mlsanityData to 7 Oct 2026

LLM Evaluators Recognize and Favor Their Own Generations

arXiv 2404.13076 ↗First posted 15 Apr 2024cs.CL cs.AIcited by 651

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

Abstract (arXiv)

Self-evaluation using large language models (LLMs) has proven valuable not only in benchmarking but also methods like reward modeling, constitutional AI, and self-refinement. But new biases are introduced due to the same LLM acting as both the evaluator and the evaluatee. One such bias is self-preference, where an LLM evaluator scores its own outputs higher than others' while human annotators consider them of equal quality. But do LLMs actually recognize their own outputs when they give those texts higher scores, or is it just a coincidence? In this paper, we investigate if self-recognition capability contributes to self-preference. We discover that, out of the box, LLMs such as GPT-4 and Llama 2 have non-trivial accuracy at distinguishing themselves from other LLMs and humans. By fine-tuning LLMs, we discover a linear correlation between self-recognition capability and the strength of self-preference bias; using controlled experiments, we show that the causal explanation resists straightforward confounders. We discuss how self-recognition can interfere with unbiased evaluations and AI safety more generally.

Key facts. Cited by 651 papers in the map (data through 7 Oct 2026). Placed in Human values and AI feedback alignment, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 37 of its 53 references that are in the map. First posted on arXiv 15 Apr 2024.

Builds on

37 of 53 references.

Closest ancestors · 2

Same direction, newest first

  1. 2022Constitutional AI: Harmlessness from AI Feedback15 Dec
  2. 2020Learning to summarize from human feedback2 Sep

Other references · 35

  1. 2020Language Models are Few-Shot Learnerscited by 32,199
  2. 2023GPT-4 Technical Reportcited by 18,132
  3. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  4. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  5. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,317
  6. 2022Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language modelscited by 1,970
  7. 2022Language Models (Mostly) Know What They Knowcited by 1,783
  8. 2018Don't Give Me the Details, Just the Summary! Topic-Aware Convolutional Neural Networks for Extreme Summarizationcited by 1,308
  9. 2016Abstractive Text Summarization Using Sequence-to-Sequence RNNs and Beyondcited by 809
  10. 2023DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvaturecited by 705
  11. 2018Scalable agent alignment via reward modeling: a research directioncited by 438
  12. 2022Self-critiquing models for assisting human evaluatorscited by 372
23 more
  1. 2023GPTScore: Evaluate as You Desirecited by 357
  2. 2021Recursively Summarizing Books with Human Feedbackcited by 287
  3. 2024Spotting LLMs With Binoculars: Zero-Shot Detection of Machine-Generated Textcited by 255
  4. 2023Large Language Models Sensitivity to The Order of Options in Multiple-Choice Questionscited by 249
  5. 2023Do Large Language Models Know What They Don't Know?cited by 238
  6. 2023Benchmarking Foundation Models with Language-Model-as-an-Examinercited by 172
  7. 2020Automatic Detection of Machine Generated Text: A Critical Surveycited by 160
  8. 2023Benchmarking Cognitive Biases in Large Language Models as Evaluatorscited by 147
  9. 2023Taken out of context: On measuring situational awareness in LLMscited by 131
  10. 2024Is LLM-as-a-Judge Robust? Investigating Universal Adversarial Attacks on Zero-shot LLM Assessmentcited by 117
  11. 2023VisIT-Bench: A Benchmark for Vision-Language Instruction Following Inspired by Real-World Usecited by 100
  12. 2023LLMs as Narcissistic Evaluators: When Ego Inflates Evaluation Scorescited by 88
  13. 2022Machine Generated Text: A Comprehensive Survey of Threat Models and Detection Methodscited by 75
  14. 2023A Survey on Detection of LLMs-Generated Contentcited by 70
  15. 2024Feedback Loops With Language Models Drive In-Context Reward Hackingcited by 70
  16. 2023Knowledge of Knowledge: Exploring Known-Unknowns Uncertainty with Large Language Modelscited by 60
  17. 2023Evaluating Instruction-Tuned Large Language Models on Code Comprehension and Generationcited by 49
  18. 2023Is GPT-4 a reliable rater? Evaluating Consistency in GPT-4 Text Ratingscited by 33
  19. 2023The ART of LLM Refinement: Ask, Refine, and Trustcited by 29
  20. 2023Towards Evaluating AI Systems for Moral Status Using Self-Reportscited by 21
  21. 2024MM-SAP: A Comprehensive Benchmark for Assessing Self-Awareness of Multimodal Large Language Models in Perceptioncited by 21
  22. 2024A Survey of AI-generated Text Forensic Systems: Detection, Attribution, and Characterizationcited by 18
  23. 2023Democratizing LLMs: An Exploration of Cost-Performance Trade-offs in Self-Refined Open-Source Modelscited by 11

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 90; 2025: 183; 2026: 378.

651 papers.

Sort:

By area:

  1. 2024BigCodeBench: Benchmarking Code Generation with Diverse Function Calls and Complex Instructionscited by 527
  2. 2024From Generation to Judgment: Opportunities and Challenges of LLM-as-a-judgecited by 480
  3. 2024LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methodscited by 451
  4. 2024Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Modelscited by 301
  5. 2024Self-Preference Bias in LLM-as-a-Judgecited by 248
  6. 2024AgentClinic: a multimodal agent benchmark to evaluate AI in simulated clinical environmentscited by 215
  7. 2025An Illusion of Progress? Assessing the Current State of Web Agentscited by 165
  8. 2025GDPval: Evaluating AI Model Performance on Real-World Economically Valuable Taskscited by 142
  9. 2025RewardBench 2: Advancing Reward Model Evaluationcited by 136
  10. 2025Preference Leakage: A Contamination Problem in LLM-as-a-judgecited by 119
  11. 2024Looking Inward: Language Models Can Learn About Themselves by Introspectioncited by 105
  12. 2024Trust or Escalate: LLM Judges with Provable Guarantees for Human Agreementcited by 99
38 more
  1. 2024CRAG -- Comprehensive RAG Benchmarkcited by 94
  2. 2024On scalable oversight with weak LLMs judging strong LLMscited by 94
  3. 2025Uncertainty Quantification and Confidence Calibration in Large Language Models: A Surveycited by 93
  4. 2024VL-RewardBench: A Challenging Benchmark for Vision-Language Generative Reward Modelscited by 85
  5. 2024Foundational Autoraters: Taming Large Language Models for Better Automatic Evaluationcited by 72
  6. 2024Systematic Evaluation of LLM-as-a-Judge in LLM Alignment Tasks: Explainable Metrics and Diverse Prompt Templatescited by 71
  7. 2024UltraMedical: Building Specialized Generalists in Biomedicinecited by 67
  8. 2024Elements of World Knowledge (EWoK): A Cognition-Inspired Framework for Evaluating Basic World Knowledge in Language Modelscited by 55
  9. 2025No Free Labels: Limitations of LLM-as-a-Judge Without Human Groundingcited by 50
  10. 2024From Lists to Emojis: How Format Bias Affects Model Alignmentcited by 48
  11. 2024Limits to scalable evaluation at the frontier: LLM as Judge won't beat twice the datacited by 46
  12. 2025ResearchQA: Evaluating Scholarly Question Answering at Scale Across 75 Fields with Survey-Mined Questions and Rubricscited by 44
  13. 2025Correlated Errors in Large Language Modelscited by 42
  14. 2024Inverse Constitutional AI: Compressing Preferences into Principlescited by 39
  15. 2025Do LLM Evaluators Prefer Themselves for a Reason?cited by 38
  16. 2024PARIKSHA: A Large-Scale Investigation of Human-LLM Evaluator Agreement on Multilingual and Multi-Cultural Datacited by 37
  17. 2025Neither Valid nor Reliable? Investigating the Use of LLMs as Judgescited by 37
  18. 2024Style Outweighs Substance: Failure Modes of LLM Judges in Alignment Benchmarkingcited by 35
  19. 2025Play Favorites: A Statistical Method to Measure Self-Bias in LLM-as-a-Judgecited by 35
  20. 2024Evaluating the Evaluator: Measuring LLMs' Adherence to Task Evaluation Instructionscited by 34
  21. 2024LLM Self-Correction with DeCRIM: Decompose, Critique, and Refine for Enhanced Following of Instructions with Multiple Constraintscited by 34
  22. 2025LoRe: Personalizing LLMs via Low-Rank Reward Modelingcited by 34
  23. 2024Direct Judgement Preference Optimizationcited by 33
  24. 2024Aligning Language Models with Demonstrated Feedbackcited by 32
  25. 2024Finding Blind Spots in Evaluator LLMs with Interpretable Checklistscited by 32
  26. 2024Weak-to-Strong Reasoningcited by 32
  27. 2024Adversaries Can Misuse Combinations of Safe Modelscited by 31
  28. 2024LLMs for Generating and Evaluating Counterfactuals: A Comprehensive Studycited by 30
  29. 2025Beyond the Surface: Measuring Self-Preference in LLM Judgmentscited by 30
  30. 2025Language Models Fail to Introspect About Their Knowledge of Languagecited by 28
  31. 2025Do Large Language Models Solve the Problems of Agent-Based Modeling? A Critical Review of Generative Social Simulationscited by 28
  32. 2024JuStRank: Benchmarking LLM Judges for System Rankingcited by 27
  33. 2025The Open Proof Corpus: A Large-Scale Study of LLM-Generated Mathematical Proofscited by 27
  34. 2025How Benchmark Prediction from Fewer Data Misses the Markcited by 26
  35. 2024Fairer Preferences Elicit Improved Human-Aligned Large Language Model Judgmentscited by 25
  36. 2025Multi-turn Evaluation of Anthropomorphic Behaviours in Large Language Modelscited by 25
  37. 2024PostMark: A Robust Blackbox Watermark for Large Language Modelscited by 24
  38. 2025Safety Aware Task Planning via Large Language Models in Roboticscited by 24

50 most cited of 651 All 651 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2024Pride and Prejudice: LLM Amplifies Self-Bias in Self-Refinement
  2. 2024SELF-[IN]CORRECT: LLMs Struggle with Discriminating Self-Generated Responses
  3. 2025The Self-Execution Benchmark: Measuring LLMs' Attempts to Overcome Their Lack of Self-Execution
  4. 2024Evaluating Human Alignment and Model Faithfulness of LLM Rationale
  5. 2025Pairwise or Pointwise? Evaluating Feedback Protocols for Bias in LLM-Based Evaluation
  6. 2025Mitigating Self-Preference by Authorship Obfuscation
  7. 2024Aligning Human and LLM Judgments: Insights from EvalAssist on Task-Specific Evaluations and AI-assisted Assessment Strategy Preferences
  8. 2024Dissecting Human and LLM Preferences