# Adapting LLM Judges to Observed Human Annotation Policy through Soft Prompt Tuning

DevFeed: [Adapting LLM Judges to Observed Human Annotation Policy through Soft Prompt Tuning](<https://devfeed.tech/articles/adapting-llm-judges-to-observed-human-annotation-policy-through-soft-prompt-tuning-120164.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.07126>)

Author: Tam\\'as Janusko, Sven Kleinkop, Ricardo Bochnia, Gunnar Hempel, Steffen Tomschke, J\\"urgen Anke, Maik Thiele

Published: 2026-11-06T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [ai alignment](<https://devfeed.tech/topics/ai-alignment.md>), [human feedback](<https://devfeed.tech/topics/human-feedback.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Enhancing AI model performance](<https://devfeed.tech/topics/enhancing-ai-model-performance.md>), [prompt](<https://devfeed.tech/topics/prompt.md>), [qwen](<https://devfeed.tech/topics/qwen.md>)

Tags: [3](<https://devfeed.tech/tags/3.md>), [adapting](<https://devfeed.tech/tags/adapting.md>), [align](<https://devfeed.tech/tags/align.md>), [annotation](<https://devfeed.tech/tags/annotation.md>), [annotations](<https://devfeed.tech/tags/annotations.md>), [automated](<https://devfeed.tech/tags/automated.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [bias](<https://devfeed.tech/tags/bias.md>), [chatbot](<https://devfeed.tech/tags/chatbot.md>), [continuous](<https://devfeed.tech/tags/continuous.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [dimensions](<https://devfeed.tech/tags/dimensions.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [hybrid](<https://devfeed.tech/tags/hybrid.md>), [language](<https://devfeed.tech/tags/language.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [large-language-models-llms](<https://devfeed.tech/tags/large-language-models-llms.md>), [llm](<https://devfeed.tech/tags/llm.md>), [policy](<https://devfeed.tech/tags/policy.md>), [prompt](<https://devfeed.tech/tags/prompt.md>)

## AI overview

The paper reports that hybrid soft-prompt tuning--combining textual prompts with learned continuous prompt embeddings--improves open-source LLM judges' agreement with human annotations over untuned baselines. Prompts were trained on FairBotBench and evaluated for bias, toxicity, and personalization. The abstract says adapted Qwen 3.5 9B judges surpassed both hosted reference judges across all three dimensions.

## Source excerpt

arXiv:2611.07126v1 Announce Type: cross Abstract: Large Language Models (LLMs) are increasingly employed as automated judges for evaluating chatbot responses. While proprietary models can achieve high agreement with human annotators, they remain costly and difficult to deploy in practical settings. Lightweight open-source LLM judges provide an attractive alternative but typically exhibit lower agreement with human ratings. In this paper, we investigate whether soft prompt tuning can align open-source LLM judges more closely with human annotation policy. We train soft prompts on the FairBotBench benchmark and evaluate the adapted judges across the dimensions of bias, toxicity, and personalization. Our findings demonstrate that hybrid soft-prompt tuning, which combines textual prompts with learned continuous prompt embeddings, consistently improves agreement with human annotations over the corresponding untuned baselines. For Qwen 3.5 9B, the adapted judges surpass both hosted reference judges across all three evaluation dimensions.