Learning Personalized Alignment for Evaluating Open-ended Text Generation

• URL: http://arxiv.org/abs/2310.03304v4
• Date: Wed, 19 Jun 2024 22:05:12 GMT
• Title: Learning Personalized Alignment for Evaluating Open-ended Text Generation
• Authors: Danqing Wang, Kevin Yang, Hanlin Zhu, Xiaomeng Yang, Andrew Cohen, Lei Li, Yuandong Tian,
• Abstract summary: We introduce an interpretable open-ended evaluation framework PerSE to assess the alignment with a specific human preference. Our 13B LLaMA-2-based PerSE shows a 15.8% increase in Kendall correlation and a 13.7% rise in accuracy on zero-shot reviewers. It also outperforms GPT-4 by 46.01% in the Kendall correlation on new domains, indicating its transferability.
• Score: 44.565686959174585
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: With rapid progress made in language qualities such as fluency and consistency via large language models (LLMs), there has been increasing interest in assessing alignment with diverse human preferences. Traditional metrics heavily rely on lexical similarity with human-written references and have been observed to suffer from a poor correlation with human evaluation. Furthermore, they ignore the diverse preferences of humans, a key aspect in evaluating open-ended tasks like story generation. Inspired by these challenges, we introduce an interpretable open-ended evaluation framework PerSE to assess the alignment with a specific human preference. It is tuned to deduce the specific preference from a given personal profile and evaluate the alignment between the generation and the personal preference. PerSE also explains its assessment by a detailed comment or several fine-grained scores. This enhances its interpretability, making it more suitable to tailor a personalized generation. Our 13B LLaMA-2-based PerSE shows a 15.8% increase in Kendall correlation and a 13.7% rise in accuracy on zero-shot reviewers compared to GPT-4. It also outperforms GPT-4 by 46.01% in the Kendall correlation on new domains, indicating its transferability.

Related papers

• When Personalization Meets Reality: A Multi-Faceted Analysis of Personalized Preference Learning [23.557084253364174]
  Reinforcement Learning from Human Feedback (RLHF) typically assumes homogeneous preferences across users, overlooking diverse human values and minority viewpoints. We present a multi-faceted evaluation framework that measures not only performance but also fairness, unintended effects, and adaptability across varying levels of preference divergence. These findings highlight the critical need for holistic evaluation approaches to advance the development of more effective and inclusive preference learning systems.
  arXiv  Detail & Related papers  (2025-02-26T14:14:58Z)
• SedarEval: Automated Evaluation using Self-Adaptive Rubrics [4.97150240417381]
  We propose a new evaluation paradigm based on self-adaptive rubrics. SedarEval consists of 1,000 meticulously crafted questions, each with its own self-adaptive rubric. We train a specialized evaluator language model (evaluator LM) to supplant human graders.
  arXiv  Detail & Related papers  (2025-01-26T16:45:09Z)
• CRoP: Context-wise Robust Static Human-Sensing Personalization [7.303478793193849]
  This work introduces CRoP, a novel static personalization approach. CRoP captures user-specific traits through adaptive pruning on a minimal sub-network while preserving generic knowledge in the remaining parameters. It demonstrates superior personalization effectiveness and intra-user robustness across four human-sensing datasets, including two from real-world health domains.
  arXiv  Detail & Related papers  (2024-09-26T16:06:38Z)
• Aligning Model Evaluations with Human Preferences: Mitigating Token Count Bias in Language Model Assessments [2.1370543868467275]
  This follow-up paper explores methods to align Large Language Models evaluator preferences with human evaluations. We employed Bayesian statistics and a t-test to quantify this bias and developed a recalibration procedure to adjust the GPTScorer. Our findings significantly improve aligning the recalibrated LLM evaluator with human evaluations across multiple use cases.
  arXiv  Detail & Related papers  (2024-07-05T09:26:40Z)
• Prometheus 2: An Open Source Language Model Specialized in Evaluating Other Language Models [92.66784679667441]
  Prometheus 2 is a more powerful evaluator LM that closely mirrors human and GPT-4 judgements. It is capable of processing both direct assessment and pairwise ranking formats grouped with a user-defined evaluation criteria. On four direct assessment benchmarks and four pairwise ranking benchmarks, Prometheus 2 scores the highest correlation and agreement with humans and proprietary LM judges.
  arXiv  Detail & Related papers  (2024-05-02T17:59:35Z)
• CritiqueLLM: Towards an Informative Critique Generation Model for Evaluation of Large Language Model Generation [87.44350003888646]
  Eval-Instruct can acquire pointwise grading critiques with pseudo references and revise these critiques via multi-path prompting. CritiqueLLM is empirically shown to outperform ChatGPT and all the open-source baselines.
  arXiv  Detail & Related papers  (2023-11-30T16:52:42Z)
• Automated Evaluation of Personalized Text Generation using Large Language Models [38.2211640679274]
  We present AuPEL, a novel evaluation method that distills three major semantic aspects of the generated text: personalization, quality and relevance, and automatically measures these aspects. We find that, compared to existing evaluation metrics, AuPEL not only distinguishes and ranks models based on their personalization abilities more accurately, but also presents commendable consistency and efficiency for this task.
  arXiv  Detail & Related papers  (2023-10-17T21:35:06Z)
• Prometheus: Inducing Fine-grained Evaluation Capability in Language Models [66.12432440863816]
  We propose Prometheus, a fully open-source Large Language Model (LLM) that is on par with GPT-4's evaluation capabilities. Prometheus scores a Pearson correlation of 0.897 with human evaluators when evaluating with 45 customized score rubrics. Prometheus achieves the highest accuracy on two human preference benchmarks.
  arXiv  Detail & Related papers  (2023-10-12T16:50:08Z)
• FLASK: Fine-grained Language Model Evaluation based on Alignment Skill Sets [69.91340332545094]
  We introduce FLASK, a fine-grained evaluation protocol for both human-based and model-based evaluation. We experimentally observe that the fine-graininess of evaluation is crucial for attaining a holistic view of model performance.
  arXiv  Detail & Related papers  (2023-07-20T14:56:35Z)
• Revisiting the Gold Standard: Grounding Summarization Evaluation with Robust Human Evaluation [136.16507050034755]
  Existing human evaluation studies for summarization either exhibit a low inter-annotator agreement or have insufficient scale. We propose a modified summarization salience protocol, Atomic Content Units (ACUs), which is based on fine-grained semantic units. We curate the Robust Summarization Evaluation (RoSE) benchmark, a large human evaluation dataset consisting of 22,000 summary-level annotations over 28 top-performing systems.
  arXiv  Detail & Related papers  (2022-12-15T17:26:05Z)
• SMURF: SeMantic and linguistic UndeRstanding Fusion for Caption Evaluation via Typicality Analysis [20.026835809227283]
  We introduce "typicality", a new formulation of evaluation rooted in information theory. We show how these decomposed dimensions of semantics and fluency provide greater system-level insight into captioner differences. Our proposed metrics along with their combination, SMURF, achieve state-of-the-art correlation with human judgment when compared with other rule-based evaluation metrics.
  arXiv  Detail & Related papers  (2021-06-02T19:58:20Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.