Compositional Zero-shot Learning (CZSL) aims to recognize novel concepts composed of known knowledge without training samples. Standard CZSL either identifies visual primitives or enhances unseen composed entities, and as a result, entanglement between state and object primitives cannot be fully utilized. Admittedly, vision-language models (VLMs) could naturally cope with CZSL through tuning prompts, while uneven entanglement leads prompts to be dragged into local optimum. In this paper, we take a further step to introduce a novel Disentangled and Recurrent Prompt Tuning framework termed DRPT to better tap the potential of VLMs in CZSL. Specifically, the state and object primitives are deemed as learnable tokens of vocabulary embedded in prompts and tuned on seen compositions. Instead of jointly tuning state and object, we devise a disentangled and recurrent tuning strategy to suppress the traction force caused by entanglement and gradually optimize the token parameters, leading to a better prompt space. Notably, we develop a progressive fine-tuning procedure that allows for incremental updates to the prompts, optimizing the object first, then the state, and vice versa. Meanwhile, the optimization of state and object is independent, thus clearer features can be learned to further alleviate the issue of entangling misleading optimization. Moreover, we quantify and analyze the entanglement in CZSL and supplement entanglement rebalancing optimization schemes. DRPT surpasses representative state-of-the-art methods on extensive benchmark datasets, demonstrating superiority in both accuracy and efficiency.

Large Language models (LLMs) are trained on large amounts of data, which can include sensitive information that may compromise personal privacy. LLMs showed to memorize parts of the training data and emit those data verbatim when an adversary prompts appropriately. Previous research has primarily focused on data preprocessing and differential privacy techniques to address memorization or prevent verbatim memorization exclusively, which can give a false sense of privacy. However, these methods rely on explicit and implicit assumptions about the structure of the data to be protected, which often results in an incomplete solution to the problem. To address this, we propose a novel framework that utilizes a reinforcement learning approach (PPO) to fine-tune LLMs to mitigate approximate memorization. Our approach utilizes a negative similarity score, such as BERTScore or SacreBLEU, as a reward signal to learn a dissimilarity policy. Our results demonstrate that this framework effectively mitigates approximate memorization while maintaining high levels of coherence and fluency in the generated samples. Furthermore, our framework is robust in mitigating approximate memorization across various circumstances, including longer context, which is known to increase memorization in LLMs.

In this paper, we release a largest ever medical Question Answering (QA) dataset with 26 million QA pairs. We benchmark many existing approaches in our dataset in terms of both retrieval and generation. Experimental results show that the existing models perform far lower than expected and the released dataset is still challenging in the pre-trained language model era. Moreover, we also experimentally show the benefit of the proposed dataset in many aspects: (i) trained models for other QA datasets in a zero-shot fashion; and (ii) as external knowledge for retrieval-augmented generation (RAG); and (iii) improving existing pre-trained language models by using the QA pairs as a pre-training corpus in continued training manner. We believe that this dataset will not only contribute to medical research but also facilitate both the patients and clinical doctors. See \url{https://github.com/FreedomIntelligence/Huatuo-26M}.

This work explores the problem of generating task graphs of real-world activities. Different from prior formulations, we consider a setting where text transcripts of instructional videos performing a real-world activity (e.g., making coffee) are provided and the goal is to identify the key steps relevant to the task as well as the dependency relationship between these key steps. We propose a novel task graph generation approach that combines the reasoning capabilities of instruction-tuned language models along with clustering and ranking components to generate accurate task graphs in a completely unsupervised manner. We show that the proposed approach generates more accurate task graphs compared to a supervised learning approach on tasks from the ProceL and CrossTask datasets.

Most existing retrieval-augmented language models (LMs) for question answering assume all retrieved information is factually correct. In this work, we study a more realistic scenario in which retrieved documents may contain misinformation, causing conflicts among them. We observe that the existing models are highly brittle to such information in both fine-tuning and in-context few-shot learning settings. We propose approaches to make retrieval-augmented LMs robust to misinformation by explicitly fine-tuning a discriminator or prompting to elicit discrimination capability in GPT-3. Our empirical results on open-domain question answering show that these approaches significantly improve LMs' robustness to knowledge conflicts. We also provide our findings on interleaving the fine-tuned model's decision with the in-context learning process, paving a new path to leverage the best of both worlds.

Fears about AI-mediated communication are grounded in different expectations for one's own versus others' use Contribution statement: MJ and NK conceived of the original idea and carried out Study 1. MJ created the stimuli for Studies 1 and 2 and performed the initial analysis of the Study 1 results. ZP was the principal investigator and led the design of Study 2 and its conceptual integration with Study 1 in consultation with all authors. ZP carried out Study 2 in consultation with all authors. MD oversaw ethical approval and pre-registration for Study 2. ZP ran the power analysis, conducted the analyses, designed the figures, and drafted the manuscript in consultation with MD and AM. All authors provided critical feedback and helped shape the design, analysis, and manuscript. Abstract The rapid development of AI-mediated communication technologies (AICTs) - digital tools that use AI to augment our interpersonal messages - has created concerns about the future of interpersonal trust and prompted urgent discussions about disclosure and uptake. We contribute to this discussion by assessing perceptions about acceptability and use of open and secret AICTs for oneself and others. In two studies with representative samples (UK: N=477, US: N=765), we found (a) that secret (i.e., undisclosed) AICT use is deemed less acceptable than open AICT use, (b) people overestimate others' AICT use, and (c) people expect others to use AICTs irresponsibly. Thus, we raise concerns about the potential for misperceptions and different expectations for others to drive self-fulfilling pessimistic outlooks about AImediated communication.

The'Godfather of Artificial Intelligence' has sensationally resigned from Google and warned the technology could upend life as we know it. Geoffrey Hinton, 75, is credited with creating the technology that became the bedrock of A.I. systems like ChatGPT and Google Bard. But the Turing prize winner now says a part of him regrets helping to make the systems, that he fears could prompt the proliferation of misinformation and replace people in the workforce. He said he had to tell himself excuses like'if I didn't build it, someone else would have' to prevent himself from being overwhelmed by guilt. He drew comparisons with the'father of the atomic bomb' Robert Oppenheimer, who was reportedly distraught by his invention and dedicated the rest of his life to stopping its proliferation.

More than 150 workers whose labor underpins the AI systems of Facebook, TikTok and ChatGPT gathered in Nairobi on Monday and pledged to establish the first African Content Moderators Union, in a move that could have significant consequences for the businesses of some of the world's biggest tech companies. The current and former workers, all employed by third party outsourcing companies, have provided content moderation services for AI tools used by Meta, Bytedance, and OpenAI--the respective owners of Facebook, TikTok and the breakout AI chatbot ChatGPT. Despite the mental toll of the work, which has left many content moderators suffering from PTSD, their jobs are some of the lowest-paid in the global tech industry, with some workers earning as little as $1.50 per hour. As news of the successful vote to register the union was read out, the packed room of workers at the Mövenpick Hotel in Nairobi burst into cheers and applause, a video from the event seen by TIME shows. Confetti fell onto the stage, and jubilant music began to play as the crowd continued to cheer.

For at least a decade, Elon Musk has tried to steer the development of artificial intelligence--only to be outmaneuvered by rivals and former allies. He has now stepped up his efforts after the success of OpenAI, an organization he co-founded but then left after a power struggle. Mr. Musk has warned for years that poorly built artificial intelligence could have catastrophic effects on humanity. Since OpenAI's ChatGPT became a viral sensation last November, Mr. Musk has denounced it as politically correct and warned it could lead AI to become too powerful for humans to control.

An AI-based decoder that can translate brain activity into a continuous stream of text has been developed, in a breakthrough that allows a person's thoughts to be read non-invasively for the first time. The decoder could reconstruct speech with uncanny accuracy while people listened to a story – or even silently imagined one – using only fMRI scan data. Previous language decoding systems have required surgical implants, and the latest advance raises the prospect of new ways to restore speech in patients struggling to communicate due to a stroke or motor neurone disease. Dr Alexander Huth, a neuroscientist who led the work at the University of Texas at Austin, said: "We were kind of shocked that it works as well as it does. I've been working on this for 15 years … so it was shocking and exciting when it finally did work."