mlsanityData to 7 Oct 2026

How Controllable Are Large Language Models? A Unified Evaluation across Behavioral Granularities

arXiv 2603.02578 ↗First posted 3 Mar 2026cs.CL cs.AI cs.HCcited by 3

Not on the map. Why?

Abstract (arXiv)

Large Language Models (LLMs) are increasingly deployed in socially sensitive domains, yet their unpredictable behaviors, ranging from misaligned intent to inconsistent personality, pose significant risks. We introduce SteerEval, a hierarchical benchmark for evaluating LLM controllability across three domains: language features, sentiment, and personality. Each domain is structured into three specification levels: L1 (what to express), L2 (how to express), and L3 (how to instantiate), connecting high-level behavioral intent to concrete textual output. Using SteerEval, we systematically evaluate contemporary steering methods, revealing that control often degrades at finer-grained levels. Our benchmark offers a principled and interpretable framework for safe and controllable LLM behavior, serving as a foundation for future research.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Builds on 30 of its 130 references that are in the map. First posted on arXiv 3 Mar 2026.

Builds on

30 of 130 references.

References · 30

  1. 2020Language Models are Few-Shot Learnerscited by 32,199
  2. 2024The Llama 3 Herd of Modelscited by 14,192
  3. 2022Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?cited by 1,413
  4. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,359
  5. 2023Inference-Time Intervention: Eliciting Truthful Answers from a Language Modelcited by 1,259
  6. 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 1,239
  7. 2023Steering Llama 2 via Contrastive Activation Additioncited by 1,009
  8. 2024Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluatorscited by 865
  9. 2023The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasetscited by 673
  10. 2025Persona Vectors: Monitoring and Controlling Character Traits in Language Modelscited by 303
  11. 2025AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoderscited by 199
  12. 2025Open Problems in Mechanistic Interpretabilitycited by 173
18 more
  1. 2023Linear Representations of Sentiment in Large Language Modelscited by 157
  2. 2024Personalized Steering of Large Language Models: Versatile Steering Vectors Through Bi-directional Preference Optimizationcited by 112
  3. 2025A Survey on Sparse Autoencoders: Interpreting the Internal Mechanisms of Large Language Modelscited by 73
  4. 2025Steering Large Language Model Activations in Sparse Spacescited by 58
  5. 2025AlphaSteer: Learning Refusal Steering with Principled Null-Space Constraintcited by 45
  6. 2025A Unified Understanding and Evaluation of Steering Methodscited by 40
  7. 2024Towards Reliable Evaluation of Behavior Steering Interventions in LLMscited by 33
  8. 2025Personalized Text Generation with Contrastive Activation Steeringcited by 30
  9. 2025Mechanistic Interpretability of Emotion Inference in Large Language Modelscited by 24
  10. 2025Belief Dynamics Reveal the Dual Nature of In-Context Learning and Activation Steeringcited by 24
  11. 2025Beyond Prompt Engineering: Robust Behavior Control in LLMs via Steering Target Atomscited by 23
  12. 2025HyperSteer: Activation Steering at Scale with Hypernetworkscited by 20
  13. 2025Representation Engineering for Large-Language Models: Survey and Research Challengescited by 18
  14. 2025Improved Representation Steering for Language Modelscited by 17
  15. 2023ChatGPT is a Remarkable Tool -- For Expertscited by 16
  16. 2026The Shape of Beliefs: Geometry, Dynamics, and Interventions along Representation Manifolds of Language Models' Posteriorscited by 15
  17. 2026The Information Geometry of Softmax: Probing and Steeringcited by 9
  18. 2026Why Steering Works: Toward a Unified View of Language Model Parameter Dynamicscited by 8

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026What Does Activation Steering Control? Attribution Across Answer Encodings and Output-Sensitive Subspacescited by 1
  2. 2026RepBench: Compiling Benchmarks into Capability Representations for Large Language Modelscited by 0
  3. 2026When Do Model Internals Help? Exploring the Role of Representation Engineering in LLM Safetycited by 0