Automatic fact verification systems increasingly rely on large language models (LLMs). We investigate how parametric knowledge biases in these models affect fact-checking outcomes of the HerO system (baseline for FEVER-25). We examine how the system is affected by: (1) potential bias in Llama 3.1's parametric knowledge and (2) intentionally injected bias. When prompted directly to perform fact-verification, Llama 3.1 labels nearly half the claims as "Not Enough Evidence". Using only its parametric knowledge it is able to reach a verdict on the remaining half of the claims. In the second experiment, we prompt the model to generate supporting, refuting, or neutral fact-checking documents. These prompts significantly influence retrieval outcomes, with approximately 50\% of retrieved evidence being unique to each perspective. Notably, the model sometimes refuses to generate supporting documents for claims it believes to be false, creating an inherent negative bias. Despite differences in retrieved evidence, final verdict predictions show stability across prompting strategies. The code is available at: https://github.com/eibakke/FEVER-8-Shared-Task

In this paper, we present an overview of ClinIQLink, a shared task, collocated with the 24th BioNLP workshop at ACL 2025, designed to stress-test large language models (LLMs) on medically-oriented question answering aimed at the level of a General Practitioner. The challenge supplies 4,978 expert-verified, medical source-grounded question-answer pairs that cover seven formats: true/false, multiple choice, unordered list, short answer, short-inverse, multi-hop, and multi-hop-inverse. Participating systems, bundled in Docker or Apptainer images, are executed on the CodaBench platform or the University of Maryland's Zaratan cluster. An automated harness (Task 1) scores closed-ended items by exact match and open-ended items with a three-tier embedding metric. A subsequent physician panel (Task 2) audits the top model responses.

Trajectory data, capturing human actions and environmental states across various modalities, holds significant potential for enhancing AI agent capabilities, particularly in GUI environments. However, how to model the representation of trajectory-level data presents a significant challenge that has not been systematically addressed amid explosive trajectory data growth. In this work, we introduce Multimodal Trajectory Retrieval, bridging the gap between universal retrieval and agent-centric trajectory modeling. We construct the Unified Agent Trajectory Dataset (UATD) from annotated demonstrations and states across diverse real-world scenarios. Based on this, we present GAE-Bench, a benchmark containing a large number of trajectory-based retrieval pairs. In addition, we propose GAE-Retriever, a multimodal retrieval framework that adopts vision-language models and incorporates optimized contrastive learning through a token selection and the GradCache mechanism. Comprehensive evaluations across multiple datasets show that GAE-Retriever consistently outperforms strong baselines in retrieval recall, highlighting its effectiveness in advancing multimodal trajectory retrieval.

The co creativity community is making significant progress in developing more sophisticated and tailored systems to support and enhance human creativity. Design considerations from prior work can serve as a valuable and efficient foundation for future systems. To support this effort, we conducted a systematic literature review of 62 papers on co-creative systems. These papers cover a diverse range of applications, including visual arts, design, and writing, where the AI acts not just as a tool but as an active collaborator in the creative process. From this review, we identified several key dimensions relevant to system design: phase of the creative process, creative task, proactive behavior of the system, user control, system embodiment, and AI model type. Our findings suggest that systems offering high user control lead to greater satisfaction, trust, and a stronger sense of ownership over creative outcomes. Furthermore, proactive systems, when adaptive and context sensitive, can enhance collaboration. We also extracted 24 design considerations, highlighting the value of encouraging users to externalize their thoughts and of increasing the system's social presence and transparency to foster trust. Despite recent advancements, important gaps remain, such as limited support for early creative phases like problem clarification, and challenges related to user adaptation to AI systems.

Since state-of-the-art LLMs often underperform in languages other than English or Chinese, improving the capability of LLMs in new languages has become an essential task. Moreover, LLMs' entire end-to-end training process remains largely unknown to the public due to proprietary reasons, technical complexity, inconsistent documentation, and ethical considerations. The complete picture remains a closely guarded secret within the industry. This paper presents methods to adapt an existing English-based LLM to Korean in a low-budget scenario. We describe the entire end-to-end process: collecting Korean datasets, preprocessing the data, training the model, creating downstream benchmarks, and conducting evaluations. The evaluation results indicate that our method can effectively and cost-efficiently add new language capabilities to existing LLMs. Our new bilingual models, Thunder-LLM and Thunder-LLM-Ins, achieve superior Korean performance compared to state-of-the-art models while utilizing minimal data and computational resources. We share our comprehensive experience and make the code publicly available.

Data is fundamental to the training of language models (LM). Recent research has been dedicated to data efficiency, which aims to maximize performance by selecting a minimal or optimal subset of training data. Techniques such as data filtering, sampling, and selection play a crucial role in this area. To complement it, we define Data Efficacy, which focuses on maximizing performance by optimizing the organization of training data and remains relatively underexplored. This work introduces a general paradigm, DELT, for considering data efficacy in LM training, which highlights the significance of training data organization. DELT comprises three components: Data Scoring, Data Selection, and Data Ordering. Among these components, we design Learnability-Quality Scoring (LQS), as a new instance of Data Scoring, which considers both the learnability and quality of each data sample from the gradient consistency perspective. We also devise Folding Ordering (FO), as a novel instance of Data Ordering, which addresses issues such as model forgetting and data distribution bias. Comprehensive experiments validate the data efficacy in LM training, which demonstrates the following: Firstly, various instances of the proposed DELT enhance LM performance to varying degrees without increasing the data scale and model size. Secondly, among these instances, the combination of our proposed LQS for data scoring and Folding for data ordering achieves the most significant improvement. Lastly, data efficacy can be achieved together with data efficiency by applying data selection. Therefore, we believe that data efficacy is a promising foundational area in LM training.

Forecasting is a challenging task that offers a clearly measurable way to study AI systems. Forecasting requires a large amount of research on the internet, and evaluations require time for events to happen, making the development of forecasting benchmarks challenging. To date, no forecasting benchmark provides a realistic, hermetic, and repeatable environment for LLM forecasters. We introduce Bench To the Future (BTF), a "pastcasting" benchmark with hundreds of high-quality questions for which the resolution is already known. Each question is accompanied by a large offline corpus of tens of thousands of relevant web pages, enabling a way to elicit realistic "forecasts" on past events from LLMs. Results suggest that our pastcasting environment can produce results comparable to those based on forecasts using the internet on at-the-time unresolved questions. We show results benchmarking agent and chain-of-thought forecasting approaches using several LLMs, including the recently-released Claude 4 models, and demonstrate BTF's ability to track steady forecasting capability progress over time. We intend this to be a living benchmark, with new questions added continually to account for increasing training data cutoff dates. We invite researchers to contact us at hello@futuresearch.ai to utilize our benchmark or tooling for their own research.

Large Language Models (LLMs) have played a pivotal role in advancing Artificial Intelligence (AI). However, despite their achievements, LLMs often struggle to explain their decision-making processes, making them a 'black box' and presenting a substantial challenge to explainability. This lack of transparency poses a significant obstacle to the adoption of LLMs in high-stakes domain applications, where interpretability is particularly essential. To overcome these limitations, researchers have developed various explainable artificial intelligence (XAI) methods that provide human-interpretable explanations for LLMs. However, a systematic understanding of these methods remains limited. To address this gap, this survey provides a comprehensive review of explainability techniques by categorizing XAI methods based on the underlying transformer architectures of LLMs: encoder-only, decoder-only, and encoder-decoder models. Then these techniques are examined in terms of their evaluation for assessing explainability, and the survey further explores how these explanations are leveraged in practical applications. Finally, it discusses available resources, ongoing research challenges, and future directions, aiming to guide continued efforts toward developing transparent and responsible LLMs.

The rapid rise of open-weight and open-source foundation models is intensifying the obligation and reshaping the opportunity to make AI systems safe. This paper reports outcomes from the Columbia Convening on AI Openness and Safety (San Francisco, 19 Nov 2024) and its six-week preparatory programme involving more than forty-five researchers, engineers, and policy leaders from academia, industry, civil society, and government. Using a participatory, solutions-oriented process, the working groups produced (i) a research agenda at the intersection of safety and open source AI; (ii) a mapping of existing and needed technical interventions and open source tools to safely and responsibly deploy open foundation models across the AI development workflow; and (iii) a mapping of the content safety filter ecosystem with a proposed roadmap for future research and development. We find that openness -- understood as transparent weights, interoperable tooling, and public governance -- can enhance safety by enabling independent scrutiny, decentralized mitigation, and culturally plural oversight. However, significant gaps persist: scarce multimodal and multilingual benchmarks, limited defenses against prompt-injection and compositional attacks in agentic systems, and insufficient participatory mechanisms for communities most affected by AI harms. The paper concludes with a roadmap of five priority research directions, emphasizing participatory inputs, future-proof content filters, ecosystem-wide safety infrastructure, rigorous agentic safeguards, and expanded harm taxonomies. These recommendations informed the February 2025 French AI Action Summit and lay groundwork for an open, plural, and accountable AI safety discipline.

The proliferation of large language models (LLMs) has significantly transformed the digital information landscape, making it increasingly challenging to distinguish between human-written and LLM-generated content. Detecting LLM-generated information is essential for preserving trust on digital platforms (e.g., social media and e-commerce sites) and preventing the spread of misinformation, a topic that has garnered significant attention in IS research. However, current detection methods, which primarily focus on identifying content generated by specific LLMs in known domains, face challenges in generalizing to new (i.e., unseen) LLMs and domains. This limitation reduces their effectiveness in real-world applications, where the number of LLMs is rapidly multiplying and content spans a vast array of domains. In response, we introduce a general LLM detector (GLD) that combines a twin memory networks design and a theory-guided detection generalization module to detect LLM-generated information across unseen LLMs and domains. Using real-world datasets, we conduct extensive empirical evaluations and case studies to demonstrate the superiority of GLD over state-of-the-art detection methods. The study has important academic and practical implications for digital platforms and LLMs.