mlsanityData to 7 Oct 2026

Improving the Validity of Automatically Generated Feedback via Reinforcement Learning

arXiv 2403.01304 ↗First posted 2 Mar 2024cs.CLcited by 20

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

Abstract (arXiv)

Automatically generating feedback via large language models (LLMs) in intelligent tutoring systems and online learning platforms has the potential to improve the learning outcomes of many students. However, both feedback generation and evaluation are challenging: feedback content has to be valid especially in subjects like math, which requires models to understand the problem, the solution, and where the student's error lies. Feedback also has to be pedagogically valid to reflect effective tutoring strategies, such as explaining possible misconceptions and encouraging the student, among other desirable features. In this work, we address both problems of automatically generating and evaluating feedback while considering both correctness and alignment. First, we propose a rubric for evaluating math feedback and show that GPT-4 is able to effectively use it to annotate human-written and LLM-generated feedback. Second, we propose a framework for feedback generation that optimizes both correctness and alignment using reinforcement learning (RL). Specifically, we use GPT-4's annotations to create preferences over feedback pairs in an augmented dataset for training via direct preference optimization (DPO). We show that our methods significantly increase the correctness and alignment of generated feedback with Llama 2, an open-source LLM, qualitatively analyze our generation and evaluation systems using case studies, and outline several areas for future work.

Key facts. Cited by 20 papers in the map (data through 7 Oct 2026). Placed in Process reward models for reasoning, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 25 of its 41 references that are in the map. First posted on arXiv 2 Mar 2024.

Builds on

25 of 41 references.

References · 25

  1. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,084
  2. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  3. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,049
  4. 2021Evaluating Large Language Models Trained on Codecited by 8,053
  5. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  6. 2019BERTScore: Evaluating Text Generation with BERTcited by 5,462
  7. 2022Large Language Models are Zero-Shot Reasonerscited by 4,323
  8. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,064
  9. 2019HuggingFace's Transformers: State-of-the-art Natural Language Processingcited by 2,201
  10. 2019Fine-Tuning Language Models from Human Preferencescited by 2,124
  11. 2022Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Taskscited by 1,057
  12. 2023Can Large Language Models Be an Alternative to Human Evaluations?cited by 729
13 more
  1. 2020Contrastive Learning with Hard Negative Samplescited by 619
  2. 2023Large Language Models Are State-of-the-Art Evaluators of Translation Qualitycited by 416
  3. 2023Automated Distractor and Feedback Generation for Math Multiple-choice Questions via In-context Learningcited by 24
  4. 2015Mathematical Language Processing: Automatic Grading and Feedback for Open Response Mathematical Questionscited by 19
  5. 2023Evaluating ChatGPT's Decimal Skills and Feedback Generation in a Digital Learning Gamecited by 19
  6. 2022Automatic Short Math Answer Grading via In-context Meta-learningcited by 18
  7. 2020Automated Personalized Feedback Improves Learning Gains in an Intelligent Tutoring Systemcited by 16
  8. 2023Novice Learner and Expert Tutor: Evaluating Math Reasoning Abilities of Large Language Models with Misconceptionscited by 15
  9. 2023Code Soliloquies for Accurate Calculations in Large Language Modelscited by 9
  10. 2024Using Large Language Models to Assess Tutors' Performance in Reacting to Students Making Math Errorscited by 9
  11. 2021Math Operation Embeddings for Open-ended Solution Analysis and Feedbackcited by 8
  12. 2021ALL-IN-ONE: Multi-Task Learning BERT models for Evaluating Peer Assessmentscited by 4
  13. 2023Algebra Error Classification with Large Language Modelscited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 5; 2025: 7; 2026: 8.

20 papers.

Sort:

By area:

  1. 2024Exploring Knowledge Tracing in Tutor-Student Dialogues using LLMscited by 44
  2. 2025Training LLM-based Tutors to Improve Student Learning Outcomes in Dialoguescited by 40
  3. 2024Math Multiple Choice Question Generation via Human-Large Language Model Collaborationcited by 9
  4. 2024Automatic Generation of Question Hints for Mathematics Problems using Large Language Models in Educational Technologycited by 9
  5. 2025From Text to Visuals: Using LLMs to Generate Math Diagrams with Vector Graphicscited by 5
  6. 2026FeedEval: Pedagogically Aligned Evaluation of LLM-Generated Essay Feedbackcited by 4
  7. 2024Revolutionizing Bridge Operation and Maintenance with LLM-based Agents: An Overview of Applications and Insightscited by 3
  8. 2026Letting Tutor Personas Speak Up for LLMs: Learning Steering Vectors from Dialogue via Preference Optimizationcited by 3
  9. 2024Interpreting Latent Student Knowledge Representations in Programming Assignmentscited by 2
  10. 2026From Untamed Black Box to Interpretable Pedagogical Orchestration: The Ensemble of Specialized LLMs Architecture for Adaptive Tutoringcited by 2
  11. 2025Concept-based Rubrics Improve LLM Formative Assessment and Data Synthesiscited by 1
  12. 2025Exploring LLM-Generated Feedback for Economics Essays: How Teaching Assistants Evaluate and Envision Its Usecited by 1
8 more
  1. 2025Narrowing the Gap: Supervised Fine-Tuning of Open-Source LLMs as a Viable Alternative to Proprietary Models for Pedagogical Toolscited by 1
  2. 2025Cultivating Helpful, Personalized, and Creative AI Tutors: A Framework for Pedagogical Alignment using Reinforcement Learningcited by 1
  3. 2025Comparing RAG and GraphRAG for Page-Level Retrieval Question Answering on a Math Textbookcited by 1
  4. 2026REFINE: Real-world Exploration of Interactive Feedback and Student Behaviourcited by 1
  5. 2026Teaching Language Models How to Code Like Learners: Conversational Serialization for Student Simulationcited by 1
  6. 2026A Multi-Agent Approach to Validate and Refine LLM-Generated Personalized Math Problemscited by 0
  7. 2026Listen, Correct, and Feed Back: Spoken Pedagogical Feedback Generationcited by 0
  8. 2026Gumbel Machine: Counterfactual Student Writing Generation via Gumbel Noise Steeringcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025MathEDU: Feedback Generation on Problem-Solving Processes for Mathematical Learning Support
  2. 2025Teaching LLMs for Step-Level Automatic Math Correction via Reinforcement Learning
  3. 2026Towards Pedagogically Aligned LLM Tutors for Math Mistake Remediation
  4. 2026Automated Feedback Generation for Undergraduate Mathematics: Development and Evaluation of an AI Teaching Assistant
  5. 2025Simulating LLM-to-LLM Tutoring for Multilingual Math Feedback
  6. 2025Automatic Large Language Models Creation of Interactive Learning Lessons
  7. 2025Efficient RL for optimizing conversation level outcomes with an LLM-based tutor
  8. 2026A Classroom Study of LLM-Generated Feedback Intervention in Introductory Programming