Causal discovery is becoming a key part in medical AI research. These methods can enhance healthcare by identifying causal links between biomarkers, demographics, treatments and outcomes. They can aid medical professionals in choosing more impactful treatments and strategies. In parallel, Large Language Models (LLMs) have shown great potential in identifying patterns and generating insights from text data. In this paper we investigate applying LLMs to the problem of determining the directionality of edges in causal discovery. Specifically, we test our approach on a deidentified set of Non Small Cell Lung Cancer(NSCLC) patients that have both electronic health record and genomic panel data. Graphs are validated using Bayesian Dirichlet estimators using tabular data. Our result shows that LLMs can accurately predict the directionality of edges in causal graphs, outperforming existing state-of-the-art methods. These findings suggests that LLMs can play a significant role in advancing causal discovery and help us better understand complex systems.

While text style transfer has many applications across natural language processing, the core premise of transferring from a single source style is unrealistic in a real-world setting. In this work, we focus on arbitrary style transfer: rewriting a text from an arbitrary, unknown style to a target style. We propose STEER: Unified Style Transfer with Expert Reinforcement, a unified frame-work developed to overcome the challenge of limited parallel data for style transfer. STEER involves automatically generating a corpus of style-transfer pairs using a product of experts during decoding. The generated offline data is then used to pre-train an initial policy before switching to online, off-policy reinforcement learning for further improvements via fine-grained reward signals. STEER is unified and can transfer to multiple target styles from an arbitrary, unknown source style, making it particularly flexible and efficient. Experimental results on a challenging dataset with text from a diverse set of styles demonstrate state-of-the-art results compared to competitive baselines. Remarkably, STEER outperforms the 175B parameter instruction-tuned GPT-3 on overall style transfer quality, despite being 226 times smaller in size. We also show STEER is robust, maintaining its style transfer capabilities on out-of-domain data, and surpassing nearly all baselines across various styles. The success of our method highlights the potential of RL algorithms when augmented with controllable decoding to overcome the challenge of limited data supervision.

This paper aims to address a critical challenge in robotics, which is enabling them to operate seamlessly in human environments through natural language interactions. Our primary focus is to equip robots with the ability to understand and execute complex instructions in coherent dialogs to facilitate intricate task-solving scenarios. To explore this, we build upon the Execution from Dialog History (EDH) task from the Teach benchmark. We employ a multi-transformer model with BART LM. We observe that our best configuration outperforms the baseline with a success rate score of 8.85 and a goal-conditioned success rate score of 14.02. In addition, we suggest an alternative methodology for completing this task. Moreover, we introduce a new task by expanding the EDH task and making predictions about game plans instead of individual actions. We have evaluated multiple BART models and an LLaMA2 LLM, which has achieved a ROGUE-L score of 46.77 for this task.

To mitigate the potential misuse of large language models (LLMs), recent research has developed watermarking algorithms, which restrict the generation process to leave an invisible trace for watermark detection. Due to the two-stage nature of the task, most studies evaluate the generation and detection separately, thereby presenting a challenge in unbiased, thorough, and applicable evaluations. In this paper, we introduce WaterBench, the first comprehensive benchmark for LLM watermarks, in which we design three crucial factors: (1) For \textbf{benchmarking procedure}, to ensure an apples-to-apples comparison, we first adjust each watermarking method's hyper-parameter to reach the same watermarking strength, then jointly evaluate their generation and detection performance. (2) For \textbf{task selection}, we diversify the input and output length to form a five-category taxonomy, covering $9$ tasks. (3) For \textbf{evaluation metric}, we adopt the GPT4-Judge for automatically evaluating the decline of instruction-following abilities after watermarking. We evaluate $4$ open-source watermarks on $2$ LLMs under $2$ watermarking strengths and observe the common struggles for current methods on maintaining the generation quality. The code and data are available at \url{https://github.com/THU-KEG/WaterBench}.

Language model (LM) prompting--a popular paradigm for solving NLP tasks-- has been shown to be susceptible to miscalibration and brittleness to slight prompt variations, caused by its discriminative prompting approach, i.e., predicting the label given the input. The framework is multivariate, as label descriptions allow us to seamlessly integrate additional contextual information about the labels to improve task performance. On various standard classification benchmarks, with six open-source LM families, we show that zero-shot classification with simple contextualization of the data source of the evaluation set consistently outperforms both zero-shot and few-shot baselines while improving robustness to prompt variations. Further, our approach enables personalizing classification in a zero-shot manner by incorporating author, subject, or reader information in the label descriptions. Language models, trained only on raw text, have been shown to perform new tasks simply by conditioning on a handful of demonstrations (Brown et al., 2020). Furthermore, ICL has been shown to be very sensitive to the choice of training examples, their order and format in the prompt (Lu et al., 2022; Sorensen et al., 2022) requiring major human effort to achieve optimal performance. In this work, we ask, "If the right demonstrations are challenging to find and only serve to implicitly prime the model, can we achieve the same performance zero-shot if we prime the language model explicitly in a robust way?" Our approach consists of two key ideas. First, most text classification methods follow a discriminative setup, which involves estimating the probability of the labels given the input, which can be sensitive to prompt or verbalizer variations. Instead, we use a generative setup, which involves estimating the probability of generating the input given different labels, which has been shown to have better worst-case performance (Min et al., 2022a).

Large language models (LLMs) have shown remarkable capabilities in various natural language understanding tasks. With only a few demonstration examples, these LLMs can quickly adapt to target tasks without expensive gradient updates. Common strategies to boost such "in-context" learning ability are to ensemble multiple model decoded results and require the model to generate an explanation along with the prediction. However, these models often treat different class predictions equally and neglect the potential discrepancy between the explanations and predictions. SE, an Explanation-Aware Soft Ensemble framework to empower in-context learning with LLMs. We design two techniques, explanation-guided ensemble, and soft probability aggregation, to mitigate the effect of unreliable explanations and improve the consistency between explanations and final predictions. Experiments on seven natural language understanding tasks and four varying-size LLMs demonstrate the effectiveness of our proposed framework. Recent advancements in Natural Language Processing (NLP) have witnessed the remarkable capabilities of Large Language Models (LLMs) (Brown et al., 2020; Tay et al., 2023; Chowdhery et al., 2022; Anil et al., 2023; Touvron et al., 2023; OpenAI, 2023). These LLMs can rapidly adapt to new tasks by learning only on a few input-output pairs (a.k.a. Yet, beyond those demonstrations, a significant facet of human learning revolves around explanations. Consequently, the integration of free-text explanations into LLM prompting holds great potentials to further enhance in-context learning performance. Recent studies have examined how to incorporate free-text explanations into LLM in-context learning scheme. For instance, the Predict-then-Explain pipeline (Lampinen et al., 2022) proposes to generate the explanation after making the prediction.

Text-based misinformation permeates online discourses, yet evidence of people's ability to discern truth from such deceptive textual content is scarce. We analyze a novel TV game show data where conversations in a high-stake environment between individuals with conflicting objectives result in lies. We investigate the manifestation of potentially verifiable language cues of deception in the presence of objective truth, a distinguishing feature absent in previous text-based deception datasets. We show that there exists a class of detectors (algorithms) that have similar truth detection performance compared to human subjects, even when the former accesses only the language cues while the latter engages in conversations with complete access to all potential sources of cues (language and audio-visual). Our model, built on a large language model, employs a bottleneck framework to learn discernible cues to determine truth, an act of reasoning in which human subjects often perform poorly, even with incentives. Our model detects novel but accurate language cues in many cases where humans failed to detect deception, opening up the possibility of humans collaborating with algorithms and ameliorating their ability to detect the truth.

Engaging in the deliberate generation of abnormal outputs from large language models (LLMs) by attacking them is a novel human activity. This paper presents a thorough exposition of how and why people perform such attacks. Using a formal qualitative methodology, we interviewed dozens of practitioners from a broad range of backgrounds, all contributors to this novel work of attempting to cause LLMs to fail. We relate and connect this activity between its practitioners' motivations and goals; the strategies and techniques they deploy; and the crucial role the community plays. As a result, this paper presents a grounded theory of how and why people attack large language models: LLM red teaming in the wild.

Current approaches to automatic summarization of scientific papers generate informative summaries in the form of abstracts. However, abstracts are not intended to show the relationship between a paper and the references cited in it. We propose a new contextualized summarization approach that can generate an informative summary conditioned on a given sentence containing the citation of a reference (a so-called "citance"). This summary outlines the content of the cited paper relevant to the citation location. Thus, our approach extracts and models the citances of a paper, retrieves relevant passages from cited papers, and generates abstractive summaries tailored to each citance. We evaluate our approach using $\textbf{Webis-Context-SciSumm-2023}$, a new dataset containing 540K~computer science papers and 4.6M~citances therein.

Large Language Models (LLMs) have reshaped natural language processing with their impressive capabilities. Their ever-increasing size, however, raised concerns about their effective deployment and the need for LLM compressions. This study introduces the Divergent Token metrics (DTMs), a novel approach for assessing compressed LLMs, addressing the limitations of traditional perplexity or accuracy measures that fail to accurately reflect text generation quality. DTMs focus on token divergence, that allow deeper insights into the subtleties of model compression, i.p. when evaluating component's impacts individually. Utilizing the First Divergent Token metric (FDTM) in model sparsification reveals that a quarter of all attention components can be pruned beyond 90% on the Llama-2 model family, still keeping SOTA performance. For quantization FDTM suggests that over 80% of parameters can naively be transformed to int8 without special outlier management. These evaluations indicate the necessity of choosing appropriate compressions for parameters individually-and that FDTM can identify those-while standard metrics result in deteriorated outcomes.