With the emergence of large language models (LLMs), multimodal models based on LLMs have demonstrated significant potential. Models such as LLaSM, X-LLM, and SpeechGPT exhibit an impressive ability to comprehend and generate human instructions. However, their performance often falters when faced with complex tasks like end-to-end speech translation (E2E-ST), a cross-language and cross-modal translation task. In comparison to single-modal models, multimodal models lag behind in these scenarios. This paper introduces LST, a Large multimodal model designed to excel at the E2E-ST task. LST consists of a speech frontend, an adapter, and a LLM backend. The training of LST consists of two stages: (1) Modality adjustment, where the adapter is tuned to align speech representation with text embedding space, and (2) Downstream task fine-tuning, where both the adapter and LLM model are trained to optimize performance on the E2EST task. Experimental results on the MuST-C speech translation benchmark demonstrate that LST-13B achieves BLEU scores of 30.39/41.55/35.33 on En-De/En-Fr/En-Es language pairs, surpassing previous models and establishing a new state-of-the-art. Additionally, we conduct an in-depth analysis of single-modal model selection and the impact of training strategies, which lays the foundation for future research. We will open up our code and models after review.

While forward reasoning (i.e. find the answer given the question) has been explored extensively in the recent literature, backward reasoning is relatively unexplored. We examine the backward reasoning capabilities of LLMs on Math Word Problems (MWPs): given a mathematical question and its answer, with some details omitted from the question, can LLMs effectively retrieve the missing information? In this paper, we formally define the backward reasoning task on math word problems and modify three datasets to evaluate this task: GSM8k, SVAMP and MultiArith. Our findings show a significant drop in the accuracy of models on backward reasoning compared to forward reasoning across four SOTA LLMs (GPT4, GPT3.5, PaLM-2, and LLaMa-2). Utilizing the specific format of this task, we propose three novel techniques that improve performance: Rephrase reformulates the given problem into a forward reasoning problem, PAL-Tools combines the idea of Program-Aided LLMs to produce a set of equations that can be solved by an external solver, and Check your Work exploits the availability of natural verifier of high accuracy in the forward direction, interleaving solving and verification steps. Finally, realizing that each of our base methods correctly solves a different set of problems, we propose a novel Bayesian formulation for creating an ensemble over these base methods aided by a verifier to further boost the accuracy by a significant margin. Extensive experimentation demonstrates that our techniques successively improve the performance of LLMs on the backward reasoning task, with the final ensemble-based method resulting in a substantial performance gain compared to the raw LLMs with standard prompting techniques such as chain-of-thought.

Visual Word Sense Disambiguation (VWSD) is a novel challenging task that lies between linguistic sense disambiguation and fine-grained multimodal retrieval. The recent advancements in the development of visiolinguistic (VL) transformers suggest some off-the-self implementations with encouraging results, which however we argue that can be further improved. To this end, we propose some knowledge-enhancement techniques towards improving the retrieval performance of VL transformers via the usage of Large Language Models (LLMs) as Knowledge Bases. More specifically, knowledge stored in LLMs is retrieved with the help of appropriate prompts in a zero-shot manner, achieving performance advancements. Moreover, we convert VWSD to a purely textual question-answering (QA) problem by considering generated image captions as multiple-choice candidate answers. Zero-shot and few-shot prompting strategies are leveraged to explore the potential of such a transformation, while Chain-of-Thought (CoT) prompting in the zero-shot setting is able to reveal the internal reasoning steps an LLM follows to select the appropriate candidate. In total, our presented approach is the first one to analyze the merits of exploiting knowledge stored in LLMs in different ways to solve WVSD.

Machine-based prediction of real-world events is garnering attention due to its potential for informed decision-making. Whereas traditional forecasting predominantly hinges on structured data like time-series, recent breakthroughs in language models enable predictions using unstructured text. In particular, (Zou et al., 2022) unveils AutoCast, a new benchmark that employs news articles for answering forecasting queries. Nevertheless, existing methods still trail behind human performance. The cornerstone of accurate forecasting, we argue, lies in identifying a concise, yet rich subset of news snippets from a vast corpus. With this motivation, we introduce AutoCast++, a zero-shot ranking-based context retrieval system, tailored to sift through expansive news document collections for event forecasting. Our approach first re-ranks articles based on zero-shot question-passage relevance, honing in on semantically pertinent news. Following this, the chosen articles are subjected to zero-shot summarization to attain succinct context. Leveraging a pre-trained language model, we conduct both the relevance evaluation and article summarization without needing domain-specific training. Notably, recent articles can sometimes be at odds with preceding ones due to new facts or unanticipated incidents, leading to fluctuating temporal dynamics. To tackle this, our re-ranking mechanism gives preference to more recent articles, and we further regularize the multi-passage representation learning to align with human forecaster responses made on different dates. Empirical results underscore marked improvements across multiple metrics, improving the performance for multiple-choice questions (MCQ) by 48% and true/false (TF) questions by up to 8%.

As of September 2023, ChatGPT correctly answers "what is 7+8" with 15, but when asked "7+8=15, True or False" it responds with "False". This inconsistency between generating and validating an answer is prevalent in language models (LMs) and erodes trust. In this paper, we propose a framework for measuring the consistency between generation and validation (which we call generator-validator consistency, or GV-consistency), finding that even GPT-4, a state-of-the-art LM, is GV-consistent only 76% of the time. To improve the consistency of LMs, we propose to finetune on the filtered generator and validator responses that are GV-consistent, and call this approach consistency fine-tuning. We find that this approach improves GV-consistency of Alpaca-30B from 60% to 93%, and the improvement extrapolates to unseen tasks and domains (e.g., GV-consistency for positive style transfers extrapolates to unseen styles like humor). In addition to improving consistency, consistency fine-tuning improves both generator quality and validator accuracy without using any labeled data. Evaluated across 6 tasks, including math questions, knowledge-intensive QA, and instruction following, our method improves the generator quality by 16% and the validator accuracy by 6.3% across all tasks. Language models (LMs) can generate high-quality responses to task prompts; however, the same model can sometimes produce contradictory responses when validating its own answers. For example, in September 2023, ChatGPT correctly responds to "what is 7+8?" with "15", but when prompted "7+8=15, True or False" it responds with "False" A consistent LM that answers "15" to the generator query should also answer "True" to the validator query, and we call this Figure 1: To measure generatorvalidator consistency between generation and validation generatorvalidator consistency, we prompt a LM consistency or GV-consistency. Then, we check if the language models, and it can be applied to a broad range same LM consistently responds to a corresponding of tasks.

Foundation models have excelled in various tasks but are often evaluated on general benchmarks. The adaptation of these models for specific domains, such as remote sensing imagery, remains an underexplored area. In remote sensing, precise building instance segmentation is vital for applications like urban planning. While Convolutional Neural Networks (CNNs) perform well, their generalization can be limited. For this aim, we present a novel approach to adapt foundation models to address existing models' generalization dropback. Among several models, our focus centers on the Segment Anything Model (SAM), a potent foundation model renowned for its prowess in class-agnostic image segmentation capabilities. We start by identifying the limitations of SAM, revealing its suboptimal performance when applied to remote sensing imagery. Moreover, SAM does not offer recognition abilities and thus fails to classify and tag localized objects. To address these limitations, we introduce different prompting strategies, including integrating a pre-trained CNN as a prompt generator. This novel approach augments SAM with recognition abilities, a first of its kind. We evaluated our method on three remote sensing datasets, including the WHU Buildings dataset, the Massachusetts Buildings dataset, and the AICrowd Mapping Challenge. For out-of-distribution performance on the WHU dataset, we achieve a 5.47% increase in IoU and a 4.81% improvement in F1-score. For in-distribution performance on the WHU dataset, we observe a 2.72% and 1.58% increase in True-Positive-IoU and True-Positive-F1 score, respectively. We intend to release our code repository, hoping to inspire further exploration of foundation models for domain-specific tasks within the remote sensing community.

Informal natural language that describes code functionality, such as code comments or function documentation, may contain substantial information about a programs intent. However, there is typically no guarantee that a programs implementation and natural language documentation are aligned. In the case of a conflict, leveraging information in code-adjacent natural language has the potential to enhance fault localization, debugging, and code trustworthiness. In practice, however, this information is often underutilized due to the inherent ambiguity of natural language which makes natural language intent challenging to check programmatically. The "emergent abilities" of Large Language Models (LLMs) have the potential to facilitate the translation of natural language intent to programmatically checkable assertions. However, it is unclear if LLMs can correctly translate informal natural language specifications into formal specifications that match programmer intent. Additionally, it is unclear if such translation could be useful in practice. In this paper, we describe LLM4nl2post, the problem leveraging LLMs for transforming informal natural language to formal method postconditions, expressed as program assertions. We introduce and validate metrics to measure and compare different LLM4nl2post approaches, using the correctness and discriminative power of generated postconditions. We then perform qualitative and quantitative methods to assess the quality of LLM4nl2post postconditions, finding that they are generally correct and able to discriminate incorrect code. Finally, we find that LLM4nl2post via LLMs has the potential to be helpful in practice; specifications generated from natural language were able to catch 70 real-world historical bugs from Defects4J.

With the emergence of foundation models (Bommasani et al., 2021), fine-tuning the pre-trained feature extractor (FE) has become a low-cost strategy to obtain superior performance in downstream tasks. Notably, GPT-3 (Brown et al., 2020) can achieve state-of-the-art (SOTA) performance on GLUE benchmarks (Wang et al., 2018) via parameterefficient fine-tuning (Hu et al., 2021). Due to the ubiquitous existence of adversarial attacks (Goodfellow et al., 2014; Madry et al., 2018), adopting pre-trained FEs to safety-critical downstream areas such as medicine (Buch et al., 2018) and autonomous cars (Kurakin et al., 2018) necessitates the strategy of robust fine-tuning (Hendrycks et al., 2019) that can yield adversarial robustness in downstream applications. Robust fine-tuning (RFT) (Hendrycks et al., 2019) that contains an adversarial objective to learn features of adversarial data (Madry et al., 2018) can gain adversarial robustness in downstream tasks. To further improve generalization, vanilla RFT (formulated in Eq. 1, shown in the left panel of Figure 1c) optimizes both adversarial and natural objectives to learn the features of adversarial and natural data simultaneously via the FE (Zhang et al., 2019; Shafahi et al., 2019; Jiang et al., 2020).

Large Language Models (LLMs) have emerged as a groundbreaking technology with their unparalleled text generation capabilities across various applications. Nevertheless, concerns persist regarding the accuracy and appropriateness of their generated content. A contemporary methodology, self-correction, has been proposed as a remedy to these issues. Building upon this premise, this paper critically examines the role and efficacy of self-correction within LLMs, shedding light on its true potential and limitations. Central to our investigation is the notion of intrinsic self-correction, whereby an LLM attempts to correct its initial responses based solely on its inherent capabilities, without the crutch of external feedback. In the context of reasoning, our research indicates that LLMs struggle to self-correct their responses without external feedback, and at times, their performance might even degrade post self-correction. Drawing from these insights, we offer suggestions for future research and practical applications in this field.

Expert feedback lays the foundation of rigorous research. However, the rapid growth of scholarly production and intricate knowledge specialization challenge the conventional scientific feedback mechanisms. High-quality peer reviews are increasingly difficult to obtain. Researchers who are more junior or from under-resourced settings have especially hard times getting timely feedback. With the breakthrough of large language models (LLM) such as GPT-4, there is growing interest in using LLMs to generate scientific feedback on research manuscripts. However, the utility of LLM-generated feedback has not been systematically studied. To address this gap, we created an automated pipeline using GPT-4 to provide comments on the full PDFs of scientific papers. We evaluated the quality of GPT-4's feedback through two large-scale studies. We first quantitatively compared GPT-4's generated feedback with human peer reviewer feedback in 15 Nature family journals (3,096 papers in total) and the ICLR machine learning conference (1,709 papers). The overlap in the points raised by GPT-4 and by human reviewers (average overlap 30.85% for Nature journals, 39.23% for ICLR) is comparable to the overlap between two human reviewers (average overlap 28.58% for Nature journals, 35.25% for ICLR). The overlap between GPT-4 and human reviewers is larger for the weaker papers. We then conducted a prospective user study with 308 researchers from 110 US institutions in the field of AI and computational biology to understand how researchers perceive feedback generated by our GPT-4 system on their own papers. Overall, more than half (57.4%) of the users found GPT-4 generated feedback helpful/very helpful and 82.4% found it more beneficial than feedback from at least some human reviewers. While our findings show that LLM-generated feedback can help researchers, we also identify several limitations.