mlsanityData to 14 Oct 2026

Bayesian Teaching Enables Probabilistic Reasoning in Large Language Models

arXiv 2503.17523 ↗First posted 21 Mar 2025cs.CL cs.AIcited by 37

On the map: LLM reasoning → Test-time compute and efficient thinking

Abstract (arXiv)

Large language models (LLMs) are increasingly used as agents that interact with users and with the world. To do so successfully, LLMs must construct representations of the world and form probabilistic beliefs about them. To provide personalized recommendations, for example, the LLM needs to infer a user's preferences from their behavior over multiple interactions. The Bayesian inference framework lays out the optimal way for an agent to update its beliefs as it receives new information. We first show that LLMs fall far short of the standard defined by the Bayesian framework. We then show that by teaching LLMs to mimic the predictions of the normative Bayesian model, we can dramatically improve their ability to update their beliefs; this ability generalizes to new tasks. We conclude that LLMs can effectively learn reasoning skills from examples and generalize those skills to new domains.

Key facts. Cited by 37 papers in the map (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 65 of its 103 references that are in the map. First posted on arXiv 21 Mar 2025.

Builds on

65 of 103 references.

Closest ancestors · 3

Same direction, newest first

  1. 2024What Are the Odds? Language Models Are Capable of Probabilistic Reasoning18 Jun
  2. 2024Doing Experiments and Revising Rules with Natural Language and Probabilistic Reasoning8 Feb
  3. 2023Tree of Thoughts: Deliberate Problem Solving with Large Language Models17 May

Other references · 62

  1. 2017Attention Is All You Needcited by 60,361
  2. 2023GPT-4 Technical Reportcited by 18,396
  3. 2022Training language models to follow instructions with human feedbackcited by 15,971
  4. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 13,028
  5. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  6. 2015Distilling the Knowledge in a Neural Networkcited by 10,324
  7. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  8. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  9. 2024Qwen2.5 Technical Reportcited by 4,942
  10. 2022Large Language Models are Zero-Shot Reasonerscited by 4,374
  11. 2017Deep reinforcement learning from human preferencescited by 3,912
  12. 2021Finetuned Language Models Are Zero-Shot Learnerscited by 3,780
38 more
  1. 2024Gemini 1.5: Unlocking multimodal understanding across millions of tokens of contextcited by 3,224
  2. 2020Learning to summarize from human feedbackcited by 2,588
  3. 2022Emergent Abilities of Large Language Modelscited by 2,512
  4. 2024Gemma 2: Improving Open Language Models at a Practical Sizecited by 1,998
  5. 2018World Modelscited by 1,513
  6. 2021Multitask Prompted Training Enables Zero-Shot Task Generalizationcited by 1,470
  7. 2022WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agentscited by 1,251
  8. 2024Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preferencecited by 1,184
  9. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 smallcited by 1,165
  10. 2021Show Your Work: Scratchpads for Intermediate Computation with Language Modelscited by 886
  11. 2016Sequence-Level Knowledge Distillationcited by 801
  12. 2018Recurrent World Models Facilitate Policy Evolutioncited by 785
  13. 2022Transformers learn in-context by gradient descentcited by 732
  14. 2022What Can Transformers Learn In-Context? A Case Study of Simple Function Classescited by 715
  15. 2022What learning algorithm is in-context learning? Investigations with linear modelscited by 684
  16. 2023Are Emergent Abilities of Large Language Models a Mirage?cited by 523
  17. 2023Trained Transformers Learn Linear Models In-Contextcited by 337
  18. 2023Transformers as Statisticians: Provable In-Context Learning with In-Context Algorithm Selectioncited by 298
  19. 2023Transformers as Algorithms: Generalization and Stability in In-context Learningcited by 222
  20. 2023Implicit Chain of Thought Reasoning via Knowledge Distillationcited by 176
  21. 2024Distilling System 2 into System 1cited by 159
  22. 2023Symbolic Chain-of-Thought Distillation: Small Models Can Also "Think" Step-by-Stepcited by 148
  23. 2025Do LLMs Recognize Your Preferences? Evaluating Personalized Preference Following in LLMscited by 148
  24. 2021Causal Analysis of Syntactic Agreement Mechanisms in Neural Language Modelscited by 131
  25. 2023From Word Models to World Models: Translating from Natural Language to the Probabilistic Language of Thoughtcited by 119
  26. 2022Language Model Cascadescited by 107
  27. 2023SCOTT: Self-Consistent Chain-of-Thought Distillationcited by 105
  28. 2024STaR-GATE: Teaching Language Models to Ask Clarifying Questionscited by 104
  29. 2023Eliciting Human Preferences with Language Modelscited by 92
  30. 2023How Do Transformers Learn In-Context Beyond Simple Functions? A Case Study on Learning with Representationscited by 75
  31. 2023Strategic Reasoning with Language Modelscited by 64
  32. 2021Building and Evaluating Open-Domain Dialogue Corpora with Clarifying Questionscited by 63
  33. 2021Counterfactual Interventions Reveal the Causal Effect of Relative Clause Representations on Agreement Predictioncited by 59
  34. 2024Large Language Models Assume People are More Rational than We Really arecited by 54
  35. 2022Inferring Rewards from Language in Contextcited by 50
  36. 2023Prompting is not a substitute for probability measurements in large language modelscited by 44
  37. 2024The broader spectrum of in-context learningcited by 35
  38. 2024Bayesian Preference Elicitation with Language Modelscited by 34

50 most cited of 62 All 62 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 10; 2026: 27.

37 papers.

Sort:

By area:

  1. 2026Learning Personalized Agents from Human Feedbackcited by 18
  2. 2025BASIL: Bayesian Assessment of Sycophancy in LLMscited by 13
  3. 2026Agentic Forecasting using Sequential Bayesian Updating of Linguistic Beliefscited by 10
  4. 2025Always Tell Me The Odds: Fine-grained Conditional Probability Estimationcited by 9
  5. 2025Beyond Markovian: Reflective Exploration via Bayes-Adaptive RL for LLM Reasoningcited by 9
  6. 2025Shoot First, Ask Questions Later? Building Rational Agents that Explore and Act Like Peoplecited by 9
  7. 2025Reasoning Under Uncertainty: Exploring Probabilistic Reasoning Capabilities of LLMscited by 8
  8. 2025Language-Informed Synthesis of Rational Agent Models for Grounded Theory-of-Mind Reasoning On-The-Flycited by 6
  9. 2025Martingale Score: An Unsupervised Metric for Bayesian Rationality in LLM Reasoningcited by 5
  10. 2026LLMs are not (consistently) Bayesian: Quantifying internal (in)consistencies of LLMs' probabilistic beliefscited by 5
  11. 2025Timely Clinical Diagnosis through Active Test Selectioncited by 4
  12. 2025Scaling Law in LLM Simulated Personality: More Detailed and Realistic Persona Profile Is All You Needcited by 3
25 more
  1. 2026Position: agentic AI orchestration should be Bayes-consistentcited by 3
  2. 2026Revealed Rationality: Label-Free Evaluation and Regularization from Representation Theoremscited by 3
  3. 2026AdaptFuse: Training-Free Sequential Preference Inference via Externalized Bayesian Computationcited by 2
  4. 2025Emergent Bayesian Behaviour and Optimal Cue Combination in LLMscited by 1
  5. 2026Adaptive Domain Models: Bayesian Evolution, Warm Rotation, and Principled Training for Geometric and Neuromorphic AIcited by 1
  6. 2026Empirical Characterization of Inference-Time Elicited Probability Transformations in Large Language Modelscited by 1
  7. 2026Analytica: Soft Propositional Reasoning for Robust and Scalable LLM-Driven Analysiscited by 1
  8. 2026Hypothesis generation and updating in large language modelscited by 1
  9. 2026ANCHOR: Abductive Network Construction with Hierarchical Orchestration for Reliable Probability Inference in Large Language Modelscited by 1
  10. 2026When Does In-Context Search Help? A Sampling-Complexity Theory of Reflection-Driven Reasoningcited by 1
  11. 2026Hypotheses-Guided Self Distillation for Continual Personalizationcited by 1
  12. 2026Tutorial on Reasoning for IR & IR for Reasoningcited by 0
  13. 2026Subjective-Graph LLM Agents for Simulating Uncertainty in Classroom Social Perceptioncited by 0
  14. 2026Decision-Centric Design for LLM Systemscited by 0
  15. 2026Medical Model Synthesis Architectures: A Case Studycited by 0
  16. 2026Turning Intent into Specifications: A Benchmark and an Interactive User-Assistant Agentcited by 0
  17. 2026Teachy Mini: Development and Preliminary Evaluation of a Knowledge-Based Generative Social Robot for Higher Educationcited by 0
  18. 2026Shape Your Feed: An LLM-based Agentic System for Conversational Recommendationcited by 0
  19. 2026Proxy reliance in large language model decisions is uncalibrated to predictive evidencecited by 0
  20. 2026Semantic Bayesian World Modelscited by 0
  21. 2026A primer on evaluation methods for large language models in healthcarecited by 0
  22. 2026Small transformers track Bayesian evidence for latent common causes via a context-invariant mechanismcited by 0
  23. 2026Diagnosing and Improving Probabilistic Reasoning in Large Language Modelscited by 0
  24. 2026Bayesian Fine-tuning Yields Language Models that are as Bayesian as their Beliefs Allowcited by 0
  25. 2026Provenance-Aware Belief Updating in Human-AI Feedback Loopscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Large Language Models as Discounted Bayesian Filters
  2. 2024BayesAgent: Bayesian Agentic Reasoning Under Uncertainty via Verbalized Probabilistic Graphical Modeling
  3. 2025Do Chains-of-Thoughts of Large Language Models Suffer from Hallucinations, Cognitive Biases, or Phobias in Bayesian Reasoning?
  4. 2024Towards Logically Consistent Language Models via Probabilistic Reasoning
  5. 2025Exploring the Potential for Large Language Models to Demonstrate Rational Probabilistic Beliefs
  6. 2026Towards a Belief-Based World Model for LLM Agents
  7. 2025Bayesian Social Deduction with Graph-Informed Language Models
  8. 2024Dual Traits in Probabilistic Reasoning of Large Language Models