Large Language Models (LLMs) often generate outputs that lack grounding in real-world facts, a phenomenon known as hallucinations. Prior research has associated hallucinations with model uncertainty, leveraging this relationship for hallucination detection and mitigation. In this paper, we challenge the underlying assumption that all hallucinations are associated with uncertainty. Using knowledge detection and uncertainty measurement methods, Figure 1: Do high-certainty hallucinations exist? An we demonstrate that models can hallucinate illustrative categorization of hallucinations based on a with high certainty even when they have the model's knowledge and certainty. Highlighted is the correct knowledge. We further show that highcertainty phenomenon of high-certainty hallucinations (purple) hallucinations are consistent across - where models confidently produce incorrect outputs, models and datasets, distinctive enough to be even when they have the correct knowledge. While other singled out, and challenge existing mitigation types of hallucinations can potentially be explained by methods. Our findings reveal an overlooked aspect the model not knowing, being mistaken, or uncertain, of hallucinations, emphasizing the need to high-certainty hallucinations are harder to rationalize, understand their origins and improve mitigation making their existence particularly intriguing.

The reasoning abilities of Large Language Models (LLMs) have demonstrated remarkable advancement and exceptional performance across diverse domains. However, leveraging these reasoning capabilities to enhance LLM safety against adversarial attacks and jailbreak queries remains largely unexplored. To bridge this gap, we propose Reasoning-to-Defend (R2D), a novel training paradigm that integrates safety reflections of queries and responses into LLMs' generation process, unlocking a safety-aware reasoning mechanism. This approach enables self-evaluation at each reasoning step to create safety pivot tokens as indicators of the response's safety status. Furthermore, in order to improve the learning efficiency of pivot token prediction, we propose Contrastive Pivot Optimization(CPO), which enhances the model's ability to perceive the safety status of dialogues. Through this mechanism, LLMs dynamically adjust their response strategies during reasoning, significantly enhancing their defense capabilities against jailbreak attacks. Extensive experimental results demonstrate that R2D effectively mitigates various attacks and improves overall safety, highlighting the substantial potential of safety-aware reasoning in strengthening LLMs' robustness against jailbreaks.

Sailor2 is a family of cutting-edge multilingual language models for South-East Asian (SEA) languages, available in 1B, 8B, and 20B sizes to suit diverse applications. Building on Qwen2.5, Sailor2 undergoes continuous pre-training on 500B tokens (400B SEA-specific and 100B replay tokens) to support 13 SEA languages while retaining proficiency in Chinese and English. Sailor2-20B model achieves a 50-50 win rate against GPT-4o across SEA languages. We also deliver a comprehensive cookbook on how to develop the multilingual model in an efficient manner, including five key aspects: data curation, pre-training, post-training, model customization and evaluation. We hope that Sailor2 model (Apache 2.0 license) will drive language development in the SEA region, and Sailor2 cookbook will inspire researchers to build more inclusive LLMs for other under-served languages.

This work studies the applicability of expensive external oracles such as large language models in answering top-k queries over predicted scores. Such scores are incurred by user-defined functions to answer personalized queries over multi-modal data. We propose a generic computational framework that handles arbitrary set-based scoring functions, as long as the functions could be decomposed into constructs, each of which sent to an oracle (in our case an LLM) to predict partial scores. At a given point in time, the framework assumes a set of responses and their partial predicted scores, and it maintains a collection of possible sets that are likely to be the true top-k. Since calling oracles is costly, our framework judiciously identifies the next construct, i.e., the next best question to ask the oracle so as to maximize the likelihood of identifying the true top-k. We present a principled probabilistic model that quantifies that likelihood. We study efficiency opportunities in designing algorithms. We run an evaluation with three large scale datasets, scoring functions, and baselines. Experiments indicate the efficacy of our framework, as it achieves an order of magnitude improvement over baselines in requiring LLM calls while ensuring result accuracy. Scalability experiments further indicate that our framework could be used in large-scale applications.

Meeting summarization suffers from limited high-quality data, mainly due to privacy restrictions and expensive collection processes. We address this gap with FAME, a dataset of 500 meetings in English and 300 in German produced by MIMIC, our new multi-agent meeting synthesis framework that generates meeting transcripts on a given knowledge source by defining psychologically grounded participant profiles, outlining the conversation, and orchestrating a large language model (LLM) debate. A modular post-processing step refines these outputs, mitigating potential repetitiveness and overly formal tones, ensuring coherent, credible dialogues at scale. We also propose a psychologically grounded evaluation framework assessing naturalness, social behavior authenticity, and transcript difficulties. Human assessments show that FAME approximates real-meeting spontaneity (4.5/5 in naturalness), preserves speaker-centric challenges (3/5 in spoken language), and introduces richer information-oriented difficulty (4/5 in difficulty). These findings highlight that FAME is a good and scalable proxy for real-world meeting conditions. It enables new test scenarios for meeting summarization research and other conversation-centric applications in tasks requiring conversation data or simulating social scenarios under behavioral constraints.

Objective speech quality models aim to predict human-perceived speech quality using automated methods. However, cross-lingual generalization remains a major challenge, as Mean Opinion Scores (MOS) vary across languages due to linguistic, perceptual, and dataset-specific differences. A model trained primarily on English data may struggle to generalize to languages with different phonetic, tonal, and prosodic characteristics, leading to inconsistencies in objective assessments. This study investigates the cross-lingual performance of two speech quality models: NISQA, a CNN-based model, and a Transformer-based Audio Spectrogram Transformer (AST) model. Both models were trained exclusively on English datasets containing over 49,000 speech samples and subsequently evaluated on speech in German, French, Mandarin, Swedish, and Dutch. We analyze model performance using Pearson Correlation Coefficient (PCC) and Root Mean Square Error (RMSE) across five speech quality dimensions: coloration, discontinuity, loudness, noise, and MOS. Our findings show that while AST achieves a more stable cross-lingual performance, both models exhibit noticeable biases. Notably, Mandarin speech quality predictions correlate highly with human MOS scores, whereas Swedish and Dutch present greater prediction challenges. Discontinuities remain difficult to model across all languages. These results highlight the need for more balanced multilingual datasets and architecture-specific adaptations to improve cross-lingual generalization.

Large Language Models (LLMs) have significantly advanced medical question-answering by leveraging extensive clinical data and medical literature. However, the rapid evolution of medical knowledge and the labor-intensive process of manually updating domain-specific resources pose challenges to the reliability of these systems. To address this, we introduce Adaptive Medical Graph-RAG (AMG-RAG), a comprehensive framework that automates the construction and continuous updating of medical knowledge graphs, integrates reasoning, and retrieves current external evidence, such as PubMed and WikiSearch. By dynamically linking new findings and complex medical concepts, AMG-RAG not only improves accuracy but also enhances interpretability in medical queries. Evaluations on the MEDQA and MEDMCQA benchmarks demonstrate the effectiveness of AMG-RAG, achieving an F1 score of 74.1 percent on MEDQA and an accuracy of 66.34 percent on MEDMCQA, outperforming both comparable models and those 10 to 100 times larger. Notably, these improvements are achieved without increasing computational overhead, highlighting the critical role of automated knowledge graph generation and external evidence retrieval in delivering up-to-date, trustworthy medical insights.

Role-Playing Agent (RPA) is an increasingly popular type of LLM Agent that simulates human-like behaviors in a variety of tasks. However, evaluating RPAs is challenging due to diverse task requirements and agent designs. This paper proposes an evidence-based, actionable, and generalizable evaluation design guideline for LLM-based RPA by systematically reviewing 1,676 papers published between Jan. 2021 and Dec. 2024. Our analysis identifies six agent attributes, seven task attributes, and seven evaluation metrics from existing literature. Based on these findings, we present an RPA evaluation design guideline to help researchers develop more systematic and consistent evaluation methods.

With the power of LLMs, we now have the ability to query data that was previously impossible to query, including text, images, and video. However, despite this enormous potential, most present-day data systems that leverage LLMs are reactive, reflecting our community's desire to map LLMs to known abstractions. Most data systems treat LLMs as an opaque black box that operates on user inputs and data as is, optimizing them much like any other approximate, expensive UDFs, in conjunction with other relational operators. Such data systems do as they are told, but fail to understand and leverage what the LLM is being asked to do (i.e. the underlying operations, which may be error-prone), the data the LLM is operating on (e.g., long, complex documents), or what the user really needs. They don't take advantage of the characteristics of the operations and/or the data at hand, or ensure correctness of results when there are imprecisions and ambiguities. We argue that data systems instead need to be proactive: they need to be given more agency -- armed with the power of LLMs -- to understand and rework the user inputs and the data and to make decisions on how the operations and the data should be represented and processed. By allowing the data system to parse, rewrite, and decompose user inputs and data, or to interact with the user in ways that go beyond the standard single-shot query-result paradigm, the data system is able to address user needs more efficiently and effectively. These new capabilities lead to a rich design space where the data system takes more initiative: they are empowered to perform optimization based on the transformation operations, data characteristics, and user intent. We discuss various successful examples of how this framework has been and can be applied in real-world tasks, and present future directions for this ambitious research agenda.

Despite the remarkable capabilities of Large Language Models (LLMs) in various NLP tasks, they remain vulnerable to hallucinations due to their limited parametric knowledge and lack of domain-specific expertise. Retrieval-Augmented Generation (RAG) addresses this challenge by incorporating external document retrieval to augment the knowledge base of LLMs. In this approach, RAG retrieves document chunks from an external corpus in response to a query, which are then used as context for the downstream language model to generate an answer. However, these retrieved knowledge sources often include irrelevant or erroneous information, undermining the effectiveness of RAG in downstream tasks. To overcome this limitation, we introduce a compact, efficient, and pluggable module designed to refine external knowledge sources before feeding them to the generator. The module reconstructs retrieved content by extracting the most relevant and supportive information and reorganising it into a concise, query-specific format. Through a three-stage training paradigm - comprising supervised fine-tuning, contrastive multi-task learning, and reinforcement learning-based alignment - it prioritises critical knowledge and aligns it with the generator's preferences. This method enables LLMs to produce outputs that are more accurate, reliable, and contextually appropriate.