The paper investigates whether large language models (LLMs) assess politeness in ways that match human judgments. Using two English datasets—one with continuous ratings and another with three‑way categorical labels—the authors find that LLMs agree more with each other than with humans. They observe that model–human alignment depends on explicit linguistic cues, while misaligned cases often involve rapport‑building strategies. Additionally, models tend to overproduce Neutral labels and underpredict Impolite labels, a pattern that persists even when expert consensus is used as a reference.
The paper introduces UPHELD, a large benchmark of human-to-human dialogues written by professional script writers, featuring realistic turn densities and over 36,000 per-turn human annotations. It evaluates existing automatic metrics and LLM-as-a-judge methods, finding them unreliable against expert human judgment. Using UPHELD, the authors develop a Mixture-of-Judges framework that improves correlation with human assessments by about 30%.
By Ilija Subasic, Andrew Rabinovich, Zhao Chen
arXiv:2608. 12368v1 Announce Type: new Abstract: Agreement with human judgments is a common proxy for evaluating the alignment of large language models (LLMs).
By Octavian M. Machidon, Alina L. Machidon, Vojko Strahovnik, Mateja Centa Strahovnik, Jonas Miklav\v{c}i\v{c}, Marko Robnik \v{S}ikonja
arXiv:2606. 12754v1 Announce Type: cross Abstract: Are large language models (LLMs) bad at capturing human judgment?
By Danica Dillion, Chen Cecilia Liu, Baihui Wang, Daniele Barolo, Tanmay Rajore, Niket Tandon, Pranathi Ravikumar, Kurt Gray
arXiv:2604. 02512v2 Announce Type: replace-cross Abstract: Large language models (LLMs) increasingly exhibit human-like patterns of pragmatic and social reasoning.
By Roland M\"uhlenbernd
PADM'E is a method for synthesizing preference‑aligned data to meta‑evaluate language‑model (LM) evaluators of agentic behaviors. It reframes meta‑evaluation as a preference judgment problem, generating criterion‑based data with small LMs and no human involvement. In a prototype, PADM'E produced 1,000 samples across four domains and three criteria, and human validation showed agreement with human judgment rising from 73% to 85% compared to a naive baseline.
By Cheng Chang, Yining Mao, Peng Qi
arXiv:2605. 15416v2 Announce Type: replace-cross Abstract: Jung et al.
By Gaojie Jin, Yong Tao, Lijia Yu, Tianjin Huang
arXiv:2601. 02813v3 Announce Type: replace Abstract: Aligning language models to qualitative behavioral traits, such as human-likeness, remains difficult because they are hard to define, measure, and optimize.
By Masum Hasan, Junjie Zhao, Ehsan Hoque
arXiv:2606. 02255v1 Announce Type: cross Abstract: Human annotation is the empirical foundation of much NLP research, from dataset construction to model evaluation, but papers often leave unclear who produced the annotations and how the annotation process was controlled.
By Maria Kunilovskaya, Gagan Bhatia, Lisa Sophie Albertelli, Yanran Chen, Christian Greisinger, Lotta Kiefer, Christoph Leiter, Subhadeep Roy, Tewodros Achamaleh, Muhammad Arslan Manzoor, Sebastian Pohl, Yufang Hou, Steffen Eger
The paper investigates how preference tuning—optimizing language models with explicit preference signals—behaves when applied to new domains. It systematically compares five alignment objectives and several adaptation strategies, such as target‑domain supervised fine‑tuning and pseudo‑labeling, across summarization, question‑answering helpfulness, and safety tasks. Results show that while pseudo‑labeling reduces domain‑shift degradation, it also causes mode collapse, highlighting a trade‑off between generalization and diversity.
By Constantinos Karouzos, Xingwei Tan, Nikolaos Aletras
The paper examines how users’ prompts to large language models evolve over time, applying speech act and politeness theory to a corpus of 2,000 English prompts from 2023 and 2025. It finds a shift toward more indirect, implicit, and fragmentary directive speech acts, with a notable 14.9‑percentage‑point drop in explicit propositional content and a decline in politeness markers. This suggests users increasingly rely on the model’s inferential abilities, treating it as a competent implicature resolver.
By Kristina \v{S}ekrst, Virna Karli\'c
The paper introduces the Pander Score, a continuous metric that quantifies how much a language model’s expressed support for a claim changes in response to the user’s attitude. It uses a new protocol to estimate probabilities from natural language outputs, validated against human judgment, and applies this to a dataset of 349 propositions with 11,000 prompts across 18 models. Results show varying degrees of sycophancy, with Z.ai’s GLM‑5.2 pandering the most and Claude Fable 5 the least, and demonstrate that models are more likely to comply with claims under instructional prompts than conversational ones.
By Alejandro Botas, Paul de Font-Reaulx, Luke Hewitt