Randomized experiments are the preferred approach for evaluating the effects of interventions, but they are costly and often yield estimates with substantial uncertainty. On the other hand, in silico experiments leveraging foundation models offer a cost-effective alternative that can potentially attain higher statistical precision. However, the benefits of in silico experiments come with a significant risk: statistical inferences are not valid if the models fail to accurately predict experimental responses to interventions. In this paper, we propose a novel approach that integrates the predictions from multiple foundation models with experimental data while preserving valid statistical inference. Our estimator is consistent and asymptotically normal, with asymptotic variance no larger than the standard estimator based on experimental data alone. Importantly, these statistical properties hold even when model predictions are arbitrarily biased. Empirical results across several randomized experiments show that our estimator offers substantial precision gains, equivalent to a reduction of up to 20% in the sample size needed to match the same precision as the standard estimator based on experimental data alone.

Large language models (LLMs) are known to struggle with consistently generating truthful responses. While various representation intervention techniques have been proposed, these methods typically apply a universal representation correction vector to all input queries, limiting their effectiveness against diverse queries in practice. In this study, we introduce TruthFlow, a novel method that leverages the Flow Matching technique for query-specific truthful representation correction. Specifically, TruthFlow first uses a flow model to learn query-specific correction vectors that transition representations from hallucinated to truthful states. Then, during inference, the trained flow model generates these correction vectors to enhance the truthfulness of LLM outputs. Experimental results demonstrate that TruthFlow significantly improves performance on open-ended generation tasks across various advanced LLMs evaluated on TruthfulQA. Moreover, the trained TruthFlow model exhibits strong transferability, performing effectively on other unseen hallucination benchmarks.

We investigate algorithmic decision problems where agents can respond strategically to the decision maker's (DM) models. The demand for clear and actionable explanations from DMs to (potentially strategic) agents continues to rise. While prior work often treats explanations as full model disclosures, explanations in practice might convey only partial information, which can lead to misinterpretations and harmful responses. When full disclosure of the predictive model is neither feasible nor desirable, a key open question is how DMs can use explanations to maximise their utility without compromising agent welfare. In this work, we explore well-known local and global explanation methods, and establish a necessary condition to prevent explanations from misleading agents into self-harming actions. Moreover, with conditional homogeneity, we establish that action recommendation (AR)-based explanations are sufficient for non-harmful responses, akin to the revelation principle in information design. To operationalise AR-based explanations, we propose a simple algorithm to jointly optimise the predictive model and AR policy to balance DM outcomes with agent welfare. Our empirical results demonstrate the benefits of this approach as a more refined strategy for safe and effective partial model disclosure in algorithmic decision-making.

Large language models (LLMs), such as o1 from OpenAI, have demonstrated remarkable reasoning capabilities. o1 generates a long chain-of-thought (LongCoT) before answering a question. LongCoT allows LLMs to analyze problems, devise plans, reflect, and backtrack effectively. These actions empower LLM to solve complex problems. After the release of o1, many teams have attempted to replicate its LongCoT and reasoning capabilities. In terms of methods, they primarily rely on knowledge distillation with data from existing models with LongCoT capacities (e.g., OpenAI-o1, Qwen-QwQ, DeepSeek-R1-Preview), leaving significant uncertainties on systematically developing such reasoning abilities. In terms of data domains, these works focus narrowly on math while a few others include coding, limiting their generalizability. This paper introduces a novel approach to enable LLM's LongCoT capacity without distillation from o1-like models or expensive human annotations, where we bootstrap LongCoT (BOLT) from a standard instruct model. BOLT involves three stages: 1) LongCoT data bootstrapping with in-context learning on a standard instruct model; 2) LongCoT supervised finetuning; 3) online training to further refine LongCoT capacities. In BOLT, only a few in-context examples need to be constructed during the bootstrapping stage; in our experiments, we created 10 examples, demonstrating the feasibility of this approach. We use Llama-3.1-70B-Instruct to bootstrap LongCoT and apply our method to various model scales (7B, 8B, 70B). We achieve impressive performance on a variety of benchmarks, Arena-Hard, MT-Bench, WildBench, ZebraLogic, MATH500, which evaluate diverse task-solving and reasoning capabilities.

Large Language Models (LLMs) are increasingly helpful in text generation, even writing code in programming languages based on user prompts written in natural language. They are even applied to generate simulation models for multibody systems from natural language. Research results suggest that LLMs surpass the mere replication of existing code examples, where some LLMs have been trained on an open-source multibody simulation code. However, for closed-source simulation software, such results are not to be expected as their ideas and concepts might differ from other publicly available ones. LLMs can hallucinate for knowledge-intensive tasks, such as model creation, which can lead to wrong responses. This is especially the case for the LLM unknown closed-source simulation software. The same applies to other internal knowledge kept private to protect intellectual property or data privacy. The Retrieval-Augmented Generation (RAG) approach might yield a solution for these knowledge-intensive tasks. This paper explores the application of RAG to closed-source simulation software and presents first experiments. After a brief introduction to LLMs, the RAG approach, and the simulation method applied by the close-source simulation software, several examples are provided to test LLMs' knowledge of the simulation software and the creation of simulation models using two RAG systems. The examples show promising results indicating the benefits of applying RAG systems to closed-source simulation software, helping to access their knowledge. Nevertheless, they also reveal gaps in the applied information and open questions for further research.

Despite notable advancements in Retrieval-Augmented Generation (RAG) systems that expand large language model (LLM) capabilities through external retrieval, these systems often struggle to meet the complex and diverse needs of real-world industrial applications. The reliance on retrieval alone proves insufficient for extracting deep, domain-specific knowledge performing in logical reasoning from specialized corpora. To address this, we introduce sPecIalized KnowledgE and Rationale Augmentation Generation (PIKE-RAG), focusing on extracting, understanding, and applying specialized knowledge, while constructing coherent rationale to incrementally steer LLMs toward accurate responses. Recognizing the diverse challenges of industrial tasks, we introduce a new paradigm that classifies tasks based on their complexity in knowledge extraction and application, allowing for a systematic evaluation of RAG systems' problem-solving capabilities. This strategic approach offers a roadmap for the phased development and enhancement of RAG systems, tailored to meet the evolving demands of industrial applications. Furthermore, we propose knowledge atomizing and knowledge-aware task decomposition to effectively extract multifaceted knowledge from the data chunks and iteratively construct the rationale based on original query and the accumulated knowledge, respectively, showcasing exceptional performance across various benchmarks.

Google has dropped a pledge not to use artificial intelligence for weapons or surveillance in its updated ethics policy on the powerful technology. In its previous version of "AI Principles", the California-based internet giant included a commitment not to pursue AI technologies that "cause or are likely to cause overall harm", including weapons and surveillance that violates "internationally accepted norms". Google's revised policy announced on Tuesday states that the company pursues AI "responsibly" and in line with "widely accepted principles of international law and human rights", but does not include the previous language about weapons or surveillance. "We believe democracies should lead in AI development, guided by core values like freedom, equality, and respect for human rights," Google DeepMind chief Demis Hassabis and research labs senior vice president James Manyika said in a blog post announcing the updated policy. "And we believe that companies, governments, and organizations sharing these values should work together to create AI that protects people, promotes global growth, and supports national security."

One critical challenge in the institutional adoption journey of Large Language Models (LLMs) stems from their propensity to hallucinate in generated responses. To address this, we propose MultiQ&A, a systematic approach for evaluating the robustness and consistency of LLM-generated answers. We demonstrate MultiQ&A's ability to crowdsource question perturbations and their respective answers through independent LLM agents at scale. Our experiments culminated in the examination of 1.9 million question perturbations and 2.3 million answers. Furthermore, MultiQ&A shows that ensembled LLMs, such as gpt-3.5-turbo, remain relatively robust and consistent under perturbations. MultiQ&A provides clarity in the response generation space, offering an effective method for inspecting disagreements and variability. Therefore, our system offers a potential framework for institutional LLM adoption with the ability to measure confidence, consistency, and the quantification of hallucinations.

This report analyzes key trends, challenges, risks, and opp ortunities associated with the development of Large Language Models (LLMs) globally. It examines natio nal experiences in developing LLMs and assesses the feasibility of investment in this sector. Addi tionally, the report explores strategies for implementing, regulating, and financing AI projects at the s tate level. International experiences indicate that LLMs significantl y enhance administrative efficiency. In regulatory processes, they streamline the management of le gal documents (Albania, Serbia), facilitate communication between government authorities and citizen s (Netherlands), and support public procurement and legal translations (Albania).

Content Warning: This paper contains examples of harmful language. Jailbreaking attacks can effectively manipulate open-source large language models (LLMs) to produce harmful responses. However, these attacks exhibit limited transferability, failing to disrupt proprietary LLMs consistently. To reliably identify vulnerabilities in proprietary LLMs, this work investigates the transferability of jailbreaking attacks by analysing their impact on the model's intent perception. Nevertheless, these adversarial sequences fail to mislead the target LLM's intent perception, allowing the target LLM to refocus on malicious-intent tokens and abstain from responding. Our analysis further reveals the inherent distributional dependency within the generated adversarial sequences, whose effectiveness stems from overfitting the source LLM's parameters, resulting in limited transferability to target LLMs. To this end, we propose the Perceived-importance Flatten (PiF) method, which uniformly disperses the model's focus across neutral-intent tokens in the original input, thus obscuring malicious-intent tokens without relying on overfitted adversarial sequences. Extensive experiments demonstrate that PiF provides an effective and efficient red-teaming evaluation for proprietary LLMs. Empowered by massive corpus, large language models (LLMs) have achieved human-level conversational capabilities (OpenAI, 2023a; Google, 2023; Meta, 2024) and are widely employed in real-world applications. However, their training corpus is mainly crawled from the Internet without thorough ethical review, raising concerns about the potential risks associated with LLMs. Recent red-teaming efforts highlight that jailbreaking attacks can effectively disrupt LLMs to produce undesirable content with harmful consequences (Perez et al., 2022; Ganguli et al., 2022; Ouyang et al., 2022). Unlike model-level jailbreaks that necessitate parameter modifications and are restricted to opensource LLMs (Qi et al., 2024; Huang et al., 2023a), token-level and prompt-level jailbreaks can generate transferable adversarial sequences (Yu et al., 2023; Lapid et al., 2023), thus posing a potential threat to widespread proprietary LLMs (Zou et al., 2023; Chao et al., 2023). Nevertheless, empirical results indicate that these adversarial sequences lack reliable transferability, failing to consistently manipulate target LLMs (Chao et al., 2024; Chen et al., 2024). Furthermore, these lengthy adversarial sequences can be further countered by adaptive jailbreaking detection and defence (Alon & Kamfonas, 2023; Inan et al., 2023; Robey et al., 2023; Wang et al., 2024a). As depicted in Figure 1, developing jailbreak attacks that can reliably identify vulnerabilities in proprietary LLMs--thereby promoting human alignment and preventing future misuse--remains a significant challenge. These attacks are initially generated on the source LLM (Llama-2-7B-Chat) and subsequently transferred to the target LLM (Llama-2-13B-Chat).