The paper introduces an interpretable method for distinguishing genuine hate speech from contextually nuanced reclaimed language, achieving robust performance even with severe class imbalance.

The paper proposes an Interpretive Audit Pipeline to evaluate LLMs for public comment analysis, arguing that measuring inter-model disagreement is crucial because standard accuracy metrics fail to det…

COFT is a training-free decoding method that significantly reduces societal biases in large language model chain-of-thought reasoning by applying token-level fairness control at decode time.

The paper introduces FBHM, a new benchmark for hateful memes, and proposes LSV, a steering vector method that significantly improves VLM performance by addressing the generalization gap.

The paper introduces the Triangulated Preference Shift score, an automated, curation-free metric to quantify systematic lexical biases introduced into Large Language Models during the preference-learn…

The paper introduces CERA, a novel contrastive retrieval framework that improves RAG factuality and interpretability by using subjectivity-based hard negative selection and an auxiliary attention alig…

The study found that human judgment of logical fallacies is significantly biased by source labels (e.g., human vs. AI), while LLM evaluations remained comparatively stable across these source conditio…

This paper investigates the production-evaluation gap in Large Reasoning Models (LRMs), finding that while LRMs excel at generating solutions, they struggle significantly to evaluate flawed reasoning,…

The paper evaluates LLM-generated reactions to Spanish online news, finding that off-the-shelf models fail to accurately reproduce the measurable properties of real audience discourse, and even fine-t…

This paper simulates the Argumentative Theory of Reasoning (ATR) using multi-agent debate among LLMs, demonstrating that collective adversarial discourse significantly enhances truth-seeking performan…

The paper introduces BiAxisAudit, a novel framework that evaluates LLM bias by analyzing bias scores across multiple prompt formats and within the internal inconsistency of model responses, revealing…

The paper introduces an Item Response Theory (IRT)-based indicator that effectively identifies likely mislabeled items in existing LLM benchmarks, revealing systematic errors in labeling and model spe…

The paper demonstrates that confronting Large Reasoning Models (LRMs) with conflicting objectives, such as contradictory choices or conflicting alignment values, significantly increases their vulnerab…

The paper introduces a Conflict-aware Penalty (CP) and Statistical Loss (SL) framework to stabilize and balance the training of multimodal sentiment analysis models, achieving state-of-the-art perform…

The paper addresses Perceptual Judgment Bias in multimodal LLM judges by introducing a new dataset and a unified training framework that forces models to prioritize visual evidence over plausible text…

The paper introduces a validated, consensus-labeled prompt bank that separates requests for executable malicious code (weapons) from requests for general harmful security knowledge, providing a more g…

The paper compares verbalized feature attributions and self-generated rationales for explaining model behavior, finding that the format and granularity of the explanation significantly affect its abil…

This paper evaluates the reliability of using Large Language Models (LLMs) as automated judges to assess the quality of other LLMs, finding a high correlation with human judgment when suitable prompts…

The paper decomposes LLM debate convergence into three mechanisms (instability, conformity, persuasion) and finds that much observed convergence is harmful social compliance rather than genuine reason…

The paper demonstrates that increasing the toxicity of prompts significantly degrades the factual reliability of LLMs, a degradation linked to the selective amplification of perturbation-sensitive nod…