Large Language Models (LLMs) have demonstrated remarkable performance on various quantitative reasoning and knowledge benchmarks. However, many of these benchmarks are losing utility as LLMs get increasingly high scores, despite not yet reaching expert performance in these domains. We introduce ARB, a novel benchmark composed of advanced reasoning problems in multiple fields. ARB presents a more challenging test than prior benchmarks, featuring problems in mathematics, physics, biology, chemistry, and law. As a subset of ARB, we introduce a challenging set of math and physics problems which require advanced symbolic reasoning and domain knowledge. We evaluate recent models such as GPT-4 and Claude on ARB and demonstrate that current models score well below 50% on more demanding tasks. In order to improve both automatic and assisted evaluation capabilities, we introduce a rubric-based evaluation approach, allowing GPT-4 to score its own intermediate reasoning steps. Further, we conduct a human evaluation of the symbolic subset of ARB, finding promising agreement between annotators and GPT-4 rubric evaluation scores.

In argumentative writing, writers must brainstorm hierarchical writing goals, ensure the persuasiveness of their arguments, and revise and organize their plans through drafting. Recent advances in large language models (LLMs) have made interactive text generation through a chat interface (e.g., ChatGPT) possible. However, this approach often neglects implicit writing context and user intent, lacks support for user control and autonomy, and provides limited assistance for sensemaking and revising writing plans. To address these challenges, we introduce VISAR, an AI-enabled writing assistant system designed to help writers brainstorm and revise hierarchical goals within their writing context, organize argument structures through synchronized text editing and visual programming, and enhance persuasiveness with argumentation spark recommendations. VISAR allows users to explore, experiment with, and validate their writing plans using automatic draft prototyping. A controlled lab study confirmed the usability and effectiveness of VISAR in facilitating the argumentative writing planning process.

The topic of Climate Change (CC) has received limited attention in NLP despite its urgency. Activists and policymakers need NLP tools to effectively process the vast and rapidly growing unstructured textual climate reports into structured form. To tackle this challenge we introduce two new large-scale climate questionnaire datasets and use their existing structure to train self-supervised models. We conduct experiments to show that these models can learn to generalize to climate disclosures of different organizations types than seen during training. We then use these models to help align texts from unstructured climate documents to the semi-structured questionnaires in a human pilot study. Finally, to support further NLP research in the climate domain we introduce a benchmark of existing climate text classification datasets to better evaluate and compare existing models.

On weekdays, between homeschooling her two children, Michelle Curtis logs on to her computer to squeeze in a few hours of work. Her screen flashes with Google Search results, the writings of a Google chatbot, and the outputs of other algorithms, and she has a few minutes to respond to each--judging the usefulness of the blue links she's been provided, checking the accuracy of an AI's description of a praying mantis, or deciding which of two chatbot-written birthday poems is better. She never knows what she will have to assess in advance, and for the AI-related tasks, which have formed the bulk of her work since February, she says she has little guidance and not enough time to do a thorough job. Curtis is an AI rater. She works for the data company Appen, which is subcontracted by Google to evaluate the outputs of the tech giant's AI products and search algorithm.

OpenAI's ChatGPT language model has gained popularity as a powerful tool for complex problem-solving and information retrieval. However, concerns arise about the reproduction of biases present in the language-specific training data. In this study, we address this issue in the context of the Israeli-Palestinian and Turkish-Kurdish conflicts. Using GPT-3.5, we employed an automated query procedure to inquire about casualties in specific airstrikes, in both Hebrew and Arabic for the former conflict and Turkish and Kurdish for the latter. Our analysis reveals that GPT-3.5 provides 27$\pm$11 percent lower fatality estimates when queried in the language of the attacker than in the language of the targeted group. Evasive answers denying the existence of such attacks further increase the discrepancy, creating a novel bias mechanism not present in regular search engines. This language bias has the potential to amplify existing media biases and contribute to information bubbles, ultimately reinforcing conflicts.

Procedural planning, which entails decomposing a high-level goal into a sequence of temporally ordered steps, is an important yet intricate task for machines. It involves integrating common-sense knowledge to reason about complex contextualized situations that are often counterfactual, e.g. "scheduling a doctor's appointment without a phone". While current approaches show encouraging results using large language models (LLMs), they are hindered by drawbacks such as costly API calls and reproducibility issues. In this paper, we advocate planning using smaller language models. We present PlaSma, a novel two-pronged approach to endow small language models with procedural knowledge and (counterfactual) planning capabilities. More concretely, we develop symbolic procedural knowledge distillation to enhance the implicit knowledge in small language models and an inference-time algorithm to facilitate more structured and accurate reasoning. In addition, we introduce a novel task, Counterfactual Planning, that requires a revision of a plan to cope with a counterfactual situation. In both the original and counterfactual setting, we show that orders-of-magnitude smaller models (770M-11B parameters) can compete and often surpass their larger teacher models' capabilities.

This paper critically evaluates the European Commission's proposed AI Act's approach to risk management and risk acceptability for high-risk AI systems that pose risks to fundamental rights and safety. The Act aims to promote "trustworthy" AI with a proportionate regulatory burden. Its provisions on risk acceptability require residual risks from high-risk systems to be reduced or eliminated "as far as possible", having regard to the "state of the art". This criterion, especially if interpreted narrowly, is unworkable and promotes neither proportionate regulatory burden, nor trustworthiness. By contrast the Parliament's most recent draft amendments to the risk management provisions introduce "reasonableness", cost-benefit analysis, and are more transparent about the value-laden and contextual nature of risk acceptability judgements. This paper argues that the Parliament's approach is more workable, and better balances the goals of proportionality and trustworthiness. It explains what reasonableness in risk acceptability judgments would entail, drawing on principles from negligence law and European medical devices regulation. And it contends that the approach to risk acceptability judgments need a firm foundation of civic legitimacy: including detailed guidance or involvement from regulators, and meaningful input from affected stakeholders.

As one of the most pervasive applications of machine learning, recommender systems are playing an important role on assisting human decision making. The satisfaction of users and the interests of platforms are closely related to the quality of the generated recommendation results. However, as a highly data-driven system, recommender system could be affected by data or algorithmic bias and thus generate unfair results, which could weaken the reliance of the systems. As a result, it is crucial to address the potential unfairness problems in recommendation settings. Recently, there has been growing attention on fairness considerations in recommender systems with more and more literature on approaches to promote fairness in recommendation. However, the studies are rather fragmented and lack a systematic organization, thus making it difficult to penetrate for new researchers to the domain. This motivates us to provide a systematic survey of existing works on fairness in recommendation. This survey focuses on the foundations for fairness in recommendation literature. It first presents a brief introduction about fairness in basic machine learning tasks such as classification and ranking in order to provide a general overview of fairness research, as well as introduce the more complex situations and challenges that need to be considered when studying fairness in recommender systems. After that, the survey will introduce fairness in recommendation with a focus on the taxonomies of current fairness definitions, the typical techniques for improving fairness, as well as the datasets for fairness studies in recommendation. The survey also talks about the challenges and opportunities in fairness research with the hope of promoting the fair recommendation research area and beyond.

Machines with human-level intelligence should be able to do most economically valuable work. This aligns a major economic incentive with the scientific grand challenge of building a human-like mind. Here we describe our approach to building and testing such a system. Our approach comprises a physical humanoid robotic system; a software based control system for robots of this type; a performance metric, which we call g+, designed to be a measure of human-like intelligence in humanoid robots; and an evolutionary algorithm for incrementally increasing scores on this performance metric. We introduce and describe the current status of each of these. We report on current and historical measurements of the g+ metric on the systems described here.

The quality of training data impacts the performance of pre-trained large language models (LMs). Given a fixed budget of tokens, we study how to best select data that leads to good downstream model performance across tasks. We develop a new framework based on a simple hypothesis: just as humans acquire interdependent skills in a deliberate order, language models also follow a natural order when learning a set of skills from their training data. If such an order exists, it can be utilized for improved understanding of LMs and for data-efficient training. Using this intuition, our framework formalizes the notion of a skill and of an ordered set of skills in terms of the associated data. First, using both synthetic and real data, we demonstrate that these ordered skill sets exist, and that their existence enables more advanced skills to be learned with less data when we train on their prerequisite skills. Second, using our proposed framework, we introduce an online data sampling algorithm, Skill-It, over mixtures of skills for both continual pre-training and fine-tuning regimes, where the objective is to efficiently learn multiple skills in the former and an individual skill in the latter. On the LEGO synthetic in the continual pre-training setting, Skill-It obtains 36.5 points higher accuracy than random sampling. On the Natural Instructions dataset in the fine-tuning setting, Skill-It reduces the validation loss on the target skill by 13.6% versus training on data associated with the target skill itself. We apply our skills framework on the recent RedPajama dataset to continually pre-train a 3B-parameter LM, achieving higher accuracy on the LM Evaluation Harness with 1B tokens than the baseline approach of sampling uniformly over data sources with 3B tokens.