Read Your Own Mind: Reasoning Helps Surface Self-Confidence Signals in LLMs

We study the source of uncertainty in DeepSeek R1-32B by analyzing its self-reported verbal confidence on question answering (QA) tasks. In the default answer-then-confidence setting, the model is regularly over-confident, whereas semantic entropy - obtained by sampling many responses - remains reliable. We hypothesize that this is because of semantic entropy's larger test-time compute, which lets us explore the model's predictive distribution. We show that granting DeepSeek the budget to explore its distribution by forcing a long chain-of-thought before the final answer greatly improves its verbal score effectiveness, even on simple fact-retrieval questions that normally require no reasoning. Furthermore, a separate reader model that sees only the chain can reconstruct very similar confidences, indicating the verbal score might be merely a statistic of the alternatives surfaced during reasoning. Our analysis concludes that reliable uncertainty estimation requires explicit exploration of the generative space, and self-reported confidence is trustworthy only after such exploration.

Beyond Reproducibility: Advancing Zero-shot LLM Reranking Efficiency with Setwise Insertion

This study presents a comprehensive reproducibility and extension analysis of the Setwise prompting methodology for zero-shot ranking with Large Language Models (LLMs), as proposed by Zhuang et al. We evaluate its effectiveness and efficiency compared to traditional Pointwise, Pairwise, and Listwise approaches in document ranking tasks. Our reproduction confirms the findings of Zhuang et al., highlighting the trade-offs between computational efficiency and ranking effectiveness in Setwise methods. Building on these insights, we introduce Setwise Insertion, a novel approach that leverages the initial document ranking as prior knowledge, reducing unnecessary comparisons and uncertainty by focusing on candidates more likely to improve the ranking results. Experimental results across multiple LLM architectures (Flan-T5, Vicuna, and Llama) show that Setwise Insertion yields a 31% reduction in query time, a 23% reduction in model inferences, and a slight improvement in reranking effectiveness compared to the original Setwise method. These findings highlight the practical advantage of incorporating prior ranking knowledge into Setwise prompting for efficient and accurate zero-shot document reranking.

Gandalf the Red: Adaptive Security for LLMs

Current evaluations of defenses against prompt attacks in large language model (LLM) applications often overlook two critical factors: the dynamic nature of adversarial behavior and the usability penalties imposed on legitimate users by restrictive defenses. We propose D-SEC (Dynamic Security Utility Threat Model), which explicitly separates attackers from legitimate users, models multi-step interactions, and rigorously expresses the security-utility in an optimizable form. We further address the shortcomings in existing evaluations by introducing Gandalf, a crowd-sourced, gamified red-teaming platform designed to generate realistic, adaptive attack datasets. Using Gandalf, we collect and release a dataset of 279k prompt attacks. Complemented by benign user data, our analysis reveals the interplay between security and utility, showing that defenses integrated in the LLM (e.g., system prompts) can degrade usability even without blocking requests. We demonstrate that restricted application domains, defense-in-depth, and adaptive defenses are effective strategies for building secure and useful LLM applications. Code is available at \href{https://github.com/lakeraai/dsec-gandalf}{\texttt{https://github.com/lakeraai/dsec-gandalf}}.

Contrastive News and Social Media Linking using BERT for Articles and Tweets across Dual Platforms

X (formerly Twitter) has evolved into a contemporary agora, offering a platform for individuals to express opinions and viewpoints on current events. The majority of the topics discussed on Twitter are directly related to ongoing events, making it an important source for monitoring public discourse. However, linking tweets to specific news presents a significant challenge due to their concise and informal nature. Previous approaches, including topic models, graph-based models, and supervised classifiers, have fallen short in effectively capturing the unique characteristics of tweets and articles. Inspired by the success of the CLIP model in computer vision, which employs contrastive learning to model similarities between images and captions, this paper introduces a contrastive learning approach for training a representation space where linked articles and tweets exhibit proximity. We present our contrastive learning approach, CATBERT (Contrastive Articles Tweets BERT), leveraging pre-trained BERT models. The model is trained and tested on a dataset containing manually labeled English and Polish tweets and articles related to the Russian-Ukrainian war. We evaluate CATBERT's performance against traditional approaches like LDA, and the novel method based on OpenAI embeddings, which has not been previously applied to this task. Our findings indicate that CATBERT demonstrates superior performance in associating tweets with relevant news articles. Furthermore, we demonstrate the performance of the models when applied to finding the main topic -- represented by an article -- of the whole cascade of tweets. In this new task, we report the performance of the different models in dependence on the cascade size.