mlsanityData to 13 Oct 2026

On the Role of Citations in Preference Data

arXiv 2608.21376 ↗First posted 6 Jul 2026cs.CL cs.AIcited by 0

On the map: LLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment

Abstract (arXiv)

Many NLP tasks require systems to provide attribution in their outputs--i.e. citations to grounding sources. Attribution serves as a bulwark against model hallucination and as a means for users to verify the credibility of model outputs. Yet, it is unclear how humans and LLMs evaluate citations when comparing outputs, a process central to reward modeling and modern LLM post-training. This paper studies the role of citations in the preferences of human judges and four open-source LLMs within the context of scientific question answering, leveraging mixed effects models to investigate the influence of citations on pairwise judgments. Among our key findings are (1) that humans prefer more diverse citations but fewer overall, and (2) that LLMs show some citation-related preferences compared to humans, despite lacking access to the sources, but these preferences depend on the data and specific models. We further discuss the implications of our findings for preference data collection.

Key facts. No citing papers in the map yet (data through 13 Oct 2026). Placed in Human values and AI feedback alignment, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 39 of its 65 references that are in the map. First posted on arXiv 6 Jul 2026.

Builds on

39 of 65 references.

Closest ancestors · 5

Same direction, newest first

  1. 2026Aligning Large Language Model Behavior with Human Citation Preferences5 Feb
  2. 2025What's In My Human Feedback? Learning Interpretable Descriptions of Preference Data30 Oct
  3. 2025Flattery, Fluff, and Fog: Diagnosing and Mitigating Idiosyncratic Biases in Preference Models5 Jun
  4. 2024Aligning with Human Judgement: The Role of Pairwise Preference in Large Language Model Evaluators25 Mar
  5. 2024Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference7 Mar

Other references · 34

  1. 2022Training language models to follow instructions with human feedbackcited by 15,856
  2. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,831
  3. 2019BERTScore: Evaluating Text Generation with BERTcited by 5,480
  4. 2024A Survey on LLM-as-a-Judgecited by 1,399
  5. 2024Humans or LLMs as the Judge? A Study on Judgement Biasescited by 317
  6. 2024JudgeBench: A Benchmark for Evaluating LLM-based Judgescited by 304
  7. 2024Judging the Judges: A Systematic Study of Position Bias in LLM-as-a-Judgecited by 258
  8. 2025DeepResearch Bench: A Comprehensive Benchmark for Deep Research Agentscited by 239
  9. 2025DR Tulu: Reinforcement Learning with Evolving Rubrics for Deep Researchcited by 106
  10. 2025ResearchRubrics: A Benchmark of Prompts and Rubrics For Evaluating Deep Research Agentscited by 81
  11. 2024LongCite: Enabling LLMs to Generate Fine-grained Citations in Long-context QAcited by 78
  12. 2025One Token to Fool LLM-as-a-Judgecited by 56
22 more
  1. 2025ResearchQA: Evaluating Scholarly Question Answering at Scale Across 75 Fields with Survey-Mined Questions and Rubricscited by 45
  2. 2024Training Language Models to Generate Text with Citations via Fine-grained Rewardscited by 40
  3. 2025Search Arena: Analyzing Search-Augmented LLMscited by 31
  4. 2025Copilot Arena: A Platform for Code LLM Evaluation in the Wildcited by 25
  5. 2025reWordBench: Benchmarking and Improving the Robustness of Reward Models with Transformed Inputscited by 25
  6. 2025Ai2 Scholar QA: Organized Literature Synthesis with Attributioncited by 24
  7. 2024RAG-RewardBench: Benchmarking Reward Models in Retrieval Augmented Generation for Preference Alignmentcited by 22
  8. 2024Explaining Datasets in Words: Statistical Models with Natural Language Parameterscited by 21
  9. 2024Large Language Models Reflect Human Citation Patterns with a Heightened Citation Biascited by 16
  10. 2025MetaFaith: Faithful Natural Language Uncertainty Expression in LLMscited by 16
  11. 2025PrefPalette: Personalized Preference Modeling with Latent Attributescited by 15
  12. 2025SurGE: A Benchmark and Evaluation Framework for Scientific Survey Generationcited by 15
  13. 2022Dialect-robust Evaluation of Generated Textcited by 14
  14. 2025Do Large Language Models Favor Recent Content? A Study on Recency Bias in LLM-Based Rerankingcited by 10
  15. 2020Natural Language Inference with Mixed Effectscited by 9
  16. 2025Can External Validation Tools Improve Annotation Quality for LLM-as-a-Judge?cited by 9
  17. 2024Citation Amnesia: On The Recency Bias of NLP and Other Academic Fieldscited by 5
  18. 2024Can Users Detect Biases or Factual Errors in Generated Responses in Conversational Information-Seeking?cited by 5
  19. 2026Deep Research, Shallow Evaluation: A Case Study in Meta-Evaluation for Long-Form QA Benchmarkscited by 5
  20. 2022Mixed-effects transformers for hierarchical adaptationcited by 2
  21. 2025Reward Models are Metrics in a Trench Coatcited by 2
  22. 2026Rethinking Literature Search Evaluation: Deep Research Helps, and Human Citation Lists Are Not a Ground Truthcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2023DecipherPref: Analyzing Influential Factors in Human Preference Judgments via GPT-4
  2. 2026Whose Facts Win? LLM Source Preferences under Knowledge Conflicts
  3. 2026How Do LLMs Cite? A Mechanistic Interpretation of Attribution in Retrieval-Augmented Generation
  4. 2024Improving Attributed Text Generation of Large Language Models via Preference Learning
  5. 2023Can LLMs Capture Human Preferences?
  6. 2025Rethinking Human Preference Evaluation of LLM Rationales
  7. 2024Attribute or Abstain: Large Language Models as Long Document Assistants
  8. 2026LLMs Contain Multitudes: How Deployment Context Reshapes Model-Level Preferences and Values