Large language models (LLMs) have significantly advanced in reasoning tasks through reinforcement learning (RL) optimization, achieving impressive capabilities across various challenging benchmarks. However, our empirical analysis reveals a critical drawback: reasoning-oriented RL fine-tuning significantly increases the prevalence of hallucinations. We theoretically analyze the RL training dynamics, identifying high-variance gradient, entropy-induced randomness, and susceptibility to spurious local optima as key factors leading to hallucinations. To address this drawback, we propose Factuality-aware Step-wise Policy Optimization (FSPO), an innovative RL fine-tuning algorithm incorporating explicit factuality verification at each reasoning step. FSPO leverages automated verification against given evidence to dynamically adjust token-level advantage values, incentivizing factual correctness throughout the reasoning process. Experiments across mathematical reasoning and hallucination benchmarks using Qwen2.5 and Llama models demonstrate that FSPO effectively reduces hallucinations while enhancing reasoning accuracy, substantially improving both reliability and performance.

Retrieval-augmented generation (RAG) based large language models (LLMs) are widely used in finance for their excellent performance on knowledge-intensive tasks. However, standardized documents (e.g., SEC filing) share similar formats such as repetitive boilerplate texts, and similar table structures. This similarity forces traditional RAG methods to misidentify near-duplicate text, leading to duplicate retrieval that undermines accuracy and completeness. To address these issues, we propose the Hierarchical Retrieval with Evidence Curation (HiREC) framework. Our approach first performs hierarchical retrieval to reduce confusion among similar texts. It first retrieve related documents and then selects the most relevant passages from the documents. The evidence curation process removes irrelevant passages. When necessary, it automatically generates complementary queries to collect missing information. To evaluate our approach, we construct and release a Large-scale Open-domain Financial (LOFin) question answering benchmark that includes 145,897 SEC documents and 1,595 question-answer pairs. Our code and data are available at https://github.com/deep-over/LOFin-bench-HiREC.

Properly evaluating the ability of Video-Language Models (VLMs) to understand long videos remains a challenge. We propose a long-context video understanding benchmark, Causal2Needles, that assesses two crucial abilities insufficiently addressed by existing benchmarks: (1) extracting information from two separate locations (two needles) in a long video and understanding them jointly, and (2) modeling the world in terms of cause and effect in human behaviors. Causal2Needles evaluates these abilities using noncausal one-needle, causal one-needle, and causal two-needle questions. The most complex question type, causal two-needle questions, require extracting information from both the cause and effect events from a long video and the associated narration text. To prevent textual bias, we introduce two complementary question formats: locating the video clip containing the answer, and verbal description of a visual detail from that video clip. Our experiments reveal that models excelling on existing benchmarks struggle with causal 2-needle questions, and the model performance is negatively correlated with the distance between the two needles. These findings highlight critical limitations in current VLMs. The dataset is available at: https://huggingface.co/datasets/causal2needles/Causal2Needles

Large Language Models (LLMs) have emerged as a promising cornerstone for the development of natural language processing (NLP) and artificial intelligence (AI). However, ensuring the robustness of LLMs remains a critical challenge. To address these challenges and advance the field, this survey provides a comprehensive overview of current studies in this area. First, we systematically examine the nature of robustness in LLMs, including its conceptual foundations, the importance of consistent performance across diverse inputs, and the implications of failure modes in real-world applications. Next, we analyze the sources of non-robustness, categorizing intrinsic model limitations, data-driven vulnerabilities, and external adversarial factors that compromise reliability. Following this, we review state-of-the-art mitigation strategies, and then we discuss widely adopted benchmarks, emerging metrics, and persistent gaps in assessing real-world reliability. Finally, we synthesize findings from existing surveys and interdisciplinary studies to highlight trends, unresolved issues, and pathways for future research.

Objective: Large language models (LLMs) are attracting increasing interest in healthcare. This commentary evaluates the potential of LLMs to improve clinical prediction models (CPMs) for diagnostic and prognostic tasks, with a focus on their ability to process longitudinal electronic health record (EHR) data. Findings: LLMs show promise in handling multimodal and longitudinal EHR data and can support multi-outcome predictions for diverse health conditions. However, methodological, validation, infrastructural, and regulatory chal- lenges remain. These include inadequate methods for time-to-event modelling, poor calibration of predictions, limited external validation, and bias affecting underrepresented groups. High infrastructure costs and the absence of clear regulatory frameworks further prevent adoption. Implications: Further work and interdisciplinary collaboration are needed to support equitable and effective integra- tion into the clinical prediction. Developing temporally aware, fair, and explainable models should be a priority focus for transforming clinical prediction workflow.

Detecting subtle forms of dishonesty like sycophancy and manipulation in Large Language Models (LLMs) remains challenging for both humans and automated evaluators, as these behaviors often appear through small biases rather than clear false statements. We introduce Judge Using Safety-Steered Alternatives (JUSSA), a novel framework that employs steering vectors not to improve model behavior directly, but to enhance LLM judges' evaluation capabilities. JUSSA applies steering vectors during inference to generate more honest alternatives, providing judges with contrastive examples that make subtle dishonest patterns easier to detect. While existing evaluation methods rely on black-box evaluation, JUSSA leverages model internals to create targeted comparisons from single examples. We evaluate our method on sycophancy detection and introduce a new manipulation dataset covering multiple types of manipulation. Our results demonstrate that JUSSA effectively improves detection accuracy over single-response evaluation in various cases. Analysis across judge models reveals that JUSSA helps weaker judges on easier dishonesty detection tasks, and stronger judges on harder tasks. Layer-wise experiments show how dishonest prompts cause representations to diverge from honest ones in middle layers, revealing where steering interventions are most effective for generating contrastive examples. By demonstrating that steering vectors can enhance safety evaluation rather than just modify behavior, our work opens new directions for scalable model auditing as systems become increasingly sophisticated.

Recurrent neural networks (RNNs) trained on neuroscience-inspired tasks offer powerful models of brain computation. However, typical training paradigms rely on open-loop, supervised settings, whereas real-world learning unfolds in closed-loop environments. Here, we develop a mathematical theory describing the learning dynamics of linear RNNs trained in closed-loop contexts. We first demonstrate that two otherwise identical RNNs, trained in either closed- or open-loop modes, follow markedly different learning trajectories. To probe this divergence, we analytically characterize the closed-loop case, revealing distinct stages aligned with the evolution of the training loss. Specifically, we show that the learning dynamics of closed-loop RNNs, in contrast to open-loop ones, are governed by an interplay between two competing objectives: short-term policy improvement and long-term stability of the agent-environment interaction. Finally, we apply our framework to a realistic motor control task, highlighting its broader applicability. Taken together, our results underscore the importance of modeling closed-loop dynamics in a biologically plausible setting.

Traditional dialogue summarization primarily focuses on dialogue content, assuming it comprises adequate information for a clear summary. However, this assumption often fails for discussions grounded in shared background, where participants frequently omit context and use implicit references. This results in summaries that are confusing to readers unfamiliar with the background. To address this, we introduce Knowledge-Grounded Discussion Summarization (KGDS), a novel task that produces a supplementary background summary for context and a clear opinion summary with clarified references. To facilitate research, we construct the first KGDS benchmark, featuring news-discussion pairs and expert-created multi-granularity gold annotations for evaluating sub-summaries. We also propose a novel hierarchical evaluation framework with fine-grained and interpretable metrics. Our extensive evaluation of 12 advanced large language models (LLMs) reveals that KGDS remains a significant challenge. The models frequently miss key facts and retain irrelevant ones in background summarization, and often fail to resolve implicit references in opinion summary integration.

Retrieval-augmented generation (RAG) aims to reduce hallucinations by grounding responses in external context, yet large language models (LLMs) still frequently introduce unsupported information or contradictions even when provided with relevant context. This paper presents two complementary efforts at Vectara to measure and benchmark LLM faithfulness in RAG. First, we describe our original hallucination leaderboard, which has tracked hallucination rates for LLMs since 2023 using our HHEM hallucination detection model. Motivated by limitations observed in current hallucination detection methods, we introduce FaithJudge, an LLM-as-a-judge framework that leverages a pool of diverse human-annotated hallucination examples to substantially improve the automated hallucination evaluation of LLMs. We introduce an enhanced hallucination leaderboard centered on FaithJudge that benchmarks LLMs on RAG faithfulness in summarization, question-answering, and data-to-text generation tasks. FaithJudge enables a more reliable benchmarking of LLM hallucinations in RAG and supports the development of more trustworthy generative AI systems: https://github.com/vectara/FaithJudge.

Generating high-quality MCQs, especially those targeting diverse cognitive levels and incorporating common misconceptions into distractor design, is time-consuming and expertise-intensive, making manual creation impractical at scale. Current automated approaches typically generate questions at lower cognitive levels and fail to incorporate domain-specific misconceptions. This paper presents a hierarchical concept map-based framework that provides structured knowledge to guide LLMs in generating MCQs with distractors. We chose high-school physics as our test domain and began by developing a hierarchical concept map covering major Physics topics and their interconnections with an efficient database design. Next, through an automated pipeline, topic-relevant sections of these concept maps are retrieved to serve as a structured context for the LLM to generate questions and distractors that specifically target common misconceptions. Lastly, an automated validation is completed to ensure that the generated MCQs meet the requirements provided. We evaluate our framework against two baseline approaches: a base LLM and a RAG-based generation. We conducted expert evaluations and student assessments of the generated MCQs. Expert evaluation shows that our method significantly outperforms the baseline approaches, achieving a success rate of 75.20% in meeting all quality criteria compared to approximately 37% for both baseline methods. Student assessment data reveal that our concept map-driven approach achieved a significantly lower guess success rate of 28.05% compared to 37.10% for the baselines, indicating a more effective assessment of conceptual understanding. The results demonstrate that our concept map-based approach enables robust assessment across cognitive levels and instant identification of conceptual gaps, facilitating faster feedback loops and targeted interventions at scale.