Saved in:
Bibliographic Details
Main Authors: Zhang, Xingjian, Gao, Tianhong, Jin, Suliang, Wang, Tianhao, Ye, Teng, Adar, Eytan, Mei, Qiaozhu
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.25860
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915808421085184
author Zhang, Xingjian
Gao, Tianhong
Jin, Suliang
Wang, Tianhao
Ye, Teng
Adar, Eytan
Mei, Qiaozhu
author_facet Zhang, Xingjian
Gao, Tianhong
Jin, Suliang
Wang, Tianhao
Ye, Teng
Adar, Eytan
Mei, Qiaozhu
contents Large language models (LLMs) are increasingly used as raters for evaluation tasks. However, their reliability is often limited for subjective tasks, when human judgments involve subtle reasoning beyond annotation labels. Thinking traces, the reasoning behind a judgment, are highly informative but challenging to collect and curate. We present a human-LLM collaborative framework to infer thinking traces from label-only annotations. The proposed framework uses a simple and effective rejection sampling method to reconstruct these traces at scale. These inferred thinking traces are applied to two complementary tasks: (1) fine-tuning open LLM raters; and (2) synthesizing clearer annotation guidelines for proprietary LLM raters. Across multiple datasets, our methods lead to significantly improved LLM-human agreement. Additionally, the refined annotation guidelines increase agreement among different LLM models. These results suggest that LLMs can serve as practical proxies for otherwise unrevealed human thinking traces, enabling label-only corpora to be extended into thinking-trace-augmented resources that enhance the reliability of LLM raters.
format Preprint
id arxiv_https___arxiv_org_abs_2510_25860
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Through the Judge's Eyes: Inferred Thinking Traces Improve Reliability of LLM Raters
Zhang, Xingjian
Gao, Tianhong
Jin, Suliang
Wang, Tianhao
Ye, Teng
Adar, Eytan
Mei, Qiaozhu
Artificial Intelligence
Computation and Language
Human-Computer Interaction
Large language models (LLMs) are increasingly used as raters for evaluation tasks. However, their reliability is often limited for subjective tasks, when human judgments involve subtle reasoning beyond annotation labels. Thinking traces, the reasoning behind a judgment, are highly informative but challenging to collect and curate. We present a human-LLM collaborative framework to infer thinking traces from label-only annotations. The proposed framework uses a simple and effective rejection sampling method to reconstruct these traces at scale. These inferred thinking traces are applied to two complementary tasks: (1) fine-tuning open LLM raters; and (2) synthesizing clearer annotation guidelines for proprietary LLM raters. Across multiple datasets, our methods lead to significantly improved LLM-human agreement. Additionally, the refined annotation guidelines increase agreement among different LLM models. These results suggest that LLMs can serve as practical proxies for otherwise unrevealed human thinking traces, enabling label-only corpora to be extended into thinking-trace-augmented resources that enhance the reliability of LLM raters.
title Through the Judge's Eyes: Inferred Thinking Traces Improve Reliability of LLM Raters
topic Artificial Intelligence
Computation and Language
Human-Computer Interaction
url https://arxiv.org/abs/2510.25860