Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others). We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes more complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations.

Large Pre-trained Language Models (PLM) have become the most desirable starting point in the field of NLP, as they have become remarkably good at solving many individual tasks. Despite such success, in this paper, we argue that current paradigms of working with PLMs are neglecting a critical aspect of modeling human intelligence: functional compositionality. Functional compositionality - the ability to compose learned tasks - has been a long-standing challenge in the field of AI (and many other fields) as it is considered one of the hallmarks of human intelligence. An illustrative example of such is cross-lingual summarization, where a bilingual person (English-French) could directly summarize an English document into French sentences without having to translate the English document or summary into French explicitly. We discuss why this matter is an important open problem that requires further attention from the field. Then, we show that current PLMs (e.g., GPT-2 and T5) don't have functional compositionality yet and it is far from human-level generalizability. Finally, we suggest several research directions that could push the field towards zero-shot functional compositionality of language models.

We propose a conceptual perspective on prompts for Large Language Models (LLMs) that distinguishes between (1) diegetic prompts (part of the narrative, e.g. "Once upon a time, I saw a fox..."), and (2) non-diegetic prompts (external, e.g. "Write about the adventures of the fox."). With this lens, we study how 129 crowd workers on Prolific write short texts with different user interfaces (1 vs 3 suggestions, with/out non-diegetic prompts; implemented with GPT-3): When the interface offered multiple suggestions and provided an option for non-diegetic prompting, participants preferred choosing from multiple suggestions over controlling them via non-diegetic prompts. When participants provided non-diegetic prompts it was to ask for inspiration, topics or facts. Single suggestions in particular were guided both with diegetic and non-diegetic information. This work informs human-AI interaction with generative models by revealing that (1) writing non-diegetic prompts requires effort, (2) people combine diegetic and non-diegetic prompting, and (3) they use their draft (i.e. diegetic information) and suggestion timing to strategically guide LLMs.

Large-scale pre-trained multi-modal models (e.g., CLIP) demonstrate strong zeroshot transfer capability in many discriminative tasks, e.g., image classification. Their adaptation to zero-shot image-conditioned text generation tasks has drawn increasing interest. Prior arts approach to zero-shot captioning by either utilizing the existing large language models (e.g., GPT-2) or pre-training the encoderdecoder network in an end-to-end manner. However, the large language models may not generate sensible descriptions due to the task discrepancy between captioning and language modeling, while the end-to-end pre-training requires paired data and extensive computational resources. In this work, we propose a simple framework, named DeCap, for zero-shot captioning. We introduce a lightweight visual-aware language decoder. This decoder is both data-efficient and computation-efficient: 1) it only requires the text data for training, easing the burden on the collection of paired data. When trained with text-only data, the decoder takes the text embedding extracted from the off-the-shelf CLIP encoder as a prefix embedding. The challenge is that the decoder is trained on the text corpus but at the inference stage, it needs to generate captions based on visual inputs. Though the CLIP text embedding and the visual embedding are correlated, the modality gap issue is widely observed in multi-modal contrastive models that prevents us from directly taking the visual embedding as the prefix embedding. We propose a training-free mechanism to reduce the modality gap. We project the visual embedding into the CLIP text embedding space, while the projected embedding retains the information of the visual input. Taking the projected embedding as the prefix embedding, the decoder generates high-quality descriptions that match the visual input. The experiments show that DeCap outperforms other zero-shot captioning methods and unpaired captioning methods by a large margin on the typical image captioning benchmarks, i.e., MSCOCO and NoCaps.

Large language models (LLMs) demonstrate strong reasoning Large language models have been demonstrated to perform capabilities across various domains, including dialogue complex tasks. However, enabling general inference in the (Glaese et al., 2022; Thoppilan et al., 2022), step-by-step real world, e.g. for robotics problems, raises the challenge reasoning (Wei et al., 2022; Kojima et al., 2022), math problem of grounding. We propose embodied language models to directly solving (Lewkowycz et al., 2022; Polu et al., 2022), and incorporate real-world continuous sensor modalities code writing (Chen et al., 2021a). However, a limitation of into language models and thereby establish the link between such models for inference in the real world is the issue of words and percepts. Input to our embodied language grounding: while training LLMs on massive textual data model are multi-modal sentences that interleave visual, continuous may lead to representations that relate to our physical world, state estimation, and textual input encodings. We connecting those representations to real-world visual and train these encodings end-to-end, in conjunction with a pretrained physical sensor modalities is essential to solving a wider large language model, for multiple embodied tasks range of grounded real-world problems in computer vision including sequential robotic manipulation planning, visual and robotics (Tellex et al., 2020).

We examine whether large neural language models, trained on very large collections of varied English text, learn the potentially long-distance dependency of British versus American spelling conventions, i.e., whether spelling is consistently one or the other within model-generated strings. In contrast to long-distance dependencies in non-surface underlying structure (e.g., syntax), spelling consistency is easier to measure both in LMs and the text corpora used to train them, which can provide additional insight into certain observed model behaviors. Using a set of probe words unique to either British or American English, we first establish that training corpora exhibit substantial (though not total) consistency. A large T5 language model does appear to internalize this consistency, though only with respect to observed lexical items (not nonce words with British/American spelling patterns). We further experiment with correcting for biases in the training data by fine-tuning T5 on synthetic data that has been debiased, and find that finetuned T5 remains only somewhat sensitive to spelling consistency. Further experiments show GPT2 to be similarly limited.

In recent years, In-context Learning (ICL) has gained increasing attention and emerged as the new paradigm for large language model (LLM) evaluation. Unlike traditional fine-tuning methods, ICL instead adapts the pre-trained models to unseen tasks without any parameter updates. However, the implementation of ICL is sophisticated due to the diverse retrieval and inference methods involved, as well as the varying pre-processing requirements for different models, datasets, and tasks. A unified and flexible framework for ICL is urgently needed to ease the implementation of the aforementioned components. To facilitate ICL research, we introduce OpenICL, an open-source toolkit for ICL and LLM evaluation. OpenICL is research-friendly with a highly flexible architecture that users can easily combine different components to suit their needs. It also provides various state-of-the-art retrieval and inference methods to streamline the process of adapting ICL to cutting-edge research. The effectiveness of OpenICL has been validated on a wide range of NLP tasks, including classification, QA, machine translation, and semantic parsing. As a side-product, we found OpenICL to be an efficient yet robust tool for LLMs evaluation. OpenICL is released at https://github.com/Shark-NLP/OpenICL

Chain-of-thought (CoT) prompting enables large language models (LLMs) to solve complex reasoning tasks by generating an explanation before the final prediction. Despite it's promising ability, a critical downside of CoT prompting is that the performance is greatly affected by the factuality of the generated explanation. To improve the correctness of the explanations, fine-tuning language models with explanation data is needed. However, there exists only a few datasets that can be used for such approaches, and no data collection tool for building them. Thus, we introduce CoTEVer, a tool-kit for annotating the factual correctness of generated explanations and collecting revision data of wrong explanations. Figure 1: Example of Explanation Verification and Answer Furthermore, we suggest several use cases Verification of GPT-3's output. Explanation Verification where the data collected with CoTEVer can requires additional knowledge which makes it be utilized for enhancing the faithfulness of hard for annotators to intuitively write a revised explanation explanations. Our toolkit is publicly available and answer.

Human models play a crucial role in human-robot interaction (HRI), enabling robots to consider the impact of their actions on people and plan their behavior accordingly. However, crafting good human models is challenging; capturing context-dependent human behavior requires significant prior knowledge and/or large amounts of interaction data, both of which are difficult to obtain. In this work, we explore the potential of large-language models (LLMs) -- which have consumed vast amounts of human-generated text data -- to act as zero-shot human models for HRI. Our experiments on three social datasets yield promising results; the LLMs are able to achieve performance comparable to purpose-built models. That said, we also discuss current limitations, such as sensitivity to prompts and spatial/numerical reasoning mishaps. Based on our findings, we demonstrate how LLM-based human models can be integrated into a social robot's planning process and applied in HRI scenarios. Specifically, we present one case study on a simulated trust-based table-clearing task and replicate past results that relied on custom models. Next, we conduct a new robot utensil-passing experiment (n = 65) where preliminary results show that planning with a LLM-based human model can achieve gains over a basic myopic plan. In summary, our results show that LLMs offer a promising (but incomplete) approach to human modeling for HRI.

When artificial intelligence software like ChatGPT writes, it considers many options for each word, taking into account the response it has written so far and the question being asked. It assigns a score to each option on the list, which quantifies how likely the word is to come next, based on the vast amount of human-written text it has analyzed. ChatGPT, which is built on what is known as a large language model, then chooses a word with a high score, and moves on to the next one. The model's output is often so sophisticated that it can seem like the chatbot understands what it is saying -- but it does not. Every choice it makes is determined by complex math and huge amounts of data.