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 101–200 of 1,000
On Reasoning Strength Planning in Large Reasoning Models
Sheng, L., Zhang, A., Wu, Z., Zhao, W., Shen, C., yi, Z., Wang, X., & Chua, T. (2025) · neural information processing systems · doi:10.52202/085713-2115 · Cited by 27
There Is More to Refusal in Large Language Models than a Single Direction
Joad, F., Hawasly, M., Boughorbel, S., Durrani, N., & Sencar, H. T. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.2606 · Cited by 31
Steering Dialogue Dynamics for Robustness against Multi-turn Jailbreaking Attacks
Hu, H., Robey, A., & Liu, C. (2026) · Transactions on machine learning research · doi:10.48550/arxiv.2503.00187 · Cited by 29
Lin, J., Liu, M., Huang, X., Li, J., Hong, H., Yuan, X., Chen, Y., Huang, L., Xue, H., Duan, R., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2601.15588 · Cited by 27
Kharinaev, A., Moskvoretskii, V., Shvetsov, E., Studenikina, K. A., Bykov, M., & Burnaev, E. V. (2026) · IEEE Access · doi:10.1109/access.2026.3703899 · Cited by 27
Dabas, M., Chen, S., Fleming, C., Jin, M., & Jia, R. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2507.04250 · Cited by 26
Sanity Checks for Sparse Autoencoders: Do SAEs Beat Random Baselines?Preprint
Korznikov, A., Galichin, A. V., Dontsov, A., Rogov, O. Y., Oseledets, I., & Tutubalina, E. V. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.14111 · Cited by 26
Improved Generation of Adversarial Examples Against Safety-aligned LLMs
Li, Q., Guo, Y., Zuo, W., & Chen, H. (2024) · neural information processing systems · doi:10.52202/079017-3054 · Cited by 26
Safety Through Reasoning: An Empirical Study of Reasoning Guardrail Models
Sreedhar, M. N., Rebedea, T. E., & Parisien, C. (2025) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2025.findings-emnlp.1193 · Cited by 26
LLM Jailbreak Detection for (Almost) Free!
Chen, G., Xia, Y., Jia, X., Li, Z., Torr, P. H. S., & Gu, J. (2025) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2025.findings-emnlp.309 · Cited by 25
MULTIGUARD: An Efficient Approach for AI Safety Moderation Across Languages and Modalities
Verma, S., Hines, K., Bilmes, J., Siska, C., Zettlemoyer, L., Gonen, H., & Singh, C. (2025) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2025.emnlp-main.819 · Cited by 26
Causal evidence that language models use confidence to drive behaviour
Kumaran, D., Daw, N. D., Osindero, S., Veličković, P., & Pătrăucean, V. (2026) · Nature Machine Intelligence · doi:10.1038/s42256-026-01293-x · Cited by 26
Shifting Perspectives: Steering Vectors for Robust Bias Mitigation in LLMsPreprint
Siddique, Z., Khalid, I., Turner, L. D., & Espinosa-Anke, L. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2503.05371 · Cited by 25
Semantic-Space Exploration and Exploitation in RLVR for LLM ReasoningPreprint
Huang, F., Huang, G., Fan, X., He, Y., Liang, X., Chen, X., Jiang, Q., Khan, F. N., Jiang, J., & Wang, Z. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2509.23808 · Cited by 25
Emergent Misalignment is Easy, Narrow Misalignment is Hard
Soligo, A., Turner, E., Rajamanoharan, S., & Nanda, N. (2026) · International Conference on Learning Representations · doi:10.48550/arxiv.2602.07852 · Cited by 26
Self-citationNaseem, U. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.11180 · Cited by 25
Stronger Universal and Transferable Attacks by Suppressing Refusals
Huang, D., Shah, A., Araujo, A., Wagner, D., & Sitawarin, C. (2025) · Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) · doi:10.18653/v1/2025.naacl-long.302 · Cited by 24
COSMIC: Generalized Refusal Direction Identification in LLM Activations
Siu, V., Crispino, N., Yu, Z., Pan, S., Wang, Z., Liu, Y., Song, D. X., & Wang, C. (2025) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2025.findings-acl.1310 · Cited by 26
Bo, J. Y., Xu, T., Chatterjee, I., Passarella-Ward, K., Kulshrestha, A., & Shin, D. (2026) · ACM Symposium on User Interface Software and Technology (UIST) · doi:10.1145/3830398.3830533 · Cited by 24
Refusal Tokens: A Simple Way to Calibrate Refusals in Large Language ModelsPreprint
Jain, N., Shrivastava, A. D., Zhu, C., Liu, D., Samuel, A., Panda, A., Kumar, A., Goldblum, M., & Goldstein, T. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2412.06748 · Cited by 25
Du, H., Li, W., Cai, M., Saraipour, K., Zhang, Z., Lakkaraju, H., Sun, Y., & Zhang, S.-C. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2504.02904 · Cited by 25
Multi-lingual Multi-turn Automated Red Teaming for LLMs
Singhania, A., Dupuy, C., Mangale, S., & Namboori, A. (2025) · Workshop on Trustworthy NLP (TrustNLP) · doi:10.18653/v1/2025.trustnlp-main.11 · Cited by 23
EasySteer: A Unified Framework for High-Performance and Extensible LLM Steering
Xu, H., Mei, X., Yan, Y., Zhou, R., Zhang, W., Lü, W., Zhuang, Y., & Shen, Y. (2026) · Conference on Empirical Methods in Natural Language Processing: System Demonstrations (EMNLP) · doi:10.18653/2026.emnlp-demos.3 · Cited by 23
Improving LLM Reasoning through Interpretable Role-Playing Steering
Wang, A., Shu, D., Wang, Y., Ma, Y., & Du, M. (2025) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2025.findings-emnlp.39 · Cited by 23
Hua, P., Li, H., Shi, S., Yu, Z., & Zhang, N. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.992 · Cited by 23
Lu, X., Liu, D., Yu, Y., Xu, L., & Shao, J. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2502.09990 · Cited by 24
Chan, Y. S., Yong, Z.-X., & Bach, S. H. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2507.12428 · Cited by 24
Emotion Concepts and their Function in a Large Language ModelPreprint
Sofroniew, N., Kauvar, I., Saunders, W., Chen, R., Henighan, T., Hydrie, S., Citro, C., Pearce, A., Tarng, J., Gurnee, W., et al. (2026) · arXiv (Cornell University) · doi:10.5281/zenodo.20025418 · Cited by 23
Self-citationCan Role Vectors Affect LLM Behaviour?
Potertì, D., Seveso, A., & Mercorio, F. (2025) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2025.findings-emnlp.963 · Cited by 22
Geisler, S., Wollschläger, T., Abdalla, M. H. I., Cohen-Addad, V., Gasteiger, J., & Günnemann, S. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2502.17254 · Cited by 23
To Steer or Not to Steer? Mechanistic Error Reduction with Abstention for Language ModelsPreprint
Hedström, A., Amoukou, S. I., Bewley, T., Mishra, S., & Veloso, M. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2510.13290 · Cited by 22
SAKE: Steering Activations for Knowledge Editing
Scialanga, M., Laugel, T., Grari, V., & Detyniecki, M. (2025) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2025.acl-long.777 · Cited by 21
Universal Jailbreak Suffixes Are Strong Attention Hijackers
Ben-Tov, M., Geva, M., & Sharif, M. (2026) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl.a.695 · Cited by 22
Large Language Models Encode Semantics and Alignment in Linearly Separable RepresentationsPreprint
Sağlam, B., Kassianik, P., Nelson, B. A., Weerawardhena, S., Singer, Y., & Karbasi, A. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2507.09709 · Cited by 21
OGLS-SD: On-Policy Self-Distillation with Outcome-Guided Logit Steering for LLM ReasoningPreprint
Yang, Y., Wang, X., & Zhang, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.12400 · Cited by 21
CoBRA: Programming Cognitive Bias in Social Agents Using Classic Social Science Experiments
Liu, X., Shang, H., & Jin, H. (2026) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3772318.3790804 · Cited by 20
Precise In-Parameter Concept Erasure in Large Language Models
Gur-Arieh, Y., Suslik, C. H., Hong, Y., Barez, F., & Geva, M. (2025) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2025.emnlp-main.960 · Cited by 20
Improving Instruction-Following in Language Models through Activation SteeringPreprint
Stolfo, A., Balachandran, V., Yousefi, S., Horvitz, E., & Nushi, B. (2024) · Qeios · doi:10.32388/btht4k · Cited by 20
Inferring Functionality of Attention Heads from their Parameters
Elhelo, A., & Geva, M. (2025) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2025.acl-long.866 · Cited by 19
Refusal Behavior in Large Language Models: A Nonlinear Perspective
Hildebrandt, F., Maier, A., Krauss, P., & Schilling, A. (2025) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn64981.2025.11228760 · Cited by 20
xJailbreak: Representation Space Guided Reinforcement Learning for Interpretable LLM JailbreakingPreprint
Lee, S., Ni, S., Wei, C., Li, S., Fan, L., Argha, A., Alinejad‐Rokny, H., Xu, R., Gong, Y., & Yang, M. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2501.16727 · Cited by 20
FairSteer: Inference Time Debiasing for LLMs with Dynamic Activation SteeringPreprint
Li, Y., Fan, Z., Chen, R., Gai, X., Gong, L., Zhang, Y., & Liu, Z. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2504.14492 · Cited by 19
Using Mechanistic Interpretability to Craft Adversarial Attacks against Large Language ModelsPreprint
Winninger, T., Addad, B., & Kapusta, K. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2503.06269 · Cited by 20
DiffuGuard: How Intrinsic Safety is Lost and Found in Diffusion Large Language Models
Li, Z., Nie, Z., Zhou, Z., Liu, Y., Zhang, Y., Cheng, Y., Wen, Q., Wang, K., Guo, Y., & Zhang, J. (2026) · International Conference on Learning Representations · doi:10.48550/arxiv.2509.24296 · Cited by 19
Spherical Steering: Geometry-Aware Activation Rotation for Language ModelsPreprint
You, Z., Deng, C., & Chen, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.08169 · Cited by 19
LLMs Encode Their Failures: Predicting Success from Pre-Generation ActivationsPreprint
Lugoloobi, W., Foster, T., Bankes, W., & Russell, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.09924 · Cited by 19
Mechanisms of Introspective AwarenessPreprint
Macar, U., Yang, L., Wang, A., Wallich, P., Ameisen, E., & Lindsey, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.21396 · Cited by 19
Understanding and Rectifying Safety Perception Distortion in VLMs
Zou, X., Kang, J., Kesidis, G., & Lu, L. (2025) · neural information processing systems · doi:10.52202/085713-3810 · Cited by 19
From Word to World: Evaluate and Mitigate Culture Bias in LLMs via Word Association Test
Dai, X., Zhou, L., Wang, B., & Li, H. (2025) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2025.emnlp-main.1246 · Cited by 19
SOM Directions Are Better than One: Multi-Directional Refusal Suppression in Language Models
Piras, G., Mura, R., Brau, F., Oneto, L., Roli, F., & Biggio, B. (2026) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v40i39.40551 · Cited by 19
Can Aha Moments be Fake? Towards Quantifying Decorative and True Thinking in Chain-of-Thought
Zhao, J., Sun, Y., Shi, W., & Song, D. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/2026.findings-emnlp.580 · Cited by 18
SAGE: An Agentic Explainer Framework for Interpreting SAE Features in Language Models
Han, J., Xu, W., Jin, M., & Du, M. (2026) · Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2026.eacl-industry.37 · Cited by 19
Why do LLaVA Vision-Language Models Reply to Images in English?
Hinck, M., Holtermann, C., Olson, M., Schneider, F., Yu, S., Bhiwandiwalla, A., Lauscher, A., Tseng, S.-Y., & Lal, V. (2024) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2024.findings-emnlp.783 · Cited by 18
West, R., & Aydin, R. C. (2025) · Communications of the ACM · doi:10.1145/3705294 · Cited by 18
Are language models aware of the road not taken? Token-level uncertainty and hidden state dynamicsPreprint
Zur, A., Geiger, A., Lubana, E. S., & Bigelow, E. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2511.04527 · Cited by 18
Steering Language Models with Weight Arithmetic
Fierro, C., & Roger, F. (2026) · International Conference on Learning Representations · doi:10.48550/arxiv.2511.05408 · Cited by 19
ICON: Indirect Prompt Injection Defense for Agents based on Inference-Time CorrectionPreprint
Wang, C., Zhang, F., Zhang, J., Zhang, Z., Wang, Y., Huang, L., Gao, J., Chen, Z., & Lim, W. Y. B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.20708 · Cited by 18
Valence-Arousal Subspace in LLMs: Circular Emotion Geometry and Multi-Behavioral ControlPreprint
Sun, L., Yan, L., Lu, X., Lee, A., Zhang, J., & Shao, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.03147 · Cited by 18
Li, Y., Fan, Z., & Zhuang, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.22864 · Cited by 19
Xie, Y., Zhang, Y., Liu, T., Ma, D., & Liu, T. (2025) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2025.findings-emnlp.956 · Cited by 18
Learning to Steer: Input-dependent Steering for Multimodal LLMs
Parekh, J., Khayatan, P., Shukor, M., Dapogny, A., Newson, A., & Cord, M. (2025) · neural information processing systems · doi:10.52202/085713-5341 · Cited by 17
Zhao, W., Guo, J., Deng, Y., Sui, X., Hu, Y., Zhao, Y., Che, W., Qin, B., Chua, T.-S., & Liu, T. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2506.15647 · Cited by 17
Activation Space Interventions Can Be Transferred Between Large Language ModelsPreprint
Oozeer, N. F., Nathawani, D., Prakash, N., Lan, M., Harrasse, A., & Abdullah, A. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2503.04429 · Cited by 17
Does Machine Unlearning Truly Remove Knowledge?Preprint
Chen, H., Zhang, Y., Bi, Y., Zhang, Y., Liu, T., Bi, J., Lan, J., Gu, J., Großer, C., Krompaß, D., et al. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2505.23270 · Cited by 18
The Information Geometry of Softmax: Probing and SteeringPreprint
Park, K., Nief, T., Choe, Y. J., & Veitch, V. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.15293 · Cited by 17
PERSONA: Dynamic and Compositional Inference-Time Personality Control via Activation Vector Algebra
Feng, X., Zhao, L., Zhong, W., Huang, Y., Gu, Y., Kong, L., Feng, X., & Qin, B. (2026) · International Conference on Learning Representations · doi:10.48550/arxiv.2602.15669 · Cited by 19
International AI Safety Report
Bengio, Y., Mindermann, S., Privitera, D., Besiroglu, T., Bommasani, R., Casper, S., Choi, Y., Fox, P., Garfinkel, B., Goldfarb, D., et al. (2025) · arXiv · doi:10.48550/arxiv.2501.17805 · Cited by 16
Analysing the Safety Pitfalls of Steering Vectors
Li, Y., Fastowski, A., Zaradoukas, E., Prenkaj, B., & Kasneci, G. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.544 · Cited by 16
Gradient-based Jailbreak Images for Multimodal Fusion ModelsPreprint
Rando, J., Korevaar, H., Brinkman, E., Evtimov, I., & Tramèr, F. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2410.03489 · Cited by 16
Countering Autonomous Cyber ThreatsPreprint
Heckel, K., & Weller, A. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2410.18312 · Cited by 16
Activation Steering with a Feedback Controller
Nguyen, D. V., Vu, H. M., Pham, N. Y., Zhang, L., & Nguyen, T. M. (2025) · International Conference on Learning Representations · doi:10.48550/arxiv.2510.04309 · Cited by 16
Guardians and Offenders: A Survey on Harmful Content Generation and Safety Mitigation of LLMPreprint
Zhang, C., Zhu, C., Xiong, J., Xu, X., Li, L., Liu, Y., & Lu, Z. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2508.05775 · Cited by 17
Semantic Structure in Large Language Model EmbeddingsPreprint
Kozlowski, A. C., Dai, C., & Boutyline, A. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2508.10003 · Cited by 17
Karny, S., Baez, A., & Pataranutaporn, P. (2026) · ACM International Conference on Intelligent User Interfaces (IUI) · doi:10.1145/3742413.3789120 · Cited by 16
Overcoming Sparsity Artifacts in Crosscoders to Interpret Chat-Tuning
Minder, J., Dumas, C., Juang, C., Chughtai, B., & Nanda, N. (2025) · neural information processing systems · doi:10.52202/085713-3551 · Cited by 15
Self-citationInterpretation Meets Safety: A Survey on Interpretation Methods and Tools for Improving LLM Safety
Lee, S., Cho, A., Kim, G., Peng, S., Phute, M., & Chau, D. H. (2025) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2025.emnlp-main.1091 · Cited by 15
Refusal Falls off a Cliff: How Safety Alignment Fails in Reasoning?Preprint
Yin, Q., Leong, C. T., Yang, L., Huang, W., Li, W., Wang, X., Yoon, J., YunXing, XingYu, & Gu, J. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2510.06036 · Cited by 16
Activation Patching for Interpretable Steering in Music GenerationPreprint
Facchiano, S., Strano, G., Crisostomi, D., Tallini, I., Mencattini, T., Galasso, F., & Rodolà, E. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2504.04479 · Cited by 15
Unforgotten Safety: Preserving Safety Alignment of Large Language Models with Continual Learning
Alssum, L., Itani, H., Hammoud, H. A. A. K., Torr, P. H. S., Bibi, A., & Ghanem, B. (2026) · Transactions on machine learning research · doi:10.48550/arxiv.2512.10150 · Cited by 16
Xiong, C., He, Z., Chen, P.-Y., Ko, C.-Y., & Ho, T.-Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.04896 · Cited by 18
Positive Alignment: Artificial Intelligence for Human FlourishingPreprint
Laukkonen, R., Krier, S., Bakalar, C., Chandaria, S., Kringelbach, M., Elwood, A., Ford, D., Rosas, F. E., Boháček, M., Franklin, M., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.10310 · Cited by 16
Linear Representations of Political Perspective Emerge in Large Language Models
Kim, J., Evans, J., & Schein, A. (2025) · arXiv · doi:10.48550/arxiv.2503.02080 · Cited by 14
Soteria: Language-Specific Functional Parameter Steering for Multilingual Safety Alignment
Banerjee, S., Layek, S., Chatterjee, P., Mukherjee, A., & Hazra, R. (2025) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2025.findings-emnlp.497 · Cited by 14
Lu, X., Liu, D., Yu, Y., Xu, L., & Shao, J. (2025) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2025.findings-emnlp.282 · Cited by 16
Detecting (Un)answerability in Large Language Models with Linear Directions
Lavi, M., Milo, T., & Geva, M. (2026) · Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2026.eacl-long.29 · Cited by 14
SafeSteer: Interpretable Safety Steering with Refusal-Evasion in LLMsPreprint
Ghosh, S., Bhattacharjee, A., Ziser, Y., & Parisien, C. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2506.04250 · Cited by 14
Enhancing LLM Steering through Sparse Autoencoder-Based Vector RefinementPreprint
Wang, A., Wu, X., Shu, D., Ma, Y., & Liu, N. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2509.23799 · Cited by 15
Steering the CensorShip: Uncovering Representation Vectors for LLM "Thought" ControlPreprint
Cyberey, H., & Evans, D. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2504.17130 · Cited by 14
Focus Directions Make Your Language Models Pay More Attention to Relevant ContextsPreprint
Zhu, Y., Li, R., Wang, D., Haehn, D., & Liang, X. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2503.23306 · Cited by 14
Yong, Z.-X., & Bach, S. H. (2026) · International Conference on Learning Representations · doi:10.48550/arxiv.2510.20956 · Cited by 15
Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMsPreprint
Singh, N. D., Müller, M., Croce, F., & Hein, M. (2025) · arXiv (Cornell University) · doi:10.15496/publikation-113058 · Cited by 14
Representation Bending for Large Language Model SafetyPreprint
Yousefpour, A., Kim, T., Kwon, R. S., Lee, S., Jeung, W., Han, S., Wan, A., Ngan, H., Yu, Y., & Choi, J. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2504.01550 · Cited by 14
LoX: Low-Rank Extrapolation Robustifies LLM Safety Against Fine-tuningPreprint
Perin, G. J., Chen, R., Chen, X., Hirata, N. S. T., Wang, Z., & Hong, J. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2506.15606 · Cited by 14
Analyzing Finetuning Representation Shift for Multimodal LLMs SteeringPreprint
Khayatan, P., Shukor, M., Parekh, J., Dapogny, A., & Cord, M. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2501.03012 · Cited by 15
Tool Calling is Linearly Readable and Steerable in Language ModelsPreprint
Wu, Z., Wang, Z., Cho, S., Yang, Y., Koshiyama, A., Bulathwela, S., & Pérez‐Ortiz, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.07990 · Cited by 14
The Pain Axis: LLMs Represent Self-Directed Harm and Act on ItPreprint
Tagliabue, V., Dung, L., & Berg, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.16247 · Cited by 16
Yi, X., Li, Y., Shi, D., Wang, L., Wang, X., & He, L. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.129101 · Cited by 13
Attention Speaks Volumes: Localizing and Mitigating Bias in Language Models
Adiga, R., Nushi, B., & Chandrasekaran, V. (2025) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2025.acl-long.1281 · Cited by 13
StructTransform: A Scalable Attack Surface for Safety-Aligned Large Language Models
Yoosuf, S., Ali, T., Lekssays, A., AlSabah, M., & Khalil, I. M. (2025) · Lecture notes in computer science · doi:10.1007/978-3-032-07884-1_25 · Cited by 13
Jailbreak Attack Initializations as Extractors of Compliance Directions
LeVi, A., Himelstein, R., Nemcovsky, Y., Mendelson, A., & Baskin, C. (2025) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2025.findings-emnlp.354 · Cited by 13
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.