Quantifying Faithful Confidence Expression in Large Reasoning Models

Reliable uncertainty communication is critical to the trustworthiness of LLMs, yet faithful calibration (FC)--the alignment between models' intrinsic and (linguistically) expressed confidence--is a persistent failure mode. This challenge is key for large reasoning models (LRMs), whose extended reasoning traces are often interpreted by users as evidence of deliberation, competence, and confidence. Despite the importance of FC and wide usage of LRMs, the extent to which LRMs can faithfully express their confidence remains poorly understood. Moreover, the prevailing paradigm to measure FC does not generalize well to the long chain-of-thought outputs generated by LRMs, which tend to lack clear step boundaries, involve inconsistent step structure, and encode complex conditional dependencies throughout the trace--complicating estimation of intrinsic confidence. To address this challenge, we introduce a novel framework to systematically quantify FC of LRMs. Our framework analyzes linguistic decisiveness relative to three sources of internal uncertainty, based on token probabilities, hidden states, and sampled response consistency. We also devise a prefix-conditioned sampling approach to control for conditional and structural variation across traces. Applying our framework to a diverse suite of leading models, datasets, and prompts, we find that faithful confidence expression is a significant challenge for LRMs. Reasoning behaviors do not automatically translate to improved FC, and prompt interventions for non-reasoning models do not improve faithfulness in the reasoning setting. Different confidence estimators further produce divergent assessments of the same traces, revealing fragility in prior evaluation methodologies. Taken together, our work establishes FC as a distinct reliability and alignment target for LRMs, particularly as such systems are increasingly deployed in high-stakes contexts.

Eye-Q: A Multilingual Benchmark for Visual Word Puzzle Solving and Image-to-Phrase Reasoning

Vision-Language Models (VLMs) have achieved strong performance on standard vision-language benchmarks, yet often rely on surface-level recognition rather than deeper reasoning. We propose visual word puzzles as a challenging alternative, as they require discovering implicit visual cues, generating and revising hypotheses, and mapping perceptual evidence to non-literal concepts in ways that are difficult to solve via literal grounding, OCR-heavy shortcuts, or simple retrieval-style matching. We introduce Eye-Q, a multilingual benchmark designed to assess this form of complex visual understanding. Eye-Q contains 1,343 puzzles in which a model observes a conceptually dense scene with a brief description and must infer a specific target word or phrase. The puzzles are intentionally unstructured and cue-implicit, with distractors and contextual relationships that demand selective attention, abstraction, and associative inference. The benchmark spans English, Persian, Arabic, and cross-lingual puzzles. We evaluate state-of-the-art VLMs using an open-ended, human-aligned protocol that probes hypothesis formation and revision under lightweight assistance. Results reveal substantial performance gaps, especially on abstract and cross-lingual puzzles, highlighting limitations in current models' ability to construct and search over appropriate conceptual representations for flexible image-to-phrase inference; maximum accuracy reaches only 60.27%.

FIRE: Faithful Interpretable Recommendation Explanations

Natural language explanations in recommender systems are often framed as a review generation task, leveraging user reviews as ground-truth supervision. While convenient, this approach conflates a user's opinion with the system's reasoning, leading to explanations that may be fluent but fail to reflect the true logic behind recommendations. In this work, we revisit the core objective of explainable recommendation: to transparently communicate why an item is recommended by linking user needs to relevant item features. Through a comprehensive analysis of existing methods across multiple benchmark datasets, we identify common limitations-explanations that are weakly aligned with model predictions, vague or inaccurate in identifying user intents, and overly repetitive or generic. To overcome these challenges, we propose FIRE, a lightweight and interpretable framework that combines SHAP-based feature attribution with structured, prompt-driven language generation. FIRE produces faithful, diverse, and user-aligned explanations, grounded in the actual decision-making process of the model. Our results demonstrate that FIRE not only achieves competitive recommendation accuracy but also significantly improves explanation quality along critical dimensions such as alignment, structure, and faithfulness. This work highlights the need to move beyond the review-as-explanation paradigm and toward explanation methods that are both accountable and interpretable.