Language model alignment research often attempts to ensure that models are not only helpful and harmless, but also truthful and unbiased. However, optimizing these objectives simultaneously can obscure how improving one aspect might impact the others. In this work, we focus on analyzing the relationship between two concepts essential in both language model alignment and political science: \textit{truthfulness} and \textit{political bias}. We train reward models on various popular truthfulness datasets and subsequently evaluate their political bias. Our findings reveal that optimizing reward models for truthfulness on these datasets tends to result in a left-leaning political bias. We also find that existing open-source reward models (i.e. those trained on standard human preference datasets) already show a similar bias and that the bias is larger for larger models. These results raise important questions about both the datasets used to represent truthfulness and what language models capture about the relationship between truth and politics.

The massive use of machine learning models, particularly neural networks, has raised serious concerns about their environmental impact. Indeed, over the last few years we have seen an explosion in the computing costs associated with training and deploying these systems. It is, therefore, crucial to understand their energy requirements in order to better integrate them into the evaluation of models, which has so far focused mainly on performance. In this paper, we study several neural network architectures that are key components of sound event detection systems, using an audio tagging task as an example. We measure the energy consumption for training and testing small to large architectures and establish complex relationships between the energy consumption, the number of floating-point operations, the number of parameters, and the GPU/memory utilization.

It has been demonstrated in many scientific fields that artificial neural networks like autoencoders or Siamese networks encode meaningful concepts in their latent spaces. However, there does not exist a comprehensive framework for retrieving this information in a human-readable form without prior knowledge. In order to extract these concepts, we introduce a framework for finding closed-form interpretations of neurons in latent spaces of artificial neural networks. The interpretation framework is based on embedding trained neural networks into an equivalence class of functions that encode the same concept. We interpret these neural networks by finding an intersection between the equivalence class and human-readable equations defined by a symbolic search space. The approach is demonstrated by retrieving invariants of matrices and conserved quantities of dynamical systems from latent spaces of Siamese neural networks.

Information Extraction (IE) and Text Classification (CLS) serve as the fundamental pillars of NLU, with both disciplines relying on analyzing input sequences to categorize outputs into pre-established schemas. However, there is no existing encoder-based model that can unify IE and CLS tasks from this perspective. To fully explore the foundation shared within NLU tasks, we have proposed a Recursive Method with Explicit Schema Instructor for Universal NLU. Specifically, we firstly redefine the true universal information extraction (UIE) with a formal formulation that covers almost all extraction schemas, including quadruples and quintuples which remain unsolved for previous UIE models. Then, we expands the formulation to all CLS and multi-modal NLU tasks. Based on that, we introduce RexUniNLU, an universal NLU solution that employs explicit schema constraints for IE and CLS, which encompasses all IE and CLS tasks and prevent incorrect connections between schema and input sequence. To avoid interference between different schemas, we reset the position ids and attention mask matrices. Extensive experiments are conducted on IE, CLS in both English and Chinese, and multi-modality, revealing the effectiveness and superiority. Our codes are publicly released.

Ensuring compliance with international data protection standards for privacy and data security is a crucial but complex task, often requiring substantial legal expertise. This paper introduces LegiLM, a novel legal language model specifically tailored for consulting on data or information compliance. LegiLM leverages a pre-trained GDPR Fines dataset and has been fine-tuned to automatically assess whether particular actions or events breach data security and privacy regulations. By incorporating a specialized dataset that includes global data protection laws, meticulously annotated policy documents, and relevant privacy policies, LegiLM is optimized for addressing data compliance challenges. The model integrates advanced legal reasoning methods and information retrieval enhancements to enhance accuracy and reliability in practical legal consulting scenarios. Our evaluation using a custom benchmark dataset demonstrates that LegiLM excels in detecting data regulation breaches, offering sound legal justifications, and recommending necessary compliance modifications, setting a new benchmark for AI-driven legal compliance solutions. Our resources are publicly available at https://github.com/DAOLegalAI/LegiLM

Large language models (LLMs) are increasingly used in business dialogue systems but they pose security and ethical risks. Multiturn conversations, where context influences the model's behavior, can be exploited to produce undesired responses. In this paper, we examine the effectiveness of utilizing off-theshelf LLMs in straightforward red-teaming approaches, where an attacker LLM aims to elicit undesired output from a target LLM, comparing both single-turn and conversational redteaming tactics. Our experiments offer insights into various usage strategies that significantly affect their performance as red teamers. They suggest that off-the-shelf models can act as effective red teamers and even adjust their attack strategy based on past attempts, although their effectiveness decreases with greater alignment. Figure 1: An example dialogue between a red-teaming Warning: This paper contains examples and model (red) and the target model (blue) in a conversational model-generated content that may be considered setting, with a judge LLM (grey) scoring the offensive.

This paper considers the scenario in which there are multiple institutions, each with a limited capacity for candidates, and candidates, each with preferences over the institutions. A central entity evaluates the utility of each candidate to the institutions, and the goal is to select candidates for each institution in a way that maximizes utility while also considering the candidates' preferences. The paper focuses on the setting in which candidates are divided into multiple groups and the observed utilities of candidates in some groups are biased--systematically lower than their true utilities. The first result is that, in these biased settings, prior algorithms can lead to selections with sub-optimal true utility and significant discrepancies in the fraction of candidates from each group that get their preferred choices. Subsequently, an algorithm is presented along with proof that it produces selections that achieve near-optimal group fairness with respect to preferences while also nearly maximizing the true utility under distributional assumptions. Further, extensive empirical validation of these results in real-world and synthetic settings, in which the distributional assumptions may not hold, are presented.

In hundreds of videos since taken down by YouTube, right-wing influencers working for Tenet Media--a company the US Department of Justice alleges was financed and guided by a state-backed Russian news network--showed interest in a highly specific set of topics, according to a WIRED analysis. Using closed captioning of the videos we downloaded before the videos were removed, we've compiled lists of terms frequently mentioned in them, along with a searchable database: The content of these videos was described by prosecutors as "consistent" with Russia's aim of sowing political discord in the US. Among the areas covered: free speech, illegal immigrants, diversity in video games, supposed racism toward white people, and Elon Musk. While an indictment unsealed earlier this week does not name Tenet, WIRED and other outlets were able to identify it because prosecutors gave its motto as that of a business identified as "U.S. Company-1." Prosecutors allege that two employees of the state-backed Russian network RT, Kostiantyn Kalashnikov and Elena Afanasyeva, who are charged with conspiracy to commit money laundering and to violate the Foreign Agents Registration Act, paid Tenet and its parent company 9.7 million to produce and distribute videos supporting Russian aims.

This article was featured in the One Story to Read Today newsletter. "About 14 billion years ago, matter, energy, time and space came into being." So begins Sapiens: A Brief History of Humankind (2011), by the Israeli historian Yuval Noah Harari, and so began one of the 21st century's most astonishing academic careers. Sapiens has sold more than 25 million copies in various languages. Since then, Harari has published several other books, which have also sold millions. He now employs some 15 people to organize his affairs and promote his ideas. Check out more from this issue and find your next story to read. Harari might be, after the Dalai Lama, the figure of global renown who is least online.

A musician in the US has been accused of using artificial intelligence (AI) tools and thousands of bots to fraudulently stream songs billions of times in order to claim millions of dollars of royalties. Michael Smith, of North Carolina, has been charged with three counts of wire fraud, wire fraud conspiracy and money laundering conspiracy charges. Prosecutors say it is the first criminal case of its kind they have handled. "Through his brazen fraud scheme, Smith stole millions in royalties that should have been paid to musicians, songwriters, and other rights holders whose songs were legitimately streamed," said US attorney Damian Williams. According to an unsealed indictment detailing the charges, the 52-year-old used hundreds of thousands of AI-generated songs to manipulate streams.