# DecipherPref: Analyzing Influential Factors in Human Preference Judgments via GPT-4

arXiv: https://arxiv.org/abs/2305.14702 · HTML: https://mlsanity.com/paper/2305.14702/

## Key facts

- Cited by 16 papers in the map (data through 14 Oct 2026).
- Builds on 38 of its 69 references that are in the map.
- First posted on arXiv 24 May 2023.

## Builds on (most cited first, 30 of 38)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [BERTScore: Evaluating Text Generation with BERT](https://mlsanity.com/paper/1904.09675/) (2019)
- [Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2204.05862/) (2022)
- [Learning to summarize from human feedback](https://mlsanity.com/paper/2009.01325/) (2020)
- [Fine-Tuning Language Models from Human Preferences](https://mlsanity.com/paper/1909.08593/) (2019)
- [WebGPT: Browser-assisted question-answering with human feedback](https://mlsanity.com/paper/2112.09332/) (2021)
- [BLEURT: Learning Robust Metrics for Text Generation](https://mlsanity.com/paper/2004.04696/) (2020)
- [OpenAssistant Conversations -- Democratizing Large Language Model Alignment](https://mlsanity.com/paper/2304.07327/) (2023)
- [SummEval: Re-evaluating Summarization Evaluation](https://mlsanity.com/paper/2007.12626/) (2020)
- [Benchmarking Large Language Models for News Summarization](https://mlsanity.com/paper/2301.13848/) (2023)
- [MoverScore: Text Generation Evaluating with Contextualized Embeddings and Earth Mover Distance](https://mlsanity.com/paper/1909.02622/) (2019)
- [News Summarization and Evaluation in the Era of GPT-3](https://mlsanity.com/paper/2209.12356/) (2022)
- [Newsroom: A Dataset of 1.3 Million Summaries with Diverse Extractive Strategies](https://mlsanity.com/paper/1804.11283/) (2018)
- [Teaching language models to support answers with verified quotes](https://mlsanity.com/paper/2203.11147/) (2022)
- [All That's 'Human' Is Not Gold: Evaluating Human Evaluation of Generated Text](https://mlsanity.com/paper/2107.00061/) (2021)
- [Towards a Unified Multi-Dimensional Evaluator for Text Generation](https://mlsanity.com/paper/2210.07197/) (2022)
- [Principled Reinforcement Learning with Human Feedback from Pairwise or K-wise Comparisons](https://mlsanity.com/paper/2301.11270/) (2023)
- [Is Reinforcement Learning (Not) for Natural Language Processing: Benchmarks, Baselines, and Building Blocks for Natural Language Policy Optimization](https://mlsanity.com/paper/2210.01241/) (2022)
- [QAFactEval: Improved QA-Based Factual Consistency Evaluation for Summarization](https://mlsanity.com/paper/2112.08542/) (2021)
- [Re-evaluating Evaluation in Text Summarization](https://mlsanity.com/paper/2010.07100/) (2020)
- [Revisiting the Gold Standard: Grounding Summarization Evaluation with Robust Human Evaluation](https://mlsanity.com/paper/2212.07981/) (2022)
- [Hallucinated but Factual! Inspecting the Factuality of Hallucinations in Abstractive Summarization](https://mlsanity.com/paper/2109.09784/) (2021)
- [Human-like Summarization Evaluation with ChatGPT](https://mlsanity.com/paper/2304.02554/) (2023)
- [On Learning to Summarize with Large Language Models as References](https://mlsanity.com/paper/2305.14239/) (2023)
- [Multi-Reward Reinforced Summarization with Saliency and Entailment](https://mlsanity.com/paper/1804.06451/) (2018)
- [Generating Personalized Recipes from Historical User Preferences](https://mlsanity.com/paper/1909.00105/) (2019)
- [Better than Average: Paired Evaluation of NLP Systems](https://mlsanity.com/paper/2110.10746/) (2021)
- [Multimodal Knowledge Alignment with Reinforcement Learning](https://mlsanity.com/paper/2205.12630/) (2022)
- [The Summary Loop: Learning to Write Abstractive Summaries Without Examples](https://mlsanity.com/paper/2105.05361/) (2021)
- [Towards Interpretable and Efficient Automatic Reference-Based Summarization Evaluation](https://mlsanity.com/paper/2303.03608/) (2023)

## Built on by (most cited first, 16 of 16)

- [Humans or LLMs as the Judge? A Study on Judgement Biases](https://mlsanity.com/paper/2402.10669/) (2024)
- [LLM-based NLG Evaluation: Current Status and Challenges](https://mlsanity.com/paper/2402.01383/) (2024)
- [Usable XAI: 10 Strategies Towards Exploiting Explainability in the LLM Era](https://mlsanity.com/paper/2403.08946/) (2024)
- [Regularized Best-of-N Sampling with Minimum Bayes Risk Objective for Language Model Alignment](https://mlsanity.com/paper/2404.01054/) (2024)
- [Position: Towards Bidirectional Human-AI Alignment](https://mlsanity.com/paper/2406.09264/) (2024)
- [What Makes a Good Story and How Can We Measure It? A Comprehensive Survey of Story Evaluation](https://mlsanity.com/paper/2408.14622/) (2024)
- [SportsMetrics: Blending Text and Numerical Data to Understand Information Fusion in LLMs](https://mlsanity.com/paper/2402.10979/) (2024)
- [AFSPP: Agent Framework for Shaping Preference and Personality with Large Language Models](https://mlsanity.com/paper/2401.02870/) (2024)
- [AdParaphrase: Paraphrase Dataset for Analyzing Linguistic Features toward Generating Attractive Ad Texts](https://mlsanity.com/paper/2502.04674/) (2025)
- [Rescue: Ranking LLM Responses with Partial Ordering to Improve Response Generation](https://mlsanity.com/paper/2311.09136/) (2023)
- [DeFine: Decision-Making with Analogical Reasoning over Factor Profiles](https://mlsanity.com/paper/2410.01772/) (2024)
- [Rethinking Human Preference Evaluation of LLM Rationales](https://mlsanity.com/paper/2509.11026/) (2025)
- [Uncovering Factor Level Preferences to Improve Human-Model Alignment](https://mlsanity.com/paper/2410.06965/) (2024)
- [From Fragments to Facts: A Curriculum-Driven DPO Approach for Generating Hindi News Veracity Explanations](https://mlsanity.com/paper/2507.05179/) (2025)
- [Structuring the Space of Sociotechnical Alignment](https://mlsanity.com/paper/2607.01250/) (2026)
- [Improving the End-to-End Efficiency of Offline Inference for Multi-LLM Applications Based on Sampling and Simulation](https://mlsanity.com/paper/2503.16893/) (2025)

## Abstract (arXiv)

Human preference judgments are pivotal in guiding large language models (LLMs) to produce outputs that align with human values. Human evaluations are also used in summarization tasks to compare outputs from various systems, complementing existing automatic metrics. Despite their significance, however, there has been limited research probing these pairwise or k-wise comparisons. The collective impact and relative importance of factors such as output length, informativeness, fluency, and factual consistency are still not well understood. It is also unclear if there are other hidden factors influencing human judgments. In this paper, we conduct an in-depth examination of a collection of pairwise human judgments released by OpenAI. Utilizing the Bradley-Terry-Luce (BTL) model, we reveal the inherent preferences embedded in these human judgments. We find that the most favored factors vary across tasks and genres, whereas the least favored factors tend to be consistent, e.g., outputs are too brief, contain excessive off-focus content or hallucinated facts. Our findings have implications on the construction of balanced datasets in human preference evaluations, which is a crucial step in shaping the behaviors of future LLMs.