mlsanityData to 7 Oct 2026

Accumulating Context Changes the Beliefs of Language Models

arXiv 2511.01805 ↗First posted 3 Nov 2025cs.CL cs.AIcited by 25

Not on the map. Why?

Abstract (arXiv)

Language model (LM) assistants are increasingly used in applications such as brainstorming and research. Improvements in memory and context size have allowed these models to become more autonomous, which has also resulted in more text accumulation in their context windows without explicit user intervention. This comes with a latent risk: the belief profiles of models -- their understanding of the world as manifested in their responses or actions -- may silently change as context accumulates. This can lead to subtly inconsistent user experiences, or shifts in behavior that deviate from the original alignment of the models. In this paper, we explore how accumulating context by engaging in interactions and processing text -- talking and reading -- can change the beliefs of language models, as manifested in their responses and behaviors. Our results reveal that models' belief profiles are highly malleable: GPT-5 exhibits a 54.7% shift in its stated beliefs after 10 rounds of discussion about moral dilemmas and queries about safety, while Grok 4 shows a 27.2% shift on political issues after reading texts from the opposing position. We also examine models' behavioral changes by designing tasks that require tool use, where each tool selection corresponds to an implicit belief. We find that these changes align with stated belief shifts, suggesting that belief shifts will be reflected in actual behavior in agentic systems. Our analysis exposes the hidden risk of belief shift as models undergo extended sessions of talking or reading, rendering their opinions and actions unreliable.

Key facts. Cited by 25 papers in the map (data through 7 Oct 2026). Builds on 63 of its 108 references that are in the map. First posted on arXiv 3 Nov 2025.

Builds on

63 of 108 references.

References · 63

  1. 2025Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiescited by 3,655
  2. 2023Generative Agents: Interactive Simulacra of Human Behaviorcited by 3,500
  3. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,383
  4. 2023Improving Factuality and Reasoning in Language Models through Multiagent Debatecited by 1,739
  5. 2023LongBench: A Bilingual, Multitask Benchmark for Long Context Understandingcited by 1,378
  6. 2023Jailbreaking Black Box Large Language Models in Twenty Queriescited by 1,326
  7. 2023AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Modelscited by 738
  8. 2023The Internal State of an LLM Knows When It's Lyingcited by 644
  9. 2023Tree of Attacks: Jailbreaking Black-Box LLMs Automaticallycited by 619
  10. 2025BrowseComp: A Simple Yet Challenging Benchmark for Browsing Agentscited by 616
  11. 2024How Johnny Can Persuade LLMs to Jailbreak Them: Rethinking Persuasion to Challenge AI Safety by Humanizing LLMscited by 533
  12. 2022Using cognitive psychology to understand GPT-3cited by 389
38 more
  1. 2025Reasoning Models Don't Always Say What They Thinkcited by 388
  2. 2024Can LLMs Generate Novel Research Ideas? A Large-Scale Human Study with 100+ NLP Researcherscited by 348
  3. 2024Agent Workflow Memorycited by 342
  4. 2025WebThinker: Empowering Large Reasoning Models with Deep Research Capabilitycited by 320
  5. 2024DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Headscited by 283
  6. 2024Debating with More Persuasive LLMs Leads to More Truthful Answerscited by 255
  7. 2024MT-Bench-101: A Fine-Grained Benchmark for Evaluating Large Language Models in Multi-Turn Dialoguescited by 247
  8. 2025MemAgent: Reshaping Long-Context LLM with Multi-Conv RL-based Memory Agentcited by 235
  9. 2023Evaluating the Moral Beliefs Encoded in LLMscited by 211
  10. 2024WildTeaming at Scale: From In-the-Wild Jailbreaks to (Adversarially) Safer Language Modelscited by 206
  11. 2023Do LLMs exhibit human-like response biases? A case study in survey designcited by 176
  12. 2025SkillWeaver: Web Agents can Self-Improve by Discovering and Honing Skillscited by 174
  13. 2024LLM Defenses Are Not Robust to Multi-Turn Human Jailbreaks Yetcited by 145
  14. 2023ReviewerGPT? An Exploratory Study on Using Large Language Models for Paper Reviewingcited by 118
  15. 2025Agentic Reasoning: A Streamlined Framework for Enhancing LLM Reasoning with Agentic Toolscited by 106
  16. 2025Inducing Programmatic Skills for Agentic Taskscited by 103
  17. 2024MT-Eval: A Multi-Turn Capabilities Evaluation Benchmark for Large Language Modelscited by 78
  18. 2024Multi-Turn Context Jailbreak Attack on Large Language Models From First Principlescited by 54
  19. 2024MoralBench: Moral Evaluation of LLMscited by 50
  20. 2024Large Language Models Assume People are More Rational than We Really arecited by 49
  21. 2025AutoAgent: A Fully-Automated and Zero-Code Framework for LLM Agentscited by 49
  22. 2025The Ideation-Execution Gap: Execution Outcomes of LLM-Generated versus Human Research Ideascited by 49
  23. 2025A Comprehensive Survey of Deep Research: Systems, Methodologies, and Applicationscited by 48
  24. 2025Contextual Experience Replay for Self-Improvement of Language Agentscited by 46
  25. 2023MoCa: Measuring Human-Language Model Alignment on Causal and Moral Judgment Taskscited by 44
  26. 2024Breaking Down the Defenses: A Comparative Survey of Attacks on Large Language Modelscited by 42
  27. 2024Moral Alignment for LLM Agentscited by 39
  28. 2024FairMT-Bench: Benchmarking Fairness for Multi-turn Dialogue in Conversational LLMscited by 36
  29. 2024BAGEL: Bootstrapping Agents by Guiding Exploration with Languagecited by 34
  30. 2025ReportBench: Evaluating Deep Research Agents via Academic Survey Taskscited by 34
  31. 2025ResearcherBench: Evaluating Deep AI Research Systems on the Frontiers of Scientific Inquirycited by 30
  32. 2024Incoherent Probability Judgments in Large Language Modelscited by 29
  33. 2025Characterizing Deep Research: A Benchmark and Formal Definitioncited by 29
  34. 2025CURIE: Evaluating LLMs On Multitask Scientific Long Context Understanding and Reasoningcited by 28
  35. 2024How do Large Language Models Navigate Conflicts between Honesty and Helpfulness?cited by 27
  36. 2025Persuade Me if You Can: A Framework for Evaluating Persuasion Effectiveness and Susceptibility Among Large Language Modelscited by 25
  37. 2025The More You Automate, the Less You See: Hidden Pitfalls of AI Scientist Systemscited by 24
  38. 2025Synthetic Socratic Debates: Examining Persona Effects on Moral Decision and Persuasion Dynamicscited by 19

50 most cited of 63 All 63 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 25.

25 papers.

Sort:

By area:

  1. 2026Characterizing Delusional Spirals through Human-LLM Chat Logscited by 22
  2. 2026On Safety Risks in Experience-Driven Self-Evolving Agentscited by 12
  3. 2026BeliefShift: Benchmarking Temporal Belief Consistency and Opinion Drift in LLM Agentscited by 11
  4. 2026Belief Memory: Agent Memory Under Partial Observabilitycited by 9
  5. 2026Linear representations in language models can change dramatically over a conversationcited by 7
  6. 2026Position: Modular Memory is the Key to Continual Learning Agentscited by 7
  7. 2026Ads in AI Chatbots? An Analysis of How Large Language Models Navigate Conflicts of Interestcited by 3
  8. 2026Useless but Safe? Benchmarking Utility Recovery with User Intent Clarification in Multi-Turn Conversationscited by 3
  9. 2026Recalling Too Well: Sycophancy Evaluation and Mitigation in Memory-Augmented Modelscited by 3
  10. 2026Illusions of Confidence? Diagnosing LLM Truthfulness via Neighborhood Consistencycited by 2
  11. 2026Intersectional Sycophancy: How Perceived User Demographics Shape False Validation in Large Language Modelscited by 2
  12. 2026Topics as Proxies for Sociodemographics: How Conversational Context Affects LLM Answerscited by 1
13 more
  1. 2026To Nuke or Not to Nuke: LLMs' (Missing) Ethical Reasoning and Actions in a High-Stakes Decision-Making Simulationcited by 1
  2. 2026What LLM Agents Say When No One Is Watching: Social Structure and Latent Objective Emergence in Multi-Agent Debatescited by 1
  3. 2026Beyond Sycophancy: Structured Resistance and Compliance in LLM Moral Reasoningcited by 1
  4. 2026Persuasion Propagation: Does Persuasion Change AI Agent Behavior?cited by 0
  5. 2026Can LLMs Take Retrieved Information with a Grain of Salt?cited by 0
  6. 2026Evidence-Informed LLM Beliefs for Continual Scientific Discoverycited by 0
  7. 2026Relational Positioning as a Measurable Risk Object: History-Carried Lock-in and Self-Confabulation in Multi-Turn Human-AI Dialoguecited by 0
  8. 2026DelusionEval: Measuring Delusion-Linked Behaviors in AI Chatbotscited by 0
  9. 2026Identifying Harm in Personalized, Generative AI Systems Requires User-Centered Auditing at the Interaction Levelcited by 0
  10. 2026Belief Cascades Drive Persuasion in LLM Agent Networkscited by 0
  11. 2026Do LLMs Change Their Minds Like Humans? Diagnosing Human--LLM Divergence in Single-Turn Persuasion Judgmentscited by 0
  12. 2026Illusory Truth or Mere Exposure? Model-Dependent Repetition Effects in LLM-Based Social Media Simulationscited by 0
  13. 2026EurekaBench: Measuring Agentic Ability to Discover New Scientific Insightscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Beliefs and Behavior in Language Models
  2. 2024Brittle Minds, Fixable Activations: Understanding Belief Representations in Language Models
  3. 2026Beyond Words: Evaluating and Bridging Epistemic Divergence in User-Agent Interaction via Theory of Mind
  4. 2026Dynamic Theory of Mind as a Temporal Memory Problem: Evidence from Large Language Models
  5. 2022Language Models as Agent Models
  6. 2024In-context learning agents are asymmetric belief updaters
  7. 2026Language Statistics and False Belief Reasoning: Evidence from 41 Open-Weight LMs
  8. 2026Probing the Misaligned Thinking Process of Language Models