mlsanityData to 14 Oct 2026

PIVOT: Preference-based Intervention Vectors for Pedagogical Tutor Steering

arXiv 2608.07509 ↗First posted 29 Jun 2026cs.HC cs.AIcited by 0

On the map: LLM alignment / RLHF / preference optimisation → Process reward models for reasoning

Abstract (arXiv)

LLMs are increasingly used for conversational tutoring, but effective tutoring requires more than correct answers. Tutors must choose when to scaffold reasoning, hint, give feedback, explain, or invite reflection. Existing prompting and training methods improve pedagogical alignment, but lack reliable inference-time control over pedagogical strategies. We introduce PIVOT, an activation-steering framework that learns preference-based intervention vectors online for frozen LLM tutors. PIVOT uses a seven-category tutor-move taxonomy and a generate-label-optimise loop, where a human-validated LLM judge identifies target and confusable non-target moves to construct preference pairs for multi-layer residual-stream steering. Across held-out and out-of-domain tutoring data, PIVOT controls tutor moves while preserving relevance and fluency, and its directions can be scaled, transferred, and composed at inference time. In a user study with 30 teachers, 73.3% of participants preferred steered conversations over neutral baseline interactions using the same prompt, and rated the controls as clear, usable, and pedagogically meaningful.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Process reward models for reasoning, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 26 of its 66 references that are in the map. First posted on arXiv 29 Jun 2026.

Builds on

26 of 66 references.

Closest ancestors · 1

Same direction, newest first

  1. 2026Letting Tutor Personas Speak Up for LLMs: Learning Steering Vectors from Dialogue via Preference Optimization7 Feb

Other references · 25

  1. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  2. 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 1,336
  3. 2023Inference-Time Intervention: Eliciting Truthful Answers from a Language Modelcited by 1,302
  4. 2023Steering Llama 2 via Contrastive Activation Additioncited by 1,085
  5. 2023Steering Language Models With Activation Engineeringcited by 922
  6. 2023MathDial: A Dialogue Tutoring Dataset with Rich Pedagogical Properties Grounded in Math Reasoning Problemscited by 131
  7. 2024Personalized Steering of Large Language Models: Versatile Steering Vectors Through Bi-directional Preference Optimizationcited by 118
  8. 2022The AI Teacher Test: Measuring the Pedagogical Ability of Blender and GPT-3 in Educational Dialoguescited by 79
  9. 2024Unifying AI Tutor Evaluation: An Evaluation Taxonomy for Pedagogical Ability Assessment of LLM-Powered AI Tutorscited by 61
  10. 2024Towards Responsible Development of Generative AI for Education: An Evaluation-Driven Approachcited by 54
  11. 2023CLASS: A Design Framework for building Intelligent Tutoring Systems based on Learning Science principlescited by 48
  12. 2025Training LLM-based Tutors to Improve Student Learning Outcomes in Dialoguescited by 43
13 more
  1. 2024Empowering Personalized Learning through a Conversation-based Tutoring System with Student Modelingcited by 34
  2. 2023Opportunities and Challenges in Neural Dialog Tutoringcited by 33
  3. 2024AutoTutor meets Large Language Models: A Language Model Tutor with Rich Pedagogy and Guardrailscited by 31
  4. 2024Book2Dial: Generating Teacher-Student Interactions from Textbooks for Cost-Effective Development of Educational Chatbotscited by 25
  5. 2024Towards the Pedagogical Steering of Large Language Models for Tutoring: A Case Study with Modeling Productive Failurecited by 24
  6. 2025Beyond Final Answers: Evaluating Large Language Models for Math Tutoringcited by 23
  7. 2024Pedagogical Alignment of Large Language Modelscited by 19
  8. 2022Computationally Identifying Funneling and Focusing Questions in Classroom Discoursecited by 15
  9. 2025The StudyChat Dataset: Analyzing Student Dialogues With ChatGPT in an Artificial Intelligence Coursecited by 9
  10. 2026Evaluating Answer Leakage Robustness of LLM Tutors against Adversarial Student Attackscited by 7
  11. 2025SCRIBE: Structured Chain Reasoning for Interactive Behaviour Explanations using Tool Callingcited by 3
  12. 2026ConvoLearn: A Learning Sciences Grounded Dataset for Fine-Tuning Dialogic AI Tutorscited by 1
  13. 2026REFINE: Real-world Exploration of Interactive Feedback and Student Behaviourcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Towards Pedagogically Aligned LLM Tutors for Math Mistake Remediation
  2. 2025Toward Automated Qualitative Analysis: Leveraging Large Language Models for Tutoring Dialogue Evaluation
  3. 2026TACT: Taxonomy-Aligned Post-Training for Pedagogically Adaptive English Tutoring
  4. 2026The Tutoring Effectiveness Index: Predicting LLM Math Tutor Quality from Four Conversation Signals
  5. 2026Mitigating Scaffolding Collapse in Socratic Tutors via Representation Alignment
  6. 2025CoDAE: Adapting Large Language Models for Education via Chain-of-Thought Data Augmentation
  7. 2025From Problem-Solving to Teaching Problem-Solving: Aligning LLMs with Pedagogy using Reinforcement Learning
  8. 2026Rewarding How Models Think Pedagogically: Integrating Pedagogical Reasoning and Thinking Rewards for LLMs in Education