We report the development of Japanese SimCSE, Japanese sentence embedding models fine-tuned with SimCSE. Since there is a lack of sentence embedding models for Japanese that can be used as a baseline in sentence embedding research, we conducted extensive experiments on Japanese sentence embeddings involving 24 pre-trained Japanese or multilingual language models, five supervised datasets, and four unsupervised datasets. In this report, we provide the detailed training setup for Japanese SimCSE and their evaluation results.

In this technical report, we present Skywork-13B, a family of large language models (LLMs) trained on a corpus of over 3.2 trillion tokens drawn from both English and Chinese texts. This bilingual foundation model is the most extensively trained and openly published LLMs of comparable size to date. We introduce a two-stage training methodology using a segmented corpus, targeting general purpose training and then domain-specific enhancement training, respectively. We show that our model not only excels on popular benchmarks, but also achieves \emph{state of the art} performance in Chinese language modeling on diverse domains. Furthermore, we propose a novel leakage detection method, demonstrating that test data contamination is a pressing issue warranting further investigation by the LLM community. To spur future research, we release Skywork-13B along with checkpoints obtained during intermediate stages of the training process. We are also releasing part of our SkyPile corpus, a collection of over 150 billion tokens of web text, which is the largest high quality open Chinese pre-training corpus to date. We hope Skywork-13B and our open corpus will serve as a valuable open-source resource to democratize access to high-quality LLMs.

Answering time-sensitive questions from long documents requires temporal reasoning over the times in questions and documents. An important open question is whether large language models can perform such reasoning solely using a provided text document, or whether they can benefit from additional temporal information extracted using other systems. We address this research question by applying existing temporal information extraction systems to construct temporal graphs of events, times, and temporal relations in questions and documents. We then investigate different approaches for fusing these graphs into Transformer models. Experimental results show that our proposed approach for fusing temporal graphs into input text substantially enhances the temporal reasoning capabilities of Transformer models with or without fine-tuning. Additionally, our proposed method outperforms various graph convolution-based approaches and establishes a new state-of-the-art performance on SituatedQA and three splits of TimeQA.

A significant increase in content creation and information exchange has been made possible by the quick development of online social media platforms, which has been very advantageous. However, these platforms have also become a haven for those who disseminate false information, propaganda, and fake news. Claims are essential in forming our perceptions of the world, but sadly, they are frequently used to trick people by those who spread false information. To address this problem, social media giants employ content moderators to filter out fake news from the actual world. However, the sheer volume of information makes it difficult to identify fake news effectively. Therefore, it has become crucial to automatically identify social media posts that make such claims, check their veracity, and differentiate between credible and false claims. In response, we presented CLAIMSCAN in the 2023 Forum for Information Retrieval Evaluation (FIRE'2023). The primary objectives centered on two crucial tasks: Task A, determining whether a social media post constitutes a claim, and Task B, precisely identifying the words or phrases within the post that form the claim. Task A received 40 registrations, demonstrating a strong interest and engagement in this timely challenge. Meanwhile, Task B attracted participation from 28 teams, highlighting its significance in the digital era of misinformation.

Evaluating text-to-image models is notoriously difficult. A strong recent approach for assessing text-image faithfulness is based on QG/A (question generation and answering), which uses pre-trained foundational models to automatically generate a set of questions and answers from the prompt, and output images are scored based on whether these answers extracted with a visual question answering model are consistent with the prompt-based answers. This kind of evaluation is naturally dependent on the quality of the underlying QG and QA models. We identify and address several reliability challenges in existing QG/A work: (a) QG questions should respect the prompt (avoiding hallucinations, duplications, and omissions) and (b) VQA answers should be consistent (not asserting that there is no motorcycle in an image while also claiming the motorcycle is blue). We address these issues with Davidsonian Scene Graph (DSG), an empirically grounded evaluation framework inspired by formal semantics. DSG is an automatic, graph-based QG/A that is modularly implemented to be adaptable to any QG/A module. DSG produces atomic and unique questions organized in dependency graphs, which (i) ensure appropriate semantic coverage and (ii) sidestep inconsistent answers. With extensive experimentation and human evaluation on a range of model configurations (LLM, VQA, and T2I), we empirically demonstrate that DSG addresses the challenges noted above. Finally, we present DSG-1k, an open-sourced evaluation benchmark that includes 1,060 prompts, covering a wide range of fine-grained semantic categories with a balanced distribution. We release the DSG-1k prompts and the corresponding DSG questions.

The rapid advancement of large language models (LLMs) has sparked interest in data synthesis techniques, aiming to generate diverse and high-quality synthetic datasets. However, these synthetic datasets often suffer from a lack of diversity and added noise. In this paper, we present TarGEN, a multi-step prompting strategy for generating high-quality synthetic datasets utilizing a LLM. An advantage of TarGEN is its seedless nature; it does not require specific task instances, broadening its applicability beyond task replication. We augment TarGEN with a method known as self-correction empowering LLMs to rectify inaccurately labeled instances during dataset creation, ensuring reliable labels. To assess our technique's effectiveness, we emulate 8 tasks from the SuperGLUE benchmark and finetune various language models, including encoder-only, encoder-decoder, and decoder-only models on both synthetic and original training sets. Evaluation on the original test set reveals that models trained on datasets generated by TarGEN perform approximately 1-2% points better than those trained on original datasets (82.84% via syn. vs. 81.12% on og. using Flan-T5). When incorporating instruction tuning, the performance increases to 84.54% on synthetic data vs. 81.49% on original data by Flan-T5. A comprehensive analysis of the synthetic dataset compared to the original dataset reveals that the synthetic dataset demonstrates similar or higher levels of dataset complexity and diversity. Furthermore, the synthetic dataset displays a bias level that aligns closely with the original dataset. Finally, when pre-finetuned on our synthetic SuperGLUE dataset, T5-3B yields impressive results on the OpenLLM leaderboard, surpassing the model trained on the Self-Instruct dataset by 4.14% points. We hope that TarGEN can be helpful for quality data generation and reducing the human efforts to create complex benchmarks.

Emotion recognition in text, the task of identifying emotions such as joy or anger, is a challenging problem in NLP with many applications. One of the challenges is the shortage of available datasets that have been annotated with emotions. Certain existing datasets are small, follow different emotion taxonomies and display imbalance in their emotion distribution. In this work, we studied the impact of data augmentation techniques precisely when applied to small imbalanced datasets, for which current state-of-the-art models (such as RoBERTa) under-perform. Specifically, we utilized four data augmentation methods (Easy Data Augmentation EDA, static and contextual Embedding-based, and ProtAugment) on three datasets that come from different sources and vary in size, emotion categories and distributions. Our experimental results show that using the augmented data when training the classifier model leads to significant improvements. Finally, we conducted two case studies: a) directly using the popular chat-GPT API to paraphrase text using different prompts, and b) using external data to augment the training set. Results show the promising potential of these methods.

This paper delves into the capabilities of large language models (LLMs), specifically focusing on advancing the theoretical comprehension of chain-of-thought prompting. We investigate how LLMs can be effectively induced to generate a coherent chain of thoughts. To achieve this, we introduce a two-level hierarchical graphical model tailored for natural language generation. Within this framework, we establish a compelling geometrical convergence rate that gauges the likelihood of an LLM-generated chain of thoughts compared to those originating from the true language. Our findings provide a theoretical justification for the ability of LLMs to produce the correct sequence of thoughts (potentially) explaining performance gains in tasks demanding reasoning skills.

Previous studies have relied on existing question-answering benchmarks to evaluate the knowledge stored in large language models (LLMs). However, this approach has limitations regarding factual knowledge coverage, as it mostly focuses on generic domains which may overlap with the pretraining data. This paper proposes a framework to systematically assess the factual knowledge of LLMs by leveraging knowledge graphs (KGs). Our framework automatically generates a set of questions and expected answers from the facts stored in a given KG, and then evaluates the accuracy of LLMs in answering these questions. We systematically evaluate the state-of-the-art LLMs with KGs in generic and specific domains. The experiment shows that ChatGPT is consistently the top performer across all domains. We also find that LLMs performance depends on the instruction finetuning, domain and question complexity and is prone to adversarial context.

Optimization problems are pervasive across various sectors, from manufacturing and distribution to healthcare. However, most such problems are still solved heuristically by hand rather than optimally by state-of-the-art solvers, as the expertise required to formulate and solve these problems limits the widespread adoption of optimization tools and techniques. We introduce OptiMUS, a Large Language Model (LLM)-based agent designed to formulate and solve MILP problems from their natural language descriptions. OptiMUS is capable of developing mathematical models, writing and debugging solver code, developing tests, and checking the validity of generated solutions. To benchmark our agent, we present NLP4LP, a novel dataset of linear programming (LP) and mixed integer linear programming (MILP) problems. Our experiments demonstrate that OptiMUS solves nearly twice as many problems as a basic LLM prompting strategy. OptiMUS code and NLP4LP dataset are available at \href{https://github.com/teshnizi/OptiMUS}{https://github.com/teshnizi/OptiMUS}