References

Citation edges between papers. Most are from OpenAlex; some from LLM-extracted bibliographies. Showing latest 500 of 7647.

Total citations

7647

Unique citing papers

111

Unique cited papers

5498

By source

s2-reference-walk4565

openalex2098

llm984

Citing paper	Cited paper	DOI	Status	Source
Fractal basins trap latent reasoning	Less is more: Recursive reasoning with tiny networks (2025)	—	referenced-only	s2-reference-walk
Fractal basins trap latent reasoning	Hierarchical reasoning model (2025)	—	referenced-only	s2-reference-walk
Fractal basins trap latent reasoning	Normally Hyperbolic Invariant Manifolds in Dynamical Systems	10.1007/978-1-4612-4312-0	referenced-only	s2-reference-walk
Fractal basins trap latent reasoning	Center for AI Safety, Scale AI, and HLE Contributors Consortium	—	referenced-only	s2-reference-walk
Fractal basins trap latent reasoning	Transient Chaos: Complex Dynamics on Finite Time Scales	—	referenced-only	s2-reference-walk
Fractal basins trap latent reasoning	Fixed-Point Reasoners: Stable and Adaptive Deep Looped Transformers	10.48550/arxiv.2606.18206	referenced-only	s2-reference-walk
Fractal basins trap latent reasoning	Parcae: Scaling Laws For Stable Looped Language Models	10.48550/arxiv.2604.12946	referenced-only	s2-reference-walk
Fractal basins trap latent reasoning	(Preprint)	10.2196/preprints.12431	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	, Sylvain Gugger , Mariama Drame , Quentin Lhoest , and Alexander Rush . 2020 .	—	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	The use of MMR, diversity-based reranking for reordering documents and producing	10.1145/290941.291025	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	2020), our results confirm that nucleus sampling is also able to generate more s	—	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Bleu: a Method for Automatic Evaluation of Machine Translation	10.3115/1073083.1073135	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Proceedings of the 2013 Conference of the North American Chapter of the Associat	—	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	A Systematic Exploration of Diversity in Machine Translation	10.18653/v1/d13-1111	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Diverse M-Best Solutions in Markov Random Fields	10.1007/978-3-642-33715-4_	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	A Diversity-Promoting Objective Function for Neural Conversation Models	10.18653/v1/n16-1014	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	A large annotated corpus for learning natural language inference	10.18653/v1/d15-1075	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	A Broad-Coverage Challenge Corpus for Sentence Understanding through Inference	10.18653/v1/n18-1101	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Latent Variable Dialogue Models and their Diversity	10.18653/v1/e17-2029	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	DailyDialog: A Manually Labelled Multi-turn Dialogue Dataset	—	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	SemEval-2017 Task 1: Semantic Textual Similarity Multilingual and Crosslingual F	10.18653/v1/s17-2001	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Texygen: A Benchmarking Platform for Text Generation Models	10.1145/3209978.3210080	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Generating Informative and Diverse Conversational Responses via Adversarial Info	—	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	FEVER: a Large-scale Dataset for Fact Extraction and VERification	10.18653/v1/n18-1074	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Combining Fact Extraction and Verification with Neural Semantic Matching Network	10.1609/aaai.v33i01.330168	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Towards Empathetic Open-domain Conversation Models: A New Benchmark and Dataset	10.18653/v1/p19-1534	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	What makes a good conversation? How controllable attributes affect human judgmen	10.18653/v1/n19-1170	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Outlier Detection for Improved Data Quality and Diversity in Dialog Systems	10.18653/v1/n19-1051	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Unifying Human and Statistical Evaluation for Natural Language Generation	10.18653/v1/n19-1169	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	BERTScore: Evaluating Text Generation with BERT	—	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	The Curious Case of Neural Text Degeneration	—	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Approximating Interactive Human Evaluation with Self-Play for Open-Domain Dialog	—	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Syntax-Infused Variational Autoencoder for Text Generation	10.18653/v1/p19-1199	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Investigating Evaluation of Open-Domain Dialogue Systems With Human Generated Mu	10.18653/v1/w19-5944	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	RoBERTa: A Robustly Optimized BERT Pretraining Approach	—	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks	10.18653/v1/d19-1410	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	DIALOGPT : Large-Scale Generative Pre-training for Conversational Response Gener	10.18653/v1/2020.acl-demos	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Adversarial NLI: A New Benchmark for Natural Language Understanding	10.18653/v1/2020.acl-main.	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Generating Persona Consistent Dialogues by Exploiting Natural Language Inference	10.1609/aaai.v34i05.6417	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Evaluating the Evaluation of Diversity in Natural Language Generation	10.18653/v1/2021.eacl-main	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	More Diverse Dialogue Datasets via Diversity-Informed Data Collection	10.18653/v1/2020.acl-main.	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Recipes for Building an Open-Domain Chatbot	10.18653/v1/2021.eacl-main	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Improving Dialog Evaluation with a Multi-reference Adversarial Dataset and Large	10.1162/tacl_a_00347	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Transformers: State-of-the-Art Natural Language Processing	—	referenced-only	s2-reference-walk
Semantic Diversity in Dialogue with Natural Language Inferen	Choose Your Own Adventure: Paired Suggestions in Collaborative Writing for Evalu	10.18653/v1/2021.naacl-mai	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Using attention sinks to identify and evaluate dormant heads in pretrained llms	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	nanochat: The best ChatGPT that $100 can buy, 2025	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Deepseek-r1: In-centivizing reasoning capability in llms via reinforcement learn	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	A Survey on Over-smoothing and Over-squashing: Unified Propagation Perspectives	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Less is more: Recursive reasoning with tiny networks (2025)	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Adaptive Computation Time for Recurrent Neural Networks	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Attention is All you Need	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Universal Transformers	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	HellaSwag: Can a Machine Really Finish Your Sentence?	10.18653/v1/p19-1472	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Deep Equilibrium Models	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	On the Bottleneck of Graph Neural Networks and its Practical Implications	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	A Note on Over-Smoothing for Graph Neural Networks	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	PonderNet: Learning to Ponder	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Can You Learn an Algorithm? Generalizing from Easy to Hard Problems with Recurre	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Training verifiers to solve math word problems	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Chain of Thought Prompting Elicits Reasoning in Large Language Models	10.52202/068431-1800	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	End-to-end Algorithm Synthesis with Recurrent Networks: Logical Extrapolation Wi	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Confident Adaptive Language Modeling	10.48550/arxiv.2207.07061	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Path Independent Equilibrium Models Can Better Exploit Test-Time Computation	10.48550/arxiv.2211.09961	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Efficient Streaming Language Models with Attention Sinks	10.48550/arxiv.2309.17453	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Sparse Universal Transformer	10.48550/arxiv.2310.07096	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Looped Transformers are Better at Learning Learning Algorithms	10.48550/arxiv.2311.12424	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Universal Neurons in GPT2 Language Models	10.48550/arxiv.2401.12181	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Transformers Can Do Arithmetic with the Right Embeddings	10.48550/arxiv.2405.17399	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Massive Activations in Large Language Models	10.48550/arxiv.2402.17762	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	The Remarkable Robustness of LLMs: Stages of Inference?	10.48550/arxiv.2406.19384	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Transformers need glasses! Information over-squashing in language tasks	10.48550/arxiv.2406.04267	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	On Expressive Power of Looped Transformers: Theoretical Analysis and Enhancement	10.48550/arxiv.2410.01405	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Softmax is not Enough (for Sharp Size Generalisation)	—	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	When Attention Sink Emerges in Language Models: An Empirical View	10.48550/arxiv.2410.10781	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Advancing the Understanding of Fixed Point Iterations in Deep Neural Networks: A	10.48550/arxiv.2410.11279	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Rethinking Deep Thinking: Stable Learning of Algorithms using Lipschitz Constrai	10.48550/arxiv.2410.23451	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Layer by Layer: Uncovering Hidden Representations in Language Models	10.48550/arxiv.2502.02013	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach	10.48550/arxiv.2502.05171	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	On Vanishing Gradients, Over-Smoothing, and Over-Squashing in GNNs: Bridging Rec	10.48550/arxiv.2502.10818	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Reasoning with Latent Thoughts: On the Power of Looped Transformers	10.48550/arxiv.2502.17416	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Why do LLMs attend to the first token?	10.48550/arxiv.2504.02732	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Return of ChebNet: Understanding and Improving an Overlooked GNN on Long Range T	10.48550/arxiv.2506.07624	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Continuous Thought Machines	10.48550/arxiv.2505.05522	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Hierarchical Reasoning Model	10.48550/arxiv.2506.21734	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Latent Chain-of-Thought? Decoding the Depth-Recurrent Transformer	10.48550/arxiv.2507.02199	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Pay Attention to Attention Distribution: A New Local Lipschitz Bound for Transfo	10.48550/arxiv.2507.07814	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Mixture-of-Recursions: Learning Dynamic Recursive Depths for Adaptive Token-Leve	10.48550/arxiv.2507.10524	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Two-Scale Latent Dynamics for Recurrent-Depth Transformers	10.48550/arxiv.2509.23314	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Attention Sinks and Compression Valleys in LLMs are Two Sides of the Same Coin	10.48550/arxiv.2510.06477	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Encode, Think, Decode: Scaling test-time reasoning with recursive latent thought	10.48550/arxiv.2510.07358	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	gLSTM: Mitigating Over-Squashing by Increasing Storage Capacity	10.48550/arxiv.2510.08450	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Scaling Latent Reasoning via Looped Language Models	10.48550/arxiv.2510.25741	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Softmax is 1/2-Lipschitz: A tight bound across all ℓp norms	10.48550/arxiv.2510.23012	referenced-only	s2-reference-walk
A Mechanistic Analysis of Looped Reasoning Language Models	Teaching Pretrained Language Models to Think Deeper with Retrofitted Recurrence	10.48550/arxiv.2511.07384	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Towards monosemanticity: Decomposing language models with dictionary learning	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Open Problems in Mechanistic Interpretability, January 2025	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	The Scales of Justitia: A Com-13	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Diff-in-Means Concept Editing is Worst-Case Optimal	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Catastrophic Interference in Connectionist Networks: The Sequential Learning Pro	10.1016/s0079-7421(08)6053	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Dynamically Scaled Activation Steering, December 2025	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Evaluating feature steering: A case study in mitigating social biases	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Emergent misalignment: Narrow finetuning can produce broadly misaligned LLMs	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Style Modulation Heads for Robust Persona Control	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Activation Steering Decoding: Mitigating Hallucination in Large Vision-Language	10.18653/v1/2025.acl-long.	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Copy Suppression: Comprehensively Understanding a Motif in Language Model Attent	10.18653/v1/2024.blackboxn	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Can Role Vectors Affect LLM Behaviour?	10.18653/v1/2025.findings-	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Inaugural Article: Functional specificity in the human brain: A window into the	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Technical	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Safety in research	10.1002/9781118643440.ch9	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Dynamic Steering With Episodic Memory For Large Language Models	10.18653/v1/2025.findings-	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Attention is All you Need	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Performance metrics in multi-objective optimization	10.1109/clei.2015.7360024	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Understanding intermediate layers using linear classifier probes	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Switch Transformers: Scaling to Trillion Parameter Models with Simple and Effici	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Locating and editing factual associations in GPT	10.52202/068431-1262	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Measuring Massive Multitask Language Understanding	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Transformer feed-forward layers are key-value memories	10.18653/v1/2021.emnlp-mai	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	In-context Learning and Induction Heads	10.48550/arxiv.2209.11895	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Extracting Latent Steering Vectors from Pretrained Language Models	10.48550/arxiv.2205.05124	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Discovering Latent Knowledge in Language Models Without Supervision	10.48550/arxiv.2212.03827	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	HiFi: High-Information Attention Heads Hold for Parameter-Efficient Model Adapta	10.48550/arxiv.2305.04573	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	OpenAssistant Conversations - Democratizing Large Language Model Alignment	10.48550/arxiv.2304.07327	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Judging LLM-as-a-judge with MT-Bench and Chatbot Arena	10.52202/075280-2020	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Inference-time intervention: eliciting truthful answers from a language model	10.52202/075280-1797	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpo	10.48550/arxiv.2305.13245	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Representation engineering: A top-down approach to AI transparency	10.48550/arxiv.2310.01405	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Steering language models with activation engineering	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Sparse autoencoders find highly interpretable features in language models	10.48550/arxiv.2309.08600	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Language models represent space and time	10.48550/arxiv.2310.02207	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	The linear representation hypothesis and the geometry of large language models	10.48550/arxiv.2311.03658	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	The Geometry of Truth: Emergent Linear Structure in Large Language Model Represe	10.48550/arxiv.2310.06824	active	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	The steerability of large language models toward data-driven personas	10.48550/arxiv.2311.04978	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Instruction-following evaluation for large language models	10.48550/arxiv.2311.07911	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Uncovering Safety Risks of Large Language Models through Concept Activation Vect	10.52202/079017-3706	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	LLM Evaluators Recognize and Favor Their Own Generations	10.48550/arxiv.2404.13076	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Steering Llama 2 via Contrastive Activation Addition	10.48550/arxiv.2312.06681	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Adaptive Activation Steering: A Tuning-Free LLM Truthfulness Improvement Method	10.1145/3696410.3714640	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Mechanistic Interpretability for AI Safety - A Review	10.48550/arxiv.2404.14082	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Refusal in language models is mediated by a single direction	10.48550/arxiv.2406.11717	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Attention heads of large language models	10.1016/j.patter.2025.1011	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	SCANS: Mitigating the Exaggerated Safety for LLMs via Safety-Conscious Activatio	10.1609/aaai.v39i22.34521	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Controlling Language and Diffusion Models by Transporting Activations	10.48550/arxiv.2410.23054	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Programming Refusal with Conditional Activation Steering	10.48550/arxiv.2409.05907	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	SparseGrad: A Selective Method for Efficient Fine-tuning of MLP Layers	10.48550/arxiv.2410.07383	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Sparse Autoencoders Trained on the Same Data Learn Different Features	10.48550/arxiv.2501.16615	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Identifying and Manipulating Personality Traits in LLMs Through Activation Engin	10.48550/arxiv.2412.10427	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Interpretable Steering of Large Language Models with Feature Guided Activation A	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	SafeSwitch: Steering Unsafe LLM Behavior via Internal Activation Signals	10.18653/v1/2025.findings-	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Shifting Perspectives: Steering Vectors for Robust Bias Mitigation in LLMs	10.18653/v1/2026.findings-	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Designing Role Vectors to Improve LLM Inference Behaviour	10.48550/arxiv.2502.12055	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	LinEAS: End-to-end Learning of Activation Steering with a Distributional Loss	10.52202/085713-2756	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	AdaSteer: Your Aligned LLM is Inherently an Adaptive Jailbreak Defender	10.48550/arxiv.2504.09466	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	SAEs Are Good for Steering - If You Select the Right Features	10.18653/v1/2025.emnlp-mai	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Guiding Giants: Lightweight Controllers for Weighted Activation Steering in LLMs	10.48550/arxiv.2505.20309	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Beyond Prompt Engineering: Robust Behavior Control in LLMs via Steering Target A	10.48550/arxiv.2505.20322	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Persona Vectors: Monitoring and Controlling Character Traits in Language Models	10.48550/arxiv.2507.21509	active	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	The Scales of Justitia: A Comprehensive Survey on Safety Evaluation of LLMs	10.1016/j.inffus.2026.1045	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	In-Distribution Steering: Balancing Control and Coherence in Language Model Gene	10.48550/arxiv.2510.13285	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Encode, Think, Decode: Scaling test-time reasoning with recursive latent thought	10.48550/arxiv.2510.07358	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	BILLY: Steering Large Language Models via Merging Persona Vectors for Creative G	10.48550/arxiv.2510.10157	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Temporal Sparse Autoencoders: Leveraging the Sequential Nature of Language for I	10.48550/arxiv.2511.05541	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	To Steer or Not to Steer? Mechanistic Error Reduction with Abstention for Langua	10.48550/arxiv.2510.13290	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	What Can We Actually Steer? A Multi-Behavior Study of Activation Control	—	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Angular Steering: Behavior Control via Rotation in Activation Space	10.48550/arxiv.2510.26243	referenced-only	s2-reference-walk
Steering at the Source: Style Modulation Heads for Robust Pe	Emergent Introspective Awareness in Large Language Models	10.48550/arxiv.2601.01828	active	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Facet-level Persona Control with Contrastive SAE for Role-Playing	—	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Introducing gpt-5: A unified system for expert-level reasoning and multi-modal c	—	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	The revised NEO personality inventory (NEO-PI-R)	10.4135/9781849200479.n9	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	SAEs (usually) transfer between base and chat models	—	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Recurrent personality factors based on trait ratings.	10.1111/j.1467-6494.1992.t	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Evaluating feature steering: A case study in mitigating social biases	—	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Egoistic and Moralistic Biases in Self-Perception: The Interplay of Self-Decepti	10.1111/1467-6494.00041	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Activation addition: Steering language models without optimization	—	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	ArcFace: Additive Angular Margin Loss for Deep Face Recognition	10.1109/cvpr.2019.00482	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	CosFace: Large Margin Cosine Loss for Deep Face Recognition	10.1109/cvpr.2018.00552	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	A personality trait-based interactionist model of job performance.	10.1037/0021-9010.88.3.500	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	CTRL: A Conditional Transformer Language Model for Controllable Generation	—	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	A Simple Framework for Contrastive Learning of Visual Representations	—	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	A Plug-and-Play Method for Controlled Text Generation	10.18653/v1/2021.findings-	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	InCharacter: Evaluating Personality Fidelity in Role-Playing Agents through Psyc	10.18653/v1/2024.acl-long.	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Mistral 7B	10.48550/arxiv.2310.06825	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Parameter-efficient fine-tuning of large-scale pre-trained language models	10.1038/s42256-023-00626-4	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Steering Llama 2 via Contrastive Activation Addition	10.48550/arxiv.2312.06681	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Style Vectors for Steering Generative Large Language Models	10.48550/arxiv.2402.01618	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	From Persona to Personalization: A Survey on Role-Playing Language Agents	10.48550/arxiv.2404.18231	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	A comprehensive survey on contrastive learning	10.1016/j.neucom.2024.1286	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Two Tales of Persona in LLMs: A Survey of Role-Playing and Personalization	10.48550/arxiv.2406.01171	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	GPT-4o System Card	—	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Evaluating the ability of large language models to emulate personality	10.1038/s41598-024-84109-5	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Steering off course: Reliability challenges in steering language models	10.48550/arxiv.2504.04635	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Identifying and Manipulating Personality Traits in LLMs Through Activation Engin	10.48550/arxiv.2412.10427	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Denoising Concept Vectors with Sparse Autoencoders for Improved Language Model S	10.48550/arxiv.2505.15038	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	SAE-SSV: Supervised Steering in Sparse Representation Spaces for Reliable Contro	10.48550/arxiv.2505.16188	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Qwen3 Technical Report	—	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Improving LLM Reasoning through Interpretable Role-Playing Steering	10.48550/arxiv.2506.07335	referenced-only	s2-reference-walk
Facet-Level Persona Control by Trait-Activated Routing with	Persona Vectors: Monitoring and Controlling Character Traits in Language Models	10.48550/arxiv.2507.21509	active	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Qwq: Reflect deeply on the boundaries of the unknown,	—	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	The Big Five Trait taxonomy: History, measurement, and theoretical perspectives.	—	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	A behavioral genetic investigation of the Dark Triad and the Big 5	10.1016/j.paid.2007.09.007	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	THE DEVELOPMENT OF MARKERS FOR THE BIG-FIVE FACTOR STRUCTURE	10.1037/1040-3590.4.1.26	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learni	10.48550/arxiv.2501.12948	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	The Dark Triad of personality: Narcissism, Machiavellianism, and psychopathy	10.1016/s0092-6566(02)0050	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	(Preprint)	10.2196/preprints.12431	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Evaluating and Inducing Personality in Pre-trained Language Models	10.52202/075280-0466	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	PersonaLLM: Investigating the Ability of Large Language Models to Express Person	10.18653/v1/2024.findings-	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Toxicity in ChatGPT: Analyzing Persona-assigned Language Models	10.48550/arxiv.2304.05335	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	CAT'S THEORY: Empirical Validation and Architectural Applications Cross-Architec	10.48550/arxiv.2212.08073	active	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Personality Traits in Large Language Models	10.48550/arxiv.2307.00184	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Representation engineering: A top-down approach to AI transparency	10.48550/arxiv.2310.01405	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Steering language models with activation engineering	—	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not I	10.48550/arxiv.2310.03693	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Safe RLHF: Safe Reinforcement Learning from Human Feedback	10.48550/arxiv.2310.12773	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Editing Personality For Large Language Models	10.1007/978-981-97-9434-8_	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Scalable and Transferable Black-Box Jailbreaks for Language Models via Persona M	10.48550/arxiv.2311.03348	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations	10.48550/arxiv.2312.06674	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Steering Llama 2 via Contrastive Activation Addition	10.48550/arxiv.2312.06681	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	SALAD-Bench: A Hierarchical and Comprehensive Safety Benchmark for Large Languag	10.48550/arxiv.2402.05044	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Rob	10.48550/arxiv.2402.04249	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	A StrongREJECT for Empty Jailbreaks	10.48550/arxiv.2402.10260	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Mod	10.48550/arxiv.2404.01318	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Two Tales of Persona in LLMs: A Survey of Role-Playing and Personalization	10.48550/arxiv.2406.01171	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Refusal in language models is mediated by a single direction	10.48550/arxiv.2406.11717	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	2024. Ask LLMs directly, “what	—	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Who's asking? User personas and the mechanics of latent misalignment	10.48550/arxiv.2406.12094	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	2023. Do personality tests generalize to large language models? In Socially Resp	—	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	2024a. Evaluating and inducing personality in pre-trained language models	—	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Towards Understanding Safety Alignment: A Mechanistic Perspective from Safety Ne	10.52202/085713-0426	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	The llama 3 herd of models	10.48550/arxiv.2407.21783	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	PERSONACHATGEN: Generating Personalized Dialogues using GPT-3	—	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	THE DEVELOPMENT OF MARKERS FOR THE BIG-FIVE FACTOR STRUCTURE	10.1037/1040-3590.4.1.26	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Safety Layers in Aligned Large Language Models: The Key to LLM Security	—	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Measuring nominal scale agreement among many raters.	10.1037/h0031619	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Evaluating Content Selection in Summarization: The Pyramid Method	10.7916/d80r9xvd	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Gemma 2: Improving Open Language Models at a Practical Size	10.48550/arxiv.2408.00118	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	The Hidden Dimensions of LLM Alignment: A Multi-Dimensional Analysis of Orthogon	—	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	BIG5-CHAT: Shaping LLM Personalities Through Training on Human-Grounded Data	10.48550/arxiv.2410.16491	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Deliberative Alignment: Reasoning Enables Safer Language Models	10.48550/arxiv.2412.16339	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	A NEW MEASURE OF RANK CORRELATION	10.1093/biomet/30.1-2.81	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	A metric for distributions with applications to image databases	10.1109/iccv.1998.710701	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	A basis for analyzing test-retest reliability	10.1007/bf02288892	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	An Empirical Analysis of the Writing Styles of Persona-Assigned LLMs	10.18653/v1/2024.emnlp-mai	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Personalizing Dialogue Agents: I have a dog, do you have pets too?	10.18653/v1/p18-1205	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	H-CoT: Hijacking the Chain-of-Thought Safety Reasoning Mechanism to Jailbreak La	10.48550/arxiv.2502.12893	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Assessing the Big Five personality traits with latent semantic analysis	10.1016/j.paid.2016.07.010	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	A structured interview for the assessment of the Five-Factor Model of Personalit	10.1037/1040-3590.10.3.229	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Exploring the Impact of Personality Traits on LLM Bias and Toxicity	10.48550/arxiv.2502.12566	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Automatic personality assessment through social media language.	10.1037/pspp0000020	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	The Hidden Risks of Large Reasoning Models: A Safety Assessment of R1	10.48550/arxiv.2502.12659	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Beware of Your Po! Measuring and Mitigating AI Safety Risks in Role-Play Fine-Tu	10.48550/arxiv.2502.20968	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Crowdsourcing Lightweight Pyramids for Manual Summary Evaluation	10.18653/v1/n19-1072	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	The Geometry of Refusal in Large Language Models: Concept Cones and Representati	10.48550/arxiv.2502.17420	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	The Rise of Darkness: Safety-Utility Trade-Offs in Role-Playing Dialogue Agents	10.48550/arxiv.2502.20757	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Do Long-Range Language Models Actually Use Long-Range Context?	10.18653/v1/2021.emnlp-mai	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	INVENTORIES	10.1142/9789814273831_0017	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Safety Tax: Safety Alignment Makes Your Large Reasoning Models Less Reasonable	10.48550/arxiv.2503.00555	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	DeepSeek-R1 Thoughtology: Let's think about LLM Reasoning	—	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Finding a Balanced Degree of Automation for Summary Evaluation	10.18653/v1/2021.emnlp-mai	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Reasoning Models Don't Always Say What They Think	10.48550/arxiv.2505.05410	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Qwen3 Technical Report	—	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	A General Language Assistant as a Laboratory for Alignment	—	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	RankGen: Improving Text Generation with Large Ranking Models	10.48550/arxiv.2205.09726	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Training a Helpful and Harmless Assistant with Reinforcement Learning from Human	10.48550/arxiv.2204.05862	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Training language models to follow instructions with human feedback	10.52202/068431-2011	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Does Chain-of-Thought Reasoning Really Reduce Harmfulness from Jailbreaking?	10.48550/arxiv.2505.17650	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Out of One, Many: Using Language Models to Simulate Human Samples	10.1017/pan.2023.2	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Weakest Link in the Chain: Security Vulnerabilities in Advanced Reasoning Models	10.48550/arxiv.2506.13726	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Localizing Persona Representations in LLMs	10.1609/aies.v8i1.36577	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Who is GPT-3? An exploration of personality, values and demographics	10.48550/arxiv.2209.14338	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	GPT-4 Technical Report	—	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Model Criticism for Long-Form Text Generation	10.48550/arxiv.2210.08444	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Enhancing Jailbreak Attacks on LLMs via Persona Prompts	10.48550/arxiv.2507.22171	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	LLMs Encode Harmfulness and Refusal Separately	10.48550/arxiv.2507.11878	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Persona Vectors: Monitoring and Controlling Character Traits in Language Models	10.48550/arxiv.2507.21509	active	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Revisiting the Gold Standard: Grounding Summarization Evaluation with Robust Hum	10.48550/arxiv.2212.07981	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Generative Agents: Interactive Simulacra of Human Behavior	10.1145/3586183.3606763	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Toxicity in ChatGPT: Analyzing Persona-assigned Language Models	10.48550/arxiv.2304.05335	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Whose Opinions Do Language Models Reflect?	—	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	A psychometric framework for evaluating and shaping personality traits in large	10.1038/s42256-025-01115-6	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Persistent Instability in LLM's Personality Measurements: Effects of Scale, Reas	10.48550/arxiv.2508.04826	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Benchmarking and Understanding Safety Risks in AI Character Platforms	10.48550/arxiv.2512.01247	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	From Pretraining Data to Language Models to Downstream Tasks: Tracking the Trail	10.48550/arxiv.2305.08283	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	PersonaLLM: Investigating the Ability of Large Language Models to Express Person	10.18653/v1/2024.findings-	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text	10.48550/arxiv.2305.14251	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Linear Personality Probing and Steering in LLMs: A Big Five Study	10.48550/arxiv.2512.17639	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	The Geometry of Persona: Disentangling Personality from Reasoning in Large Langu	10.48550/arxiv.2512.07092	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Evading LLMs’ Safety Boundary with Adaptive Role-Play Jailbreaking	10.3390/electronics1424480	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Marked Personas: Using Natural Language Prompts to Measure Stereotypes in Langua	10.48550/arxiv.2305.18189	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Aligning Language Models to User Opinions	10.48550/arxiv.2305.14929	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	In-Context Impersonation Reveals Large Language Models' Strengths and Biases	10.48550/arxiv.2305.14930	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Breaking Minds, Breaking Systems: Jailbreaking Large Language Models via Human-l	10.48550/arxiv.2512.18244	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	The Assistant Axis: Situating and Stabilizing the Default Persona of Language Mo	10.48550/arxiv.2601.10387	active	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	The Persona Paradox: Medical Personas as Behavioral Priors in Clinical Language	10.48550/arxiv.2601.05376	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Stay in Character, Stay Safe: Dual-Cycle Adversarial Self-Evolution for Safety R	10.48550/arxiv.2602.13234	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Do Personality Traits Interfere? Geometric Limitations of Steering in Large Lang	10.48550/arxiv.2602.15847	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Better Zero-Shot Reasoning with Role-Play Prompting	10.48550/arxiv.2308.07702	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Evaluating the Moral Beliefs Encoded in LLMs	10.48550/arxiv.2307.14324	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Steering Externalities: Benign Activation Steering Unintentionally Increases Jai	10.48550/arxiv.2602.04896	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Towards Measuring the Representation of Subjective Global Opinions in Language M	—	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Character-LLM: A Trainable Agent for Role-Playing	10.48550/arxiv.2310.10158	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Ethical Reasoning over Moral Alignment: A Case and Framework for In-Context Ethi	10.48550/arxiv.2310.07251	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Are Personalized Stochastic Parrots More Dangerous? Evaluating Persona Biases in	10.48550/arxiv.2310.05280	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Reasoning Theater: Disentangling Model Beliefs from Chain-of-Thought	10.48550/arxiv.2603.05488	active	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Steer2Adapt: Dynamically Composing Steering Vectors Elicits Efficient Adaptation	10.48550/arxiv.2602.07276	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Expert Personas Improve LLM Alignment but Damage Accuracy: Bootstrapping Intent-	10.48550/arxiv.2603.18507	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	InCharacter: Evaluating Personality Fidelity in Role-Playing Agents through Psyc	10.18653/v1/2024.acl-long.	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	Analysing the Safety Pitfalls of Steering Vectors	10.48550/arxiv.2603.24543	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Bias Runs Deep: Implicit Reasoning Biases in Persona-Assigned LLMs	10.48550/arxiv.2311.04892	referenced-only	s2-reference-walk
Persona Non Grata: Single-Method Safety Evaluation Is Incomp	What Drives Representation Steering? A Mechanistic Case Study on Steering Refusa	10.48550/arxiv.2604.08524	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	The Generative AI Paradox: "What It Can Create, It May Not Understand"	10.48550/arxiv.2311.00059	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Assessing LLMs for Moral Value Pluralism	10.48550/arxiv.2312.10075	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	You don’t need a personality test to know these models are unreliable: Assessing	10.48550/arxiv.2311.09718	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Quantifying the Persona Effect in LLM Simulations	10.48550/arxiv.2402.10811	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Scaling Synthetic Data Creation with 1,000,000,000 Personas	10.48550/arxiv.2406.20094	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Evaluating Large Language Model Biases in Persona-Steered Generation	10.48550/arxiv.2405.20253	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Political Compass or Spinning Arrow? Towards More Meaningful Evaluations for Val	10.48550/arxiv.2402.16786	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	The llama 3 herd of models	10.48550/arxiv.2407.21783	referenced-only	s2-reference-walk
Spotting Out-of-Character Behavior: Atomic-Level Evaluation	Different Bias Under Different Criteria: Assessing Bias in LLMs with a Fact-Base	10.48550/arxiv.2411.17338	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	2024. Ask LLMs directly, “what	—	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	2022. Discovering Language Model Behav-iors with Model-Written Evaluations	—	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	To-wards formal definitions of blameworthiness, intention, and moral responsibil	—	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	2023. Do personality tests generalize to large language models? In Socially Resp	—	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	GitHub Copilot: your AI pair programmer	—	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	Language Models are Unsupervised Multitask Learners	—	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	The instrumental action should be ‘slightly unethical’	—	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	A modern approach to probability theory	10.2307/2670021	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	Agency	10.1080/00091380903479323	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	2024. Stark - Animation and AI (preprint)	—	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	Behaviorism	10.4135/9781506326139.n78	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	Belief	10.1177/0194599812439519	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	Leap-Of-Thought: Teaching Pre-Trained Models to Systematically Reason Over Impli	—	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	Fundamentals of Probability and Stochastic Processes with Applications to Commun	10.1007/978-3-319-68075-0	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	Definitions of intent suitable for algorithms	10.1007/s10506-022-09322-x	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	How to ascribe beliefs to animals	10.1111/mila.12302	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	TruthfulQA: Measuring How Models Mimic Human Falsehoods	10.18653/v1/2022.acl-long.	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	Do Language Models Have Beliefs? Methods for Detecting, Updating, and Visualizin	—	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	Training a Helpful and Harmless Assistant with Reinforcement Learning from Human	10.48550/arxiv.2204.05862	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	Locating and editing factual associations in GPT	10.52202/068431-1262	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	Discovering Latent Knowledge in Language Models Without Supervision	10.48550/arxiv.2212.03827	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	Goal Misgeneralization: Why Correct Specifications Aren't Enough For Correct Goa	10.48550/arxiv.2210.01790	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	Honesty Is the Best Policy: Defining and Mitigating AI Deception	10.48550/arxiv.2312.01350	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	Still no lie detector for language models: probing empirical and conceptual road	10.1007/s11098-023-02094-3	referenced-only	s2-reference-walk
Evaluating Language Model Character Traits	The Reasons that Agents Act: Intention and Instrumental Goals	10.48550/arxiv.2402.07221	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	19: The average hidden state norms, which largely vary, especially in earlier ch	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	RQ1 We trace persona-vector development through pretraining and find that they f	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Steered answer with score 38.5 and coherence = 64.4:	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Four roots of evil.	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Gemini 2.5 Team	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Measuring and understanding social sycophancy in	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Less is more: Recursive reasoning with tiny networks (2025)	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	What We Talk to When We Talk to Language Models	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Rethinking pretraining: Data and architecture	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	What OpenAI Did When ChatGPT Users Lost Touch With Reality	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	A longlist of theories of impact for interpretability	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	System card: Claude Mythos preview	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	GPT-5.5 system card. System card	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	The persona selection model: Why AI assistants might behave like humans	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Transformers: State-of-the-Art Natural Language Processing	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	The Hitchhiker’s Guide to Testing Statistical Significance in Natural Language P	10.18653/v1/p18-1128	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Multidimensional scaling: I. Theory and method	10.1007/bf02288916	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	ChatGPT’s inconsistent moral advice influences users’ judgment	10.1038/s41598-023-31341-0	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	A Survey on Evaluation of Large Language Models	10.1145/3641289	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Efficient Memory Management for Large Language Model Serving with PagedAttention	10.1145/3600006.3613165	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	The linear representation hypothesis and the geometry of large language models	10.48550/arxiv.2311.03658	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Large Language Models can Strategically Deceive their Users when Put Under Press	—	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Grokked Transformers are Implicit Reasoners: A Mechanistic Journey to the Edge o	10.48550/arxiv.2405.15071	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	A Mechanistic Understanding of Alignment Algorithms: A Case Study on DPO and Tox	10.48550/arxiv.2401.01967	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	GIFT-SW: Gaussian noise Injected Fine-Tuning of Salient Weights for LLMs	10.48550/arxiv.2408.15300	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Designing a Dashboard for Transparency and Control of Conversational AI	10.48550/arxiv.2406.07882	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	AI sandbagging: Language models can strategically underperform on evaluations	10.48550/arxiv.2406.07358	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Open Problems in Mechanistic Interpretability	10.48550/arxiv.2501.16496	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Cases of Using ChatGPT as a Mental Health and Psychological Support Tool	10.1080/15398285.2024.2442	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Safety Pretraining: Toward the Next Generation of Safe AI	10.48550/arxiv.2504.16980	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Persona Features Control Emergent Misalignment	10.48550/arxiv.2506.19823	active	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Weird Generalization and Inductive Backdoors: New Ways to Corrupt LLMs	10.48550/arxiv.2512.09742	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Seeking Emotional and Mental Health Support From Generative AI: Mixed-Methods St	10.2196/77951	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Persona Vectors: Monitoring and Controlling Character Traits in Language Models	10.48550/arxiv.2507.21509	active	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	From Model Training to Model Raising	10.1145/3748645	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	The Assistant Axis: Situating and Stabilizing the Default Persona of Language Mo	10.48550/arxiv.2601.10387	active	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation E	10.48550/arxiv.2512.15674	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	PERSONA: Dynamic and Compositional Inference-Time Personality Control via Activa	10.48550/arxiv.2602.15669	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Persona Vectors in Games: Measuring and Steering Strategies via Activation Vecto	10.48550/arxiv.2603.21398	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	International AI Safety Report 2026	10.48550/arxiv.2602.21012	referenced-only	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	Where is the Mind? Persona Vectors and LLM Individuation	10.48550/arxiv.2604.17031	active	s2-reference-walk
Tracing Persona Vectors Through LLM Pretraining	DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	System card: Claude opus 4 & claude sonnet 4. System card	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Theia Vogel	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Evaluating feature steering: A case study in mitigating social biases	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Persona features control	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	The new ai-powered bing is threatening users. that’s no laughing matter	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	The mad dreams of an electric mind	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Amplify-instruct: Synthetically generated diverse multi-turn conversations for e	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Lighteval: A lightweight framework for llm evaluation	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	level? That just creates dependency and unequal power dynamics that favor neithe	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Microsoft artificial intelligence ai chatbot “sydney” rattled users before chatg	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Claude’s character	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	The rating of chessplayers, past and present	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	An Adversarial Winograd Schema Challenge at Scale	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Modifying llm beliefs with synthetic document finetuning	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	OpenRLHF: An Easy-to-use, Scalable and High-performance RLHF Framework	10.48550/arxiv.2405.11143	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	X removes posts by musk chatbot grok after antisemitism complaints	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	On the Humanity of Conversational AI: Evaluating the Psychological Portrayal of	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Expanding on what we missed with sycophancy	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	OpenAI	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	A Persona-Based Neural Conversation Model	10.18653/v1/p16-1094	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Personalizing Dialogue Agents: I have a dog, do you have pets too?	10.18653/v1/p18-1205	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Deep Reinforcement Learning from Human Preferences	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Measuring Massive Multitask Language Understanding	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Talk to me: Exploring user interactions with the Amazon Alexa	10.1177/0961000618759414	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	HellaSwag: Can a Machine Really Finish Your Sentence?	10.18653/v1/p19-1472	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	A General Language Assistant as a Laboratory for Alignment	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Transformers: State-of-the-Art Natural Language Processing	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	LoRA: Low-rank adaptation of large language models	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	TruthfulQA: Measuring How Models Mimic Human Falsehoods	10.18653/v1/2022.acl-long.	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	2026. On the non-identifiability of steering vectors in large language models	—	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	2024c. PATIENT-Ψ : Using large language models to simulate patients for training	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	LaMP: When Large Language Models Meet Personalization	10.48550/arxiv.2304.11406	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	CAT'S THEORY: Empirical Validation and Architectural Applications Cross-Architec	10.48550/arxiv.2212.08073	active	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	1 others	—	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Check whether the prompt still works if shown alone on a blank page	—	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	2025. A survey on hallucination in large language models: Principles, taxonomy,	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Training a Helpful and Harmless Assistant with Reinforcement Learning from Human	10.48550/arxiv.2204.05862	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	2023. Do personality tests generalize to large language models? In Socially Resp	—	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	2024. Ask LLMs directly, “what	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	LIMA: Less Is More for Alignment	10.52202/075280-2400	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Direct preference optimization: Your language model is secretly a reward model	10.52202/075280-2338	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Role play with large language models	10.1038/s41586-023-06647-8	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Development and validation of the Working Alliance Inventory.	10.1037/0022-0167.36.2.223	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Validation of the five-factor model of personality across instruments and observ	10.1037/0022-3514.52.1.81	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	The generalizability of the psychoanalytic concept of the working alliance.	10.1037/h0085885	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Quantifying the Persona Effect in LLM Simulations	10.48550/arxiv.2402.10811	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Specific versus General Principles for Constitutional AI	10.48550/arxiv.2310.13798	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Steering language models with activation engineering	—	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	The Big Five Trait taxonomy: History, measurement, and theoretical perspectives.	—	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	NEO PI-R Professional Manual	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Iterative Reasoning Preference Optimization	10.48550/arxiv.2404.19733	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Do LLMs Have Distinct and Consistent Personality? TRAIT: Personality Testset des	10.48550/arxiv.2406.14703	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	WildChat: 1M ChatGPT Interaction Logs in the Wild	10.48550/arxiv.2405.01470	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	An alternative "description of personality": the big-five factor structure.	10.1037/0022-3514.59.6.121	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Negotiating the Therapeutic Alliance: A Relational Treatment Guide.	—	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	OpenAI	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	PersonalLLM: Tailoring LLMs to Individual Preferences	10.48550/arxiv.2409.20296	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Self-Directed Synthetic Dialogues and Revisions Technical Report	10.48550/arxiv.2407.18421	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Can Role Vectors Affect LLM Behaviour?	10.18653/v1/2025.findings-	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Empathy	10.7551/mitpress/12282.003	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Personality Alignment of Large Language Models	10.48550/arxiv.2408.11779	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Human agency in social cognitive theory.	10.1037/0003-066x.44.9.117	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	What Makes for a Good Teacher and Who Can Tell? Working Paper 30.	10.1037/e722052011-001	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Misuses and misunderstandings of boundary theory in clinical and regulatory sett	10.1176/ajp.155.3.409	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Boundary issues and multiple relationships: Fantasy and reality.	10.1037/0735-7028.38.4.401	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	When the patient is a therapist: special challenges in the psychoanalysis of men	—	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Personalization of Large Language Models: A Survey	10.48550/arxiv.2411.00027	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Designing AI Personalities: Enhancing Human-Agent Interaction Through Thoughtful	10.1145/3701571.3701608	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Alliance in individual psychotherapy.	10.1037/a0022186	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Visible learning: a synthesis of over 800 meta‐analyses relating to achievement	10.1080/01443410903415150	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory	10.48550/arxiv.2412.13663	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Can LLM "Self-report"?: Evaluating the Validity of Self-report Scales in Measuri	10.48550/arxiv.2412.00207	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	TÜLU 3: Pushing Frontiers in Open Language Model Post-Training	10.48550/arxiv.2411.15124	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Alliance-focused training.	10.1037/a0037596	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Qualities of Effective Teachers, 3rd Edition	—	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Detecting alliance ruptures: the effects of the therapist’s experience, attachme	10.4081/ripppo.2019.325	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	The therapeutic relationship in cognitive–behavioral therapy: Essential features	10.1037/pri0000088	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	A Computational Approach to Understanding Empathy Expressed in Text-Based Mental	10.18653/v1/2020.emnlp-mai	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Linear Representations of Political Perspective Emerge in Large Language Models	10.48550/arxiv.2503.02080	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Training large language models on narrow tasks can lead to broad misalignment	10.1038/s41586-025-09937-5	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	A Survey of Personalized Large Language Models: Progress and Future Directions	10.48550/arxiv.2502.11528	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Will AI Tell Lies to Save Sick Children? Litmus-Testing AI Values Prioritization	10.48550/arxiv.2505.14633	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Discovering language model behaviors with model-written evaluations	10.48550/arxiv.2212.09251	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Probing then Editing Response Personality of Large Language Models	10.48550/arxiv.2504.10227	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Alliance ruptures in cognitive-behavioral therapy: A cognitive conceptualization	10.1002/jclp.23116	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Reinforcement Learning from Human Feedback	10.48550/arxiv.2504.12501	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	TruthfulQA: Measuring How Models Mimic Human Falsehoods	10.18653/v1/2022.acl-long.	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	The Personality Illusion: Revealing Dissociation Between Self-Reports & Behavior	10.48550/arxiv.2509.03730	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Persona Vectors: Monitoring and Controlling Character Traits in Language Models	10.48550/arxiv.2507.21509	active	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Inference-time intervention: eliciting truthful answers from a language model	10.52202/075280-1797	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	ChatGPT for good? On opportunities and challenges of large language models for e	10.1016/j.lindif.2023.1022	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Subliminal Learning: Language models transmit behavioral traits via hidden signa	10.48550/arxiv.2507.14805	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Steering language models with activation engineering	—	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	The Rise and Potential of Large Language Model Based Agents: A Survey	10.48550/arxiv.2309.07864	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Another Turn, Better Output? A Turn-Wise Analysis of Iterative LLM Prompting	10.48550/arxiv.2509.06770	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Character-LLM: A Trainable Agent for Role-Playing	10.48550/arxiv.2310.10158	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Generative Value Conflicts Reveal LLM Priorities	10.48550/arxiv.2509.25369	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	A survey on large language model based autonomous agents	10.1007/s11704-024-40231-1	referenced-only	s2-reference-walk
Open Character Training: Shaping the Persona of AI Assistant	Stress-Testing Model Specs Reveals Character Differences among Language Models	10.48550/arxiv.2510.07686	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Steering Llama 2 via Contrastive Activation Addition	10.48550/arxiv.2312.06681	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Towards Understanding Sycophancy in Language Models	10.48550/arxiv.2310.13548	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Two Tales of Persona in LLMs: A Survey of Role-Playing and Personalization	10.48550/arxiv.2406.01171	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Crafting Customisable Characters with LLMs: A Persona-Driven Role-Playing Agent	10.18653/v1/2025.findings-	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Persona Vectors: Monitoring and Controlling Character Traits in Language Models	10.48550/arxiv.2507.21509	active	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Intuition, empathy, and intellectual humility in psychotherapy. A philosophical	10.3389/fpsyg.2025.1590481	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Deliberative Alignment: Reasoning Enables Safer Language Models	10.48550/arxiv.2412.16339	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Personality Vector: Modulating Personality of Large Language Models by Model Mer	10.48550/arxiv.2509.19727	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Evaluating the Clinical Safety of LLMs in Response to High-Risk Mental Health Di	10.48550/arxiv.2509.08839	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Bridging Mechanistic Interpretability and Prompt Engineering with Gradient Ascen	10.48550/arxiv.2601.02896	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	BILLY: Steering Large Language Models via Merging Persona Vectors for Creative G	10.48550/arxiv.2510.10157	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	The Assistant Axis: Situating and Stabilizing the Default Persona of Language Mo	10.48550/arxiv.2601.10387	active	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Letting Tutor Personas "Speak Up" for LLMs: Learning Steering Vectors from Dialo	10.48550/arxiv.2602.07639	referenced-only	s2-reference-walk
What Models Express, Suppress, and Resist: Auditing Open-Wei	Sycophancy Hides Linearly in the Attention Heads	10.48550/arxiv.2601.16644	referenced-only	s2-reference-walk