems in Mechanistic Interpretability (2025) · cites this paper as Arditi et al., 2024
Citations
Cites
The publisher hasn’t deposited this paper’s reference list.
Cited by
Items 301–400 of 1,000
Paths Not Taken: Understanding and Mending the Multilingual Factual Recall PipelinePreprint
Meng, L., Zhang, R., Eickhoff, C., & Pavlick, E. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2505.20546 · Cited by 8
A Granular Study of Safety Pretraining under Model Abliteration
Agnihotri, S., Jakubassa, J., Dey, P., Goyal, S., Schiele, B., Radhakrishnan, V. B., & Keuper, M. (2025) · arXiv
Physics Steering: Causal Control of Cross-Domain Concepts in a Physics Foundation ModelPreprint
Fear, R. A., Mukhopadhyay, P., McCabe, M., Bietti, A., & Cranmer, M. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2511.20798 · Cited by 8
Unified Defense for Large Language Models against Jailbreak and Fine-Tuning Attacks in EducationPreprint
Yi, X., Li, Y., Shi, D., Wang, L., Wang, X., & He, L. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2511.14423 · Cited by 8
From Narrow Unlearning to Emergent Misalignment: Causes, Consequences, and Containment in LLMsPreprint
Mushtaq, E., Ramakrishna, A., Krishna, S., Sahai, S., Goyal, P., Chang, K.-W., Zhang, T., & Gupta, R. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2511.14017 · Cited by 8
Ni, Y., & Yang, T. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2512.18776 · Cited by 8
Probe-Based Data Attribution: Discovering and Mitigating Undesirable Behaviors in LLM Post-TrainingPreprint
Xiao, F., & Aranguri, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.11079 · Cited by 8
Wang, K., Li, Z., Zhou, Z., Zhang, Y., Mi, Y., Yang, K., Zhang, Y., Dong, J., Sun, Z., Li, Q., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.10161 · Cited by 8
The Geometry of Alignment Collapse: When Fine-Tuning Breaks SafetyPreprint
Springer, M., Lee, C. P., Metevier, B., Castleman, J., Turbal, B., Jung, H., Shen, Z., & Korolova, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.15799 · Cited by 8
Minimizing Collateral Damage in Activation SteeringPreprint
Nguyen, T., Nguyen, T. A., Alemohammad, S., & Baraniuk, R. G. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.01167 · Cited by 8
ResearchArena: Evaluating Sabotage and Monitoring in Automated AI R&DPreprint
Libon, L., Rank, B., Yeon, J., Schmotz, D., Qin, J., Donnelly, D., Prinzhorn, D., & Andriushchenko, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.19321 · Cited by 8
Synthetic Persona Pretraining: Alignment from Token ZeroPreprint
Minder, J., Moskvoretskii, V., Singhal, R., Jiao, D., Arditi, A., Cui, S., Borjigin, Y., Bali, K., Krsteski, S., Raj, H., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.13482 · Cited by 8
Self-citationPersona Vectors: Monitoring and Controlling Character Traits in Language Models
Chen, R., Arditi, A., Sleight, H., Evans, O., & Lindsey, J. (2025) · arXiv · doi:10.48550/arxiv.2507.21509 · Cited by 7
Self-citationCircuit Stability Characterizes Language Model Generalization
Sun, A. (2025) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2025.acl-long.442 · Cited by 7
Weng, J., Zheng, H., Zhang, H., Zhou, E., He, Q., Tao, J., Xue, H., Chu, Z., & Wang, X. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.944 · Cited by 7
SafeScientist: Enhancing AI Scientist Safety for Risk-Aware Scientific Discovery
Zhu, K., Zhang, J., Qi, Z., Shang, N., Liu, Z., Han, P., Su, Y., Yu, H., & You, J. (2025) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2025.emnlp-main.116 · Cited by 8
Guarding the Guardrails: A Taxonomy-Driven Approach to Jailbreak Detection
Giarrusso, F., Sorokoletova, O., Suriani, V., & Nardi, D. (2026) · Proceedings of IASEAI Conference · doi:10.1609/iaseai.v2i1.43024 · Cited by 7
Efficient and accurate steering of large language models through attention-guided feature learning
Davarmanesh, P., Wilson, A., & Radhakrishnan, A. (2026) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2609978123 · Cited by 7
Adaptive Helpfulness–Harmlessness Alignment with Preference Vectors
Liang, R.-W., Hsu, C. T., Yu, C.-H., Agrawal, S., Huang, S.-C., Lin, C.-Y., Chen, S.-T., Huang, K.-H., & Sun, S.-H. (2026) · Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2026.eacl-long.77 · Cited by 7
Shifting Perspectives: Steering Vectors for Robust Bias Mitigation in LLMs
Siddique, Z., Khalid, I., Turner, L. D., & Espinosa-Anke, L. (2026) · Findings of the Association for Computational Linguistics: EACL · doi:10.18653/v1/2026.findings-eacl.41 · Cited by 7
Feature Drift: How Fine-Tuning Repurposes Representations in LLMs
Galichin, A. V., Korznikov, A., Dontsov, A., Rogov, O. Y., Tutubalina, E., & Oseledets, I. (2026) · Findings of the Association for Computational Linguistics: EACL · doi:10.18653/v1/2026.findings-eacl.96 · Cited by 7
What Drives Representation Steering? A Mechanistic Case Study on Steering Refusal
Cheng, S. S., Wiegreffe, S., & Manocha, D. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.79 · Cited by 7
Mueller, A., Lee, A., Joshi, S., Lubana, E. S., Sridhar, D., & Reizinger, P. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.782 · Cited by 7
When Correct Beliefs Collapse: Epistemic Resilience of LLMs under Clinical Pressure
Xiao, B., Tian, X., Song, X., Wang, H., Song, G., Zhao, S., & Qin, B. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.395 · Cited by 8
Between a Rock and a Hard Place: The Tension Between Ethical Reasoning and Safety Alignment in LLMs
Chua, S. P., Thai, Z. L., Teh, K. J., Li, X., Ren, Q., & Hu, X. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.197 · Cited by 7
Consiglieres in the Shadow: Understanding the Use of Uncensored Large Language Models in CybercrimesPreprint
Lin, Z., Li, Z., Liao, X., & Wang, X. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2508.12622 · Cited by 7
Noise Injection Systemically Degrades Large Language Model Safety GuardrailsPreprint
Shahani, P. S., Miandoab, K. E., & Scheutz, M. J. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2505.13500 · Cited by 7
The Safety Gap Toolkit: Evaluating Hidden Dangers of Open-Source ModelsPreprint
Dombrowski, A.-K., Bowen, D., Gleave, A., & Cundy, C. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2507.11544 · Cited by 9
Enhancing Cross-task Transfer of Large Language Models via Activation SteeringPreprint
Tang, X., Lv, Z., Cheng, X., Li, J., Zhao, W. X., Wen, Z., Zhang, Z., & Zhou, J. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2507.13236 · Cited by 7
Hallucination reduction with CASAL: Contrastive Activation Steering For Amortized Learning
Wannan, Yang, Qiu, X., Lei, Y., Zhang, Y., Yang, A., Kokhlikyan, N., Cancedda, N., & García-Olano, D. (2026) · International Conference on Learning Representations · doi:10.48550/arxiv.2510.02324 · Cited by 8
On the Effectiveness and Generalization of Race Representations for Debiasing High-Stakes DecisionsPreprint
Nguyen, D., & Tan, C. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2504.06303 · Cited by 7
Dynamic Token Reweighting for Robust Vision-Language ModelsPreprint
Jiang, T., Liang, J., Zhu, R., Zhou, J., Ma, F., & Wang, T. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2505.17132 · Cited by 7
Multilingual Political Views of Large Language Models: Identification and SteeringPreprint
Gurgurov, D., Trinley, K., Vykopal, I., van Genabith, J., Ostermann, S., & Zamparelli, R. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2507.22623 · Cited by 7
AbsTopK: Rethinking Sparse Autoencoders For Bidirectional Features
Zhu, X., Khalili, M. M., & Zhu, Z. (2025) · International Conference on Learning Representations · doi:10.48550/arxiv.2510.00404 · Cited by 7
Knowing without Acting: The Disentangled Geometry of Safety Mechanisms in Large Language ModelsPreprint
Wu, J., Xie, Y., Lin, S., Zhao, S., & Chen, X. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.05773 · Cited by 7
Steering at the Source: Style Modulation Heads for Robust Persona ControlPreprint
Izawa, Y., Minegishi, G., Eguchi, K., Hosokawa, S., & Taura, K. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.13249 · Cited by 7
The Unreasonable Effectiveness of Text Embedding Interpolation for Continuous Image SteeringPreprint
Ekin, Y., & Gandelsman, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.17998 · Cited by 8
State-Dependent Safety Failures in Multi-Turn Language Model InteractionPreprint
Li, P., (64655), J. Z., Zhang, T., Qiu, H., kejun, Z., Zhang, W., Yu, N., & Zhou, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.15684 · Cited by 7
Steerable but Not Decodable: Function Vectors Operate Beyond the Logit LensPreprint
Nadaf, M. S. B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.02608 · Cited by 7
Steered LLM Activations are Non-SurjectivePreprint
Mishra, A., Khashabi, D., & Liu, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.09839 · Cited by 8
Predicting Where Steering Vectors SucceedPreprint
Billa, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.15557 · Cited by 7
Semantic Denial of Service in LLM-controlled robotsPreprint
Steinberg, J., & Gal, O. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.24790 · Cited by 7
TACT: Mitigating Overthinking and Overacting in Coding Agents via Activation SteeringPreprint
Sui, Y., Chen, Y., Li, Y., Jiang, X., He, Y., Dong, Y., He, X., Tianyu, G., & Hooi, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.05980 · Cited by 7
Probing Persona-Dependent Preferences in Language ModelsPreprint
Gilg, O., Beckmann, P., Paleka, D., & Butlin, P. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.13339 · Cited by 7
Self-citationContextEcho: A Benchmark for Persona Drift in Long Agentic-Coding SessionsPreprint
Ding, X., Yu, Y., Liu, C., Zhao, B., Chen, L., & Chen, T. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.24279 · Cited by 7
Learning to Attack and Defend: Adaptive Red Teaming of Language Models via GRPOPreprint
Bullwinkel, B., Kim, E., Minnich, A., & Russinovich, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.09701 · Cited by 7
Position: Anthropomorphic Misalignment Research Needs Stronger EvidencePreprint
Gupta, V., Nutter, P., Stante, S., Krause, A., Tramèr, F., Fluri, L., Chen, X., & Hedström, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.07612 · Cited by 7
RepBench: Compiling Benchmarks into Capability Representations for Large Language ModelsPreprint
Li, Y., Bai, X., Liu, S., & Zhang, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.28008 · Cited by 7
Inducing language models to assert their own consciousness restores human beliefs and valuesPreprint
Kim, J., Street, W., Rocca, R., Korngiebel, D. M., Waytz, A., Evans, J. A., & Keeling, G. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.28607 · Cited by 8
Automated Researchers Can Mitigate Well-characterized Alignment FailuresPreprint
Yueh-Han, C., Wen, J., & Kirchner, J. H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.28945 · Cited by 7
Dialz: A Python Toolkit for Steering Vectors
Siddique, Z., Turner, L. D., & Espinosa-Anke, L. (2025) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2025.acl-demo.35 · Cited by 6
Representation-based Broad Hallucination Detectors Fail to Generalize Out of Distribution
Dubanowska, Z., Żelaszczyk, M., Brzozowski, M., Mandica, P., & Karpowicz, M. (2025) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2025.findings-emnlp.952 · Cited by 6
“I’ve Decided to Leak”: Probing Internals Behind Prompt Leakage Intents
Dong, J., Zhang, Y., Yan, L., Zhong, Z., Wei, T., Xu, K., Huang, M., Zhang, C., & Qiu, H. (2025) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2025.emnlp-main.1082 · Cited by 6
Liu, G., Li, F., Bono, C. A., Yang, S., Zhu, M., & Pierri, F. (2026) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/3819074 · Cited by 6
Why does Weak-OOD Help? A Further Step Towards Understanding Jailbreaking VLMs
Zhou, Y., Peng, Y., Bai, Y., Gao, K., Zhang, Y., Zhang, Y., Chen, X., Tao, Y., Dai, T., & Xia, S. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.1980 · Cited by 6
PII-Scope: A Comprehensive Study on Training Data Privacy Leakage in Pretrained LLMs
Nakka, K. K., Frikha, A., Mendes, R., Jiang, X., & Zhou, X. (2025) · International Joint Conference on Natural Language Processing and Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (IJCNLP-AACL) · doi:10.18653/v1/2025.ijcnlp-long.195 · Cited by 6
Rethinking Deep Alignment Through the Lens of Incomplete Safety Learning
Bach, T., Nguyen, D., Le, T. M., & Tran, T. (2026) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v40i36.40248 · Cited by 6
SafeSteer: Localized On-Policy Distillation for Efficient Safety Alignment
Li, H., An, J., Song, Z., Zhu, P., Li, R., Wang, H., Feng, W., Liu, Y., Li, L., Yao, J.-G., et al. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.555 · Cited by 6
TamperBench : Systematically Stress-Testing LLM Safety Under Fine-Tuning and Tampering
Hossain, S., Tseng, T., Pandey, P. S., Vajpayee, S., Kowal, M., Nonta, N., Simko, S., Casper, S., Jin, Z., Pelrine, K., et al. (2026) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3770855.3817557 · Cited by 9
Compiling Activation Steering into Weights via Null-Space Constraints for Stealthy Backdoors
Yin, R., Han, T., Xu, N., Li, C., He, P., Zhou, C., Wang, J., Fu, Z., Du, T., Li, J., et al. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.1206 · Cited by 7
Selective Steering: Norm-Preserving Control Through Discriminative Layer Selection
Dang, Q.-A., & Ngo, C. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.529 · Cited by 6
Multilingual Refusal Alignment for Safer Large Language Models
Krasnodębska, A., Kusa, W., & Lipani, A. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.1537 · Cited by 6
HumorReject: Decoupling LLM Safety from Refusal Prefix via A Little HumorPreprint
Wu, Z., Gao, H., Luo, J., & Liu, Z. (2025) · Preprints.org · doi:10.20944/preprints202501.1736.v1 · Cited by 6
Where Do Reasoning Models Refuse?Preprint
Yamaguchi, K., Etheridge, B., & Arditi, A. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2507.03167 · Cited by 6
Self-citationSilent Tokens, Loud Effects: Padding in LLMsPreprint
Himelstein, R., LeVi, A., Belinkov, Y., & Mendelson, A. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2510.01238 · Cited by 6
Designing Synthetic Discussion Generation Systems: A Case Study for Online FacilitationPreprint
Tsirmpas, D., Androutsopoulos, I., & Pavlopoulos, J. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2503.16505 · Cited by 6
Logit-Gap Steering: A Forward-Pass Diagnostic for Alignment RobustnessPreprint
Li, T.-L., & Liu, H. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2506.24056 · Cited by 6
Probing the Robustness of Large Language Models Safety to Latent PerturbationsPreprint
Gu, T., Huang, K., Wang, Z., Wang, Y., Li, J., Yao, Y., Yao, Y., Yang, Y., Teng, Y., & Wang, Y. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2506.16078 · Cited by 6
Latent Concept Disentanglement in Transformer-based Language Models
Hong, G. Z., Vasudeva, B., Sharan, V., Rashtchian, C., Raghavan, P., & Panigrahy, R. (2025) · International Conference on Learning Representations · doi:10.48550/arxiv.2506.16975 · Cited by 6
The Geometry of Harmfulness in LLMs through Subconcept ProbingPreprint
Shah, M., Angeline, S., Kumar, A. R., Chheda, N., Zhu, K., Sharma, V., O’Brien, S., & Cai, W. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2507.21141 · Cited by 6
Mitigating Jailbreaks with Intent-Aware LLMsPreprint
Yeo, W. J., Satapathy, R., & Cambria, E. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2508.12072 · Cited by 6
Read the Scene, Not the Script: Outcome-Aware Safety for LLMsPreprint
Wu, R., Quan, Y., Shi, Z., Wang, Z., Li, Y., & Tang, R. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2510.04320 · Cited by 6
Selective Steering: Norm-Preserving Control Through Discriminative Layer SelectionPreprint
Dang, Q.-A., & Ngo, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2601.19375 · Cited by 7
Bao, Y., Zhang, X., Chen, J., Su, G., Cai, Y., Peng, H., Sun, B., Weng, H., Yan, L., & Yin, J. (2026) · International Conference on Learning Representations · doi:10.48550/arxiv.2602.05234 · Cited by 7
Exposing the Systematic Vulnerability of Open-Weight Models to Prefill AttacksPreprint
Struppek, L., Gleave, A., & Pelrine, K. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.14689 · Cited by 6
Braun, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.17881 · Cited by 6
PSF-Med: Measuring and Explaining Paraphrase Sensitivity in Medical Vision Language ModelsPreprint
Sadanandan, B., & Behzadan, V. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.21428 · Cited by 6
Tucano 2 Cool: Better Open Source LLMs for PortuguesePreprint
Corrêa, N. K., Sen, A., Fatimah, S., Falk, S., Landgraf, L., Kastner, J., & Flek, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.03543 · Cited by 8
Censored LLMs as a Natural Testbed for Secret Knowledge ElicitationPreprint
Casademunt, H., Cywiński, B., Tran, K., Jakkli, A., Marks, S., & Nanda, N. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.05494 · Cited by 6
Self-citationA Comprehensive Information-Decomposition Analysis of Large Vision-Language Models
Xiu, L., Luo, X., & Nakayama, H. (2026) · International Conference on Learning Representations · doi:10.48550/arxiv.2603.29676 · Cited by 6
Jailbreaking Vision-Language Models Through the Visual ModalityPreprint
Azulay, A., Dubiński, J., Li, Z., Mittal, A., & Gandelsman, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.00583 · Cited by 6
Duan, E. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.15980 · Cited by 6
Evaluation Awareness Is Not One Capability: Evidence from Open Language ModelsPreprint
Nayan, N., Kumar, A. S., Girmal, R., Anilkumar, S., Vaidyanathan, S., Palacio, D. A. N., Ghosh, R., & Srinivasan, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.23583 · Cited by 6
ORBIT: Training-Free Multi-Attribute Behavioral Steering via Orthogonal Subspace RotationPreprint
Ghasemi, N., Ziashahabi, A., Avestimehr, S., & May, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.22357 · Cited by 6
RecirculationPreprint
Mozer, M. C., Siddiqui, S. A., Sawyer, D., Sanyal, S., & Liu, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.17981 · Cited by 8
Heiding, F., Lermen, S., Kao, A., Schneier, B., & Vishwanath, A. (2024) · arXiv · doi:10.48550/arxiv.2412.00586 · Cited by 5
Beyond Syntax: How Do LLMs Understand Code?
North, M., Atapour–Abarghouei, A., & Bencomo, N. (2025) · IEEE/ACM International Conference on Software Engineering: New Ideas and Emerging Results (ICSE-NIER) · doi:10.1109/icse-nier66352.2025.00023 · Cited by 5
Architectural Backdoors for Within-Batch Data Stealing and Model Inference Manipulation
Küchler, N., Petrov, I., Grobler, C., & Shumailov, I. (2026) · IEEE Conference on Secure and Trustworthy Machine Learning (SaTML) · doi:10.1109/satml68715.2026.00069 · Cited by 5
DISCO: Disentangled Communication Steering for Large Language Models
Torop, M., Masoomi, A., Eskandar, M., & Dy, J. (2025) · neural information processing systems · doi:10.52202/085713-1938 · Cited by 5
Attention Misses Visual Risk: Risk-Adaptive Steering for Multimodal Safety Alignment
Park, J., Seo, M., Yeo, C., & Choi, J. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37314-4_3 · Cited by 5
Agentic Reinforcement Learning for Search Misaligns Instruction-Tuning
Yang, Y., Padarha, S., Ball, S., Lee, A., & Mahdi, A. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.2440 · Cited by 5
Lock on Target! Precision Unlearning via Directional Control
Wen, Y., Feng, R., Guo, F., Wang, Y., Le, R., Song, Y., Gao, S., & Shang, S. (2025) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2025.findings-emnlp.1021 · Cited by 5
Attention Eclipse: Manipulating Attention to Bypass LLM Safety-Alignment
Zaree, P., Mamun, M. A. A., Alam, Q. M., Dong, Y., Alouani, I., & Abu‐Ghazaleh, N. (2025) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2025.emnlp-main.842 · Cited by 5
Step-Wise Refusal Dynamics in Autoregressive and Diffusion Language Models
Rahimi, E., Hirshel, E., LeVi, A., Himelstein, R., Mendelson, A., & Baskin, C. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.617 · Cited by 5
Lasnier, T., Antoun, W., Kulumba, F., Sagot, B., & Seddah, D. (2026) · International Joint Conference on Natural Language Processing and Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (IJCNLP-AACL) · doi:10.18653/2026.aacl-short.45 · Cited by 5
Causal Activation Steering via Sparse Mediation
Doan, T., Le, U., & Nguyen, T. (2026) · Findings of the Association for Computational Linguistics: EACL · doi:10.18653/v1/2026.findings-eacl.57 · Cited by 5
SafeCOMM: A Study on Safety Degradation in Fine-Tuned Telecom Large Language Models
Djuhera, A., Kadhe, S., Ahmed, F., Zawad, S., Koch, F. L., Saad, W., & Boche, H. (2026) · IEEE Wireless Communications and Networking Conference · doi:10.1109/wcnc65185.2026.11555432 · Cited by 5
Do LLMs Know Tool Irrelevance? Demystifying Structural Alignment Bias in Tool Invocations
Liu, Y., Lin, X., Cao, P., Zhang, G., Fang, F., & Cao, Y. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.1473 · Cited by 5
Interpretable Safety Alignment via SAE-Constructed Low-Rank Subspace Adaptation
Wang, D., Ma, Q., Shang, Y., Lu, Z., Xu, Z., Ning, L., Wu, H., & He, Z. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.215 · Cited by 5
FineSteer: A Unified Framework for Fine-Grained Inference-Time Steering in Large Language Models
Weng, Z., Zhang, J., Cai, K., Li, Y., Wang, P., & Tian, Y. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.852 · Cited by 5
Metadata
- Authors
- Arditi, A., Obeso, O., Syed, A., Paleka, D., Panickssery, N., Gurnee, W., & Nanda, N.
- ORCID
- Gurnee, Wes
- Venue
- Advances in Neural Information Processing Systems 37
- Published
- 2024
- Pages
- 136037-136083
- ISSN
- 1049-5258 (ISSN-L)
- OpenAlex
- W4415800641
- Cited by
- 70
- References
- 0
- Citations by year
- 2026: 61 · 2025: 9
- FWCI
- 386.89
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- neural information processing systems (publishedVersion)
- Topics
- Social Sciences › Arts and Humanities › Language and Linguistics › Language, Discourse, Communication Strategies
- Social Sciences › Arts and Humanities › Language and Linguistics › Linguistics and Language Analysis
- Physical Sciences › Computer Science › Artificial Intelligence › Speech and dialogue systems
- Keywords
- refusal, language models, large language models
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.