Role-playing large language models (LLMs) are expected to adopt a character’s style while also respecting that character’s knowledge boundaries. Prior evaluations detect character hallucination but rarely distinguish whether errors arise from failure to recognize a boundary or from failure to comply despite recognition. We introduce CHARM, a multicultural benchmark of 40 real and fictional characters drawn from five cultural-linguistic regions, and validated by native reviewers. It probes two boundary types, Temporal (historical vs. modern) and Cross-Universe (entities outside a character’s narrative or historical universe), using abstention-enabled multiple-choice questions. We propose a two-stage evaluation that separates Boundary-Awareness (explicit recognition that a query is out of scope) from Boundary-Compliance (abstention when answering concrete questions). Evaluations across six LLMs show that hallucination is driven predominantly by compliance failures. Models frequently acknowledge that a query lies outside the character’s knowledge yet still provide factual, out-of-character answers. By re-posing the same questions to the target character, we confirm that a large fraction of these cases are verified parametric overrides; the model stores the relevant fact but fails to suppress it. We also observe systematic cultural variation in these failures, consistent with imbalances in how characters from different regions are represented in model knowledge.

Multilingual Large Language Models (MLLMs) are increasingly expected to handle Code-Switched (CS) inputs, yet mixing languages frequently degrades performance relative to source- or target-language monolingual counterparts. To understand this degradation, we use grammar-forced CS as a controlled diagnostic setting for locating CS representations relative to their source and target counterparts. We introduce Anchor Bias, a geometric measure that quantifies language anchoring—whether a CS hidden state aligns closer to its source or target language counterpart. Across diverse MLLMs, Anchor Bias reveals a consistent grammar-frame effect: source-framed CS stays source-anchored, whereas target-framed CS shifts target-ward and shows larger Question Answering (QA) degradation. Motivated by this representational pattern, we propose CANVAS (Contextual Anchor-based Neural Vector Alignment Steering), an inference-time intervention that extracts a source-side canvas from the input and softly steers target-language hidden states toward the source anchor during prefill. CANVAS consistently recovers QA F1 across MLLMs and CS conditions, showing that internal anchoring signals provide an actionable target for mitigating CS inference failures.

Video Multimodal Large Language Models (VideoMLLMs) have achieved remarkable progress in both Video-to-Text and Text-to-Video tasks. However, they often suffer from hallucinations, generating content that contradicts the visual input. Existing evaluation methods are limited to one task (V2T) and also fail to assess hallucinations in open-ended, free-form responses. To address this gap, we propose FIFA, a unified FaIthFulness evAluation framework that extracts comprehensive descriptive facts, models their semantic dependencies via a Spatio-Temporal Semantic Dependency Graph, and verifies them using VideoQA models. We further introduce , a tool-based correction framework that revises hallucinated content. Extensive experiments demonstrate that FIFA aligns more closely with human judgment than existing evaluation methods, and that   effectively improves factual consistency in both text and video generation.

Real-world multi-hop QA is naturally linked with ambiguity, where a single query can trigger multiple reasoning paths that require independent resolution. Since ambiguity can occur at any stage, models must navigate layered uncertainty throughout the entire reasoning chain. Despite its prevalence in real-world user queries, previous benchmarks have primarily focused on single-hop ambiguity, leaving the complex interaction between multi-step inference and layered ambiguity underexplored. In this paper, we introduce MARCH, a benchmark for their intersection, with 2,209 multi-hop ambiguous questions curated via multi-LLM verification and validated by human annotation with strong agreement. Our experiments reveal that even state-of-the-art models struggle with MARCH, confirming that combining ambiguity resolution with multi-step reasoning is a significant challenge. To address this, we propose CLARION, a two-stage agentic framework that explicitly decouples ambiguity planning from evidence-driven reasoning, significantly outperforms existing approaches, and paves the way for robust reasoning systems.

This paper describes VILLAIN, a multimodal fact-checking system that verifies image-text claims through prompt-based multi-agent collaboration. For the AVerImaTeC shared task, VILLAIN employs vision-language model agents across multiple stages of fact-checking. Textual and visual evidence is retrieved from the knowledge store enriched through additional web collection. To identify key information and address inconsistencies among evidence items, modality-specific and cross-modal agents generate analysis reports. In the subsequent stage, question-answer pairs are produced based on these reports. Finally, the Verdict Prediction agent produces the verification outcome based on the image-text claim and the generated question-answer pairs. Our system ranked first on the leaderboard across all evaluation metrics. The source code is publicly available at https://github.com/ssu-humane/VILLAIN.

Instruction tuning is a widely used approach to improve the instruction-following ability of large language models (LLMs). Instruction-tuning datasets typically include a mixture of context-augmented and context-free examples, yet prior work has largely combined these data types without examining their distinct effects. In this paper, we investigate how training LLMs with or without context affects model behavior and downstream performance. First, in the text domain, we show that LLMs trained with context attend more strongly to the provided knowledge, achieving better grounding. We also observe that context-augmented training shifts how LLMs use knowledge: models store and leverage less on parametric knowledge and instead depend more on the provided context. Second, we observe that using LLM trained with context-augmented data as the backbone for vision-language models reduces hallucination and improves grounding in the visual domain. Finally, we explore practical strategies for real-world deployments where context availability varies. We show that maintaining separate context-augmented and context-free models and routing inputs between them yields more robust overall performance than training a single mixed model, as it better preserves their complementary strengths.

User simulation has long played a vital role in computer science due to its potential to support a wide range of applications. Language, as the primary medium of human communication, forms the foundation of social interaction and behavior. Consequently, simulating conversational behavior has become a key area of study. Recent advancements in large language models (LLMs) have significantly catalyzed progress in this domain by enabling high-fidelity generation of synthetic user conversation. In this paper, we survey recent advancements in LLM-based conversational user simulation. We introduce a novel taxonomy covering user granularity and simulation objectives. Additionally, we systematically analyze core techniques and evaluation methodologies. We aim to keep the research community informed of the latest advancements in conversational user simulation and to further facilitate future research by identifying open challenges and organizing existing work under a unified framework.

LLMs operating in dynamic real-world contexts often encounter knowledge that evolves continuously or emerges incrementally. To remain accurate and effective, models must adapt to newly arriving information on the fly. We introduce Online Adaptation to Continual Knowledge Streams(OAKS) to evaluate this capability, establishing a benchmark for online adaptation over streaming, continually updating knowledge. Specifically, the benchmark is structured as a sequence of fine-grained context chunks where facts change dynamically across time intervals. OAKS comprises two datasets: OAKS-BABI and OAKS-Novel, where individual facts evolve multiple times across context chunks. These datasets include dense annotations to measure whether models track changes accurately. Evaluating 14 models with varied inference approaches, we observe significant limitations in current methodologies. Both state-of-the-art models and agentic memory systems fail to adapt robustly on OAKS, demonstrating delays in state-tracking and susceptibility to distraction within streaming environments. We will open-source the code and datasets.

Inductive reasoning — the process of inferring general rules from a small number of observations — is a fundamental aspect of human intelligence. Recent works suggest that large language models (LLMs) can engage in inductive reasoning by sampling multiple hypotheses about the rules and selecting the one that best explains the observations. However, due to the IID sampling, semantically redundant hypotheses are frequently generated, leading to significant wastage of compute. In this paper, we 1) demonstrate that increasing the temperature to enhance the diversity is limited due to text degeneration issue, and 2) propose a novel method to improve the diversity while maintaining text quality. We first analyze the effect of increasing the temperature parameter, which is regarded as the LLM’s diversity control, on IID hypotheses. Our analysis shows that as temperature rises, diversity and accuracy of hypotheses increase up to a certain point, but this trend saturates due to text degeneration. To generate hypotheses that are more semantically diverse and of higher quality, we propose a novel approach inspired by human inductive reasoning, which we call Mixture of Concepts (MoC). When applied to several inductive reasoning benchmarks, MoC demonstrated significant performance improvements compared to standard IID sampling and other approaches.

In a real-world corpus, knowledge frequently recurs across documents but often contains inconsistencies due to ambiguous naming, outdated information, or errors, leading to complex interrelationships between contexts. Previous research has shown that language models struggle with these complexities, typically focusing on single factors in isolation. We classify these relationships into four types: distracting, ambiguous, counterfactual, and duplicated. Our analysis reveals that no single approach effectively addresses all these interrelationships simultaneously. Therefore, we introduce Context Organizer (COrg), a framework that organizes multiple contexts into independently processed groups. This design allows the model to efficiently find all relevant answers while ensuring disambiguation. COrg consists of three key components: a graph constructor, a reranker, and an aggregator. Our results demonstrate that COrg balances performance and efficiency effectively, outperforming existing grouping methods and achieving comparable results to more computationally intensive, single-context approaches.

Large Vision-Language Models (LVLMs) have demonstrated outstanding performance across various multimodal tasks. However, they suffer from a problem known as language prior, where responses are generated based solely on textual patterns while disregarding image information. Addressing the issue of language prior is crucial, as it can lead to undesirable biases or hallucinations when dealing with images that are out of training distribution. Despite its importance, current methods for accurately measuring language priors in LVLMs are poorly studied. Although existing benchmarks based on counterfactual or out-of-distribution images can partially be used to measure language priors, they fail to disentangle language priors from other confounding factors. To this end, we propose a new benchmark called VLind-Bench, which is the first benchmark specifically designed to measure the language priors, or blindness, of LVLMs. It not only includes tests on counterfactual images to assess language priors but also involves a series of tests to evaluate more basic capabilities such as commonsense knowledge, visual perception, and commonsense biases. For each instance in our benchmark, we ensure that all these basic tests are passed before evaluating the language priors, thereby minimizing the influence of other factors on the assessment. The evaluation and analysis of recent LVLMs in our benchmark reveal that almost all models exhibit a significant reliance on language priors, presenting a strong challenge in the field.

Query expansion methods powered by large language models (LLMs) have demonstrated effectiveness in zero-shot retrieval tasks. These methods assume that LLMs can generate hypothetical documents that, when incorporated into a query vector, enhance the retrieval of real evidence. However, we challenge this assumption by investigating whether knowledge leakage in benchmarks contributes to the observed performance gains. Using fact verification as a testbed, we analyze whether the generated documents contain information entailed by ground-truth evidence and assess their impact on performance. Our findings indicate that, on average, performance improvements consistently occurred for claims whose generated documents included sentences entailed by gold evidence. This suggests that knowledge leakage may be present in fact-verification benchmarks, potentially inflating the perceived performance of LLM-based query expansion methods.

Graphical User Interface (GUI) agents, powered by Large Foundation Models, have emerged as a transformative approach to automating human-computer interaction. These agents autonomously interact with digital systems via GUIs, emulating human actions such as clicking, typing, and navigating visual elements across diverse platforms. Motivated by the growing interest and fundamental importance of GUI agents, we provide a comprehensive survey that categorizes their benchmarks, evaluation metrics, architectures, and training methods. We propose a unified framework that delineates their perception, reasoning, planning, and acting capabilities. Furthermore, we identify important open challenges and discuss key future directions. Finally, this work serves as a basis for practitioners and researchers to gain an intuitive understanding of current progress, techniques, benchmarks, and critical open problems that remain to be addressed.

This paper presents HerO 2, Team HUMANE’s system for the AVeriTeC shared task at the FEVER-25 workshop. HerO 2 is an enhanced version of HerO, the best-performing open-source model from the previous year’s challenge. It improves evidence quality through document summarization and answer reformulation, optimizes veracity prediction via post-training quantization under computational constraints, and enhances overall system performance by integrating updated language model (LM) backbones. HerO 2 ranked second on the leaderboard while achieving the shortest runtime among the top three systems, demonstrating both high efficiency and strong potential for real-world fact verification. The code is available at https://github.com/ssu-humane/HerO2.

Various studies have attempted to remove sensitive or private knowledge from a language model to prevent its unauthorized exposure. However, prior studies have overlooked the inherent complexity and interconnectedness of knowledge, which requires careful examination. To resolve this problem, we first define a new concept called superficial unlearning, which refers to the phenomenon where an unlearning method either fails to erase the interconnected knowledge it should remove or unintentionally erases irrelevant knowledge. Based on the definition, we introduce a novel benchmark, FaithUn, to analyze and evaluate the faithfulness of unlearning in real-world knowledge QA settings. Furthermore, we propose a novel unlearning method, KLUE, which updates only knowledge-related neurons to achieve faithful unlearning. KLUE leverages a regularized explainability method to localize contextual knowledge neurons, updating only these neurons using carefully selected unforgotten samples. Experimental results demonstrate that existing unlearning methods fail to ensure faithful unlearning, while our method shows significant effectiveness in real-world QA unlearning.

Active Learning (AL) has been a powerful paradigm for improving model efficiency and performance by selecting the most informative data points for labeling and training. In recent active learning frameworks, Large Language Models (LLMs) have been employed not only for selection but also for generating entirely new data instances and providing more cost-effective annotations. Motivated by the increasing importance of high-quality data and efficient model training in the era of LLMs, we present a comprehensive survey on LLM-based Active Learning. We introduce an intuitive taxonomy that categorizes these techniques and discuss the transformative roles LLMs can play in the active learning loop. We further examine the impact of AL on LLM learning paradigms and its applications across various domains. Finally, we identify open challenges and propose future research directions. This survey aims to serve as an up-to-date resource for researchers and practitioners seeking to gain an intuitive understanding of LLM-based AL techniques and deploy them to new applications.

While small language models (SLMs) show promises for mobile deployment, their real world performance and applications on smartphones remain underexplored. We present SlimLM, a series of SLMs optimized for document assistance tasks on mobile devices. Through extensive experiments on a Samsung Galaxy S24, we identify the sweet spot between model size (ranging from 125M to 8B parameters), context length, and inference time for efficient on-device processing. SlimLM is pretrained on SlimPajama-627B and fine-tuned on DocAssist, our constructed dataset for summarization, question answering, and suggestion tasks. Our smallest model demonstrates efficient performance on S24, while larger variants offer enhanced capabilities within mobile constraints. We evaluate SlimLM against existing SLMs, showing comparable or superior performance and offering a benchmark for future research in on-device language models. We provide an Android application allowing users to experience SlimLM’s document assistance capabilities, offering valuable insights for mobile developers, researchers, and companies seeking privacy-preserving on-device alternatives to server-based language models.

Question-answering for domain-specific applications has recently attracted much interest due to the latest advancements in large language models (LLMs). However, accurately assessing the performance of these applications remains a challenge, mainly due to the lack of suitable benchmarks that effectively simulate real-world scenarios. To address this challenge, we introduce two product question-answering (QA) datasets focused on Adobe Acrobat and Photoshop products to help evaluate the performance of existing models on domain-specific product QA tasks. Additionally, we propose a novel knowledge-driven RAG-QA framework to enhance the performance of the models in the product QA task. Our experiments demonstrated that inducing domain knowledge through query reformulation allowed for increased retrieval and generative performance when compared to standard RAG-QA methods. This improvement, however, is slight, and thus illustrates the challenge posed by the datasets introduced.

Application of LLM to database queries on natural language sentences has demonstrated impressive results in both single and multi-hop scenarios.In the existing methodologies, the requirement to re-encode query vectors at each stage for processing multi-hop queries presents a significant bottleneck to the inference speed.This paper proposes VKGFR (Virtual Knowledge Graph based Fact Retriever) that leverages large language models to extract representations corresponding to a sentence’s knowledge graph, significantly enhancing inference speed for multi-hop reasoning without performance loss.Given that both the queries and natural language database sentences can be structured as a knowledge graph, we suggest extracting a Virtual Knowledge Graph (VKG) representation from sentences with LLM.Over the pre-constructed VKG, our VKGFR conducts retrieval with a tiny model structure, showing performance improvements with higher computational efficiency. We evaluate VKGFR on the WikiNLDB and MetaQA dataset, designed for multi-hop database reasoning over text. The results indicate 13x faster inference speed on the WikiNLDB dataset without performance loss.

Contrastive language-image pre-training (CLIP) models have demonstrated considerable success across various vision-language tasks, such as text-to-image retrieval, where the model is required to effectively process natural language input to produce an accurate visual output. However, current models still face limitations in dealing with linguistic variations in input queries, such as paraphrases, making it challenging to handle a broad range of user queries in real-world applications. In this study, we introduce a straightforward fine-tuning approach to enhance the representations of CLIP models for paraphrases. Our approach involves a two-step paraphrase generation process, where we automatically create two categories of paraphrases from web-scale image captions by leveraging large language models. Subsequently, we fine-tune the CLIP text encoder using these generated paraphrases while freezing the image encoder. Our resulting model, which we call ParaCLIP, exhibits significant improvements over baseline CLIP models across various tasks, including paraphrased retrieval (with rank similarity scores improved by up to 7.6% and 9.6%), Visual Genome Relation and Attribution, as well as seven semantic textual similarity tasks.

This paper addresses the critical challenge of assessing the representativeness of news thumbnail images, which often serve as the first visual engagement for readers when an article is disseminated on social media. We focus on whether a news image represents the actors discussed in the news text. To serve the challenge, we introduce NewsTT, a manually annotated dataset of 1000 news thumbnail images and text pairs. We found that the pretrained vision and language models, such as BLIP-2, struggle with this task. Since news subjects frequently involve named entities or proper nouns, the pretrained models could have a limited capability to match news actors’ visual and textual appearances. We hypothesize that learning to contrast news text with its counterfactual, of which named entities are replaced, can enhance the cross-modal matching ability of vision and language models. We propose CFT-CLIP, a contrastive learning framework that updates vision and language bi-encoders according to the hypothesis. We found that our simple method can boost the performance for assessing news thumbnail representativeness, supporting our assumption. Code and data can be accessed at https://github.com/ssu-humane/news-images-acl24.

To tackle the AVeriTeC shared task hosted by the FEVER-24, we introduce a system that only employs publicly available large language models (LLMs) for each step of automated fact-checking, dubbed the Herd of Open LLMs for verifying real-world claims (HerO). HerO employs multiple LLMs for each step of automated fact-checking. For evidence retrieval, a language model is used to enhance a query by generating hypothetical documents that check the veracity of a claim. We fine-tune LLMs for question generation and veracity prediction by crafting prompts with retrieved in-context samples. HerO achieved 2nd place on the leaderboard with the AVeriTeC score of 0.57, suggesting the potential of open LLMs for verifying real-world claims. For future research, we make our code publicly available at https://github.com/ssu-humane/HerO.

Through the advent of pre-trained language models, there have been notable advancements in abstractive summarization systems. Simultaneously, a considerable number of novel methods for evaluating factual consistency in abstractive summarization systems has been developed. But these evaluation approaches incorporate substantial limitations, especially on refinement and interpretability. In this work, we propose highly effective and interpretable factual inconsistency detection method FIZZ (Factual Inconsistency Detection by Zoom-in Summary and Zoom-out Document) for abstractive summarization systems that is based on fine-grained atomic facts decomposition. Moreover, we align atomic facts decomposed from the summary with the source document through adaptive granularity expansion. These atomic facts represent a more fine-grained unit of information, facilitating detailed understanding and interpretability of the summary’s factual inconsistency. Experimental results demonstrate that our proposed factual consistency checking system significantly outperforms existing systems. We release the code at https://github.com/plm3332/FIZZ.

Extractive summarization plays a pivotal role in natural language processing due to its wide-range applications in summarizing diverse content efficiently, while also being faithful to the original content. Despite significant advancement achieved in extractive summarization by Large Language Models (LLMs), these summaries frequently exhibit incoherence. An important aspect of the coherent summary is its readability for intended users. Although there have been many datasets and benchmarks proposed for creating coherent extractive summaries, none of them currently incorporate user intent to improve coherence in extractive summarization. Motivated by this, we propose a systematically created human-annotated dataset consisting of coherent summaries for five publicly available datasets and natural language user feedback, offering valuable insights into how to improve coherence in extractive summaries. We utilize this dataset for aligning LLMs through supervised fine-tuning with natural language human feedback to enhance the coherence of their generated summaries. Preliminary experiments with Falcon-40B and Llama-2-13B show significant performance improvements (~10% Rouge-L) in terms of producing coherent summaries. We further utilize human feedback to benchmark results over instruction-tuned models such as FLAN-T5 which resulted in several interesting findings.

Large Language Models (LLMs) have issues with document question answering (QA) in situations where the document is unable to fit in the small context length of an LLM. To overcome this issue, most existing works focus on retrieving the relevant context from the document, representing them as plain text. However, documents such as PDFs, web pages, and presentations are naturally structured with different pages, tables, sections, and so on. Representing such structured documents as plain text is incongruous with the user’s mental model of these documents with rich structure. When a system has to query the document for context, this incongruity is brought to the fore, and seemingly trivial questions can trip up the QA system. To bridge this fundamental gap in handling structured documents, we propose an approach called PDFTriage that enables models to retrieve the context based on either structure or content. Our experiments demonstrate the effectiveness of the proposed PDFTriage-augmented models across several classes of questions where existing retrieval-augmented LLMs fail. To facilitate further research on this fundamental problem, we release our benchmark dataset consisting of 900+ human-generated questions over 80 structured documents from 10 different categories of question types for document QA. Our code and datasets will be released soon on Github.

With the ubiquitous use of online meeting platforms and robust automatic speech recognition systems, meeting transcripts have emerged as a promising domain for natural language tasks. Most recent works on meeting transcripts primarily focus on summarization and extraction of action items. However, meeting discussions also have a useful question-answering (QA) component, crucial to understanding the discourse or meeting content, and can be used to build interactive interfaces on top of long transcripts. Hence, in this work, we leverage this inherent QA component of meeting discussions and introduce MeetingQA, an extractive QA dataset comprising of questions asked by meeting participants and corresponding responses. As a result, questions can be open-ended and actively seek discussions, while the answers can be multi-span and distributed across multiple speakers. Our comprehensive empirical study of several robust baselines including long-context language models and recent instruction-tuned models reveals that models perform poorly on this task (F1 = 57.3) and severely lag behind human performance (F1 = 84.6), thus presenting a challenging new task for the community to improve upon.

Most weakly supervised named entity recognition (NER) models rely on domain-specific dictionaries provided by experts. This approach is infeasible in many domains where dictionaries do not exist. While a phrase retrieval model was used to construct pseudo-dictionaries with entities retrieved from Wikipedia automatically in a recent study, these dictionaries often have limited coverage because the retriever is likely to retrieve popular entities rather than rare ones. In this study, we present a novel framework, HighGEN, that generates NER datasets with high-coverage pseudo-dictionaries. Specifically, we create entity-rich dictionaries with a novel search method, called phrase embedding search, which encourages the retriever to search a space densely populated with various entities. In addition, we use a new verification process based on the embedding distance between candidate entity mentions and entity types to reduce the false-positive noise in weak labels generated by high-coverage dictionaries. We demonstrate that HighGEN outperforms the previous best model by an average F1 score of 4.7 across five NER benchmark datasets.

Modern image captioning models are usually trained with text similarity objectives. However, since reference captions in public datasets often describe the most salient common objects, models trained with the text similarity objectives tend to ignore specific and detailed aspects of an image that distinguish it from others. Towards more descriptive and distinctive caption generation, we propose to use CLIP, a multimodal encoder trained on huge image-text pairs from the web, to calculate multi-modal similarity and use it as a reward function. We also propose a simple finetuning strategy of CLIP text encoder to improve grammar that does not require extra text annotation. This completely eliminates the need for reference captions during the reward computation. To comprehensively evaluate descriptive captions, we introduce FineCapEval, a new dataset for caption evaluation with fine-grained criteria: overall, background, object, relations. In our experiments on text-to-image retrieval and FineCapEval, the proposed CLIP-guided model generates more distinctive captions than the CIDEroptimized model. We also show that our unsupervised grammar finetuning of the CLIP text encoder alleviates the degeneration problem of the naive CLIP reward. Lastly, we show human analysis where the annotators strongly prefer CLIP reward to CIDEr and MLE objectives on diverse criteria.

Individuals, educational institutions, and businesses are prolific at generating instructional video content such as “how-to” and tutorial guides. While significant progress has been made in basic video understanding tasks, identifying procedural intent within these instructional videos is a challenging and important task that remains unexplored but essential to video summarization, search, and recommendations. This paper introduces the problem of instructional intent identification and extraction from software instructional livestreams. We construct and present a new multimodal dataset consisting of software instructional livestreams and containing manual annotations for both detailed and abstract procedural intent that enable training and evaluation of joint video and text understanding models. We then introduce a multimodal cascaded cross-attention model to efficiently combine the weaker and noisier video signal with the more discriminative text signal. Our experiments show that our proposed model brings significant gains compared to strong baselines, including large-scale pretrained multimodal models. Our analysis further identifies that the task benefits from spatial as well as motion features extracted from videos, and provides insight on how the video signal is preferentially used for intent discovery. We also show that current models struggle to comprehend the nature of abstract intents, revealing important gaps in multimodal understanding and paving the way for future work.