While language models (LMs) have proven remarkably adept at generating code, many programs are challenging for LMs to generate using their parametric knowledge alone. Providing external contexts such as library documentation can facilitate generating accurate and functional code. Despite the success of retrieval-augmented generation (RAG) in various text-oriented tasks, its potential for improving code generation remains under-explored. In this work, we conduct a systematic, large-scale analysis by asking: in what scenarios can retrieval benefit code generation models? and what challenges remain? We first curate a comprehensive evaluation benchmark, CodeRAG-Bench, encompassing three categories of code generation tasks, including basic programming, open-domain, and repository-level problems. We aggregate documents from five sources for models to retrieve contexts: competition solutions, online tutorials, library documentation, StackOverflow posts, and GitHub repositories. We examine top-performing models on CodeRAG-Bench by providing contexts retrieved from one or multiple sources. While notable gains are made in final code generation by retrieving high-quality contexts across various settings, our analysis reveals room for improvement -- current retrievers still struggle to fetch useful contexts especially with limited lexical overlap, and generators fail to improve with limited context lengths or abilities to integrate additional contexts. We hope CodeRAG-Bench serves as an effective testbed to encourage further development of advanced code-oriented RAG methods.

With the emergence of large language models (LLMs) and their ability to perform a variety of tasks, their application in recommender systems (RecSys) has shown promise. However, we are facing significant challenges when deploying LLMs into RecSys, such as limited prompt length, unstructured item information, and un-constrained generation of recommendations, leading to sub-optimal performance. To address these issues, we propose a novel method using a taxonomy dictionary. This method provides a systematic framework for categorizing and organizing items, improving the clarity and structure of item information. By incorporating the taxonomy dictionary into LLM prompts, we achieve efficient token utilization and controlled feature generation, leading to more accurate and contextually relevant recommendations. Our Taxonomy-guided Recommendation (TaxRec) approach features a two-step process: one-time taxonomy categorization and LLM-based recommendation, enabling zero-shot recommendations without the need for domain-specific fine-tuning. Experimental results demonstrate TaxRec significantly enhances recommendation quality compared to traditional zero-shot approaches, showcasing its efficacy as personal recommender with LLMs. Code is available at https://github.com/yueqingliang1/TaxRec.

In this report, we present TAGLAS, an atlas of text-attributed graph (TAG) datasets and benchmarks. TAGs are graphs with node and edge features represented in text, which have recently gained wide applicability in training graph-language or graph foundation models. In TAGLAS, we collect and integrate more than 23 TAG datasets with domains ranging from citation graphs to molecule graphs and tasks from node classification to graph question-answering. Unlike previous graph datasets and benchmarks, all datasets in TAGLAS have a unified node and edge text feature format, which allows a graph model to be simultaneously trained and evaluated on multiple datasets from various domains. Further, we provide a standardized, efficient, and simplified way to load all datasets and tasks. We also provide useful utils like text-to-embedding conversion, and graph-to-text conversion, which can facilitate different evaluation scenarios. Finally, we also provide standard and easy-to-use evaluation utils. The project is open-sourced at https://github.com/JiaruiFeng/TAGLAS and is still under construction. Please expect more datasets/features in the future.

Language agency is an important aspect of evaluating social biases in texts. While several studies approached agency-related bias in human-written language, very limited research has investigated such biases in Large Language Model (LLM)-generated content. In addition, previous research often relies on string-matching techniques to identify agentic and communal words within texts, which fall short of accurately classifying language agency. We introduce the novel Language Agency Bias Evaluation (LABE) benchmark, which comprehensively evaluates biases in LLMs by analyzing agency levels attributed to different demographic groups in model generations. LABE leverages 5,400 template-based prompts, an accurate agency classifier, and corresponding bias metrics to test for gender, racial, and intersectional language agency biases in LLMs on 3 text generation tasks: biographies, professor reviews, and reference letters. To build better and more accurate automated agency classifiers, we also contribute and release the Language Agency Classification (LAC) dataset, consisting of 3,724 agentic and communal sentences. Using LABE, we unveil previously under-explored language agency social biases in 3 recent LLMs: ChatGPT, Llama3, and Mistral. We observe that: (1) For the same text category, LLM generations demonstrate higher levels of gender bias than human-written texts; (2) On most generation tasks, models show remarkably higher levels of intersectional bias than the other bias aspects. Those who are at the intersection of gender and racial minority groups -- such as Black females -- are consistently described by texts with lower levels of agency; (3) Among the 3 LLMs investigated, Llama3 demonstrates greatest overall bias in language agency; (4) Not only does prompt-based mitigation fail to resolve language agency bias in LLMs, but it frequently leads to the exacerbation of biases in generated texts.

Scientific Natural Language Inference (NLI) is the task of predicting the semantic relation between a pair of sentences extracted from research articles. The automatic annotation method based on distant supervision for the training set of SciNLI (Sadat and Caragea, 2022b), the first and most popular dataset for this task, results in label noise which inevitably degenerates the performance of classifiers. In this paper, we propose a novel co-training method that assigns weights based on the training dynamics of the classifiers to the distantly supervised labels, reflective of the manner they are used in the subsequent training epochs. That is, unlike the existing semi-supervised learning (SSL) approaches, we consider the historical behavior of the classifiers to evaluate the quality of the automatically annotated labels. Furthermore, by assigning importance weights instead of filtering out examples based on an arbitrary threshold on the predicted confidence, we maximize the usage of automatically labeled data, while ensuring that the noisy labels have a minimal impact on model training. The proposed method obtains an improvement of 1.5% in Macro F1 over the distant supervision baseline, and substantial improvements over several other strong SSL baselines. We make our code and data available on Github.

Ilya Sutskever, a co-founder and former chief scientist of OpenAI, announced on Wednesday that he's launching a new venture dubbed Safe Superintelligence Inc. Sutskever said on X that the new lab will focus solely on building a safe "superintelligence"--an industry term for a hypothetical system that's smarter than humans. Sutskever is joined at Safe SuperIntelligence Inc. by co-founders Daniel Gross, an investor and engineer who worked on AI at Apple till 2017, and Daniel Levy, another former OpenAI employee. The new American-based firm will have offices in Palo Alto, Calif., and Tel Aviv, according to a description Sutskever shared. I am starting a new company: https://t.co/BG3K3SI3A1 Sutskever was one of OpenAI's founding members, and was chief scientist during the company's meteoric rise following the release of ChatGPT.

An enduring problem with today's generative artificial intelligence (AI) tools, like ChatGPT, is that they often confidently assert false information. Computer scientists call this behavior "hallucination," and it's a key barrier to AI's usefulness. Hallucinations have led to some embarrassing public slip-ups. In February, AirCanada was forced by a tribunal to honor a discount that its customer-support chatbot had mistakenly offered to a passenger. In May, Google was forced to make changes to its new "AI overviews" search feature, after the bot told some users that it was safe to eat rocks. And last June, two lawyers were fined 5,000 by a U.S. judge after one of them admitted he had used ChatGPT to help write a court filing.

For four years, Jacob Hilton worked for one of the most influential startups in the Bay Area -- OpenAI. His research helped test and improve the truthfulness of AI models such as ChatGPT. He believes artificial intelligence can benefit society, but he also recognizes the serious risks if the technology is left unchecked. Hilton was among 13 current and former OpenAI and Google employees who this month signed an open letter that called for more whistleblower protections, citing broad confidentiality agreements as problematic. "The basic situation is that employees, the people closest to the technology, they're also the ones with the most to lose from being retaliated against for speaking up," says Hilton, 33, now a researcher at the nonprofit Alignment Research Center, who lives in Berkeley.

Last week, Apple held its Worldwide Developers Conference, the annual event that is often used to showcase the company's most significant innovations. Much of the presentation this year was devoted to A.I., or, as the company is branding it, Apple Intelligence. Whereas Google and Microsoft have leaped headlong into A.I. with their Gemini and OpenAI products, respectively, Apple is so far taking a narrower approach. The A.I. model it is unveiling on iPhone hardware is relatively weak. A.I. models are measured on their number of "parameters," or the variables adjusted during the training process; while OpenAI's GPT-4 has more than one and a half trillion parameters, Apple's model has three billion.

The short-video platform, which has over 600 million active users, announced the new tool on June 6. Like OpenAI's Sora model, Kling is able to generate videos "up to two minutes long with a frame rate of 30fps and video resolution up to 1080p," the company says on its website. But unlike Sora, which still remains inaccessible to the public four months after OpenAI trialed it, Kling soon started letting people try the model themselves. I got access to it after downloading Kuaishou's video-editing tool, signing up with a Chinese number, getting on a waitlist, and filling out an additional form through Kuaishou's user feedback groups. The model can't process prompts written entirely in English, but you can get around that by either translating the phrase you want to use into Chinese or including one or two Chinese words.