Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80\% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. We will publicly release MT-bench questions, 3K expert votes, and 30K conversations with human preferences from Chatbot Arena.
相關內容
Generic sentence embeddings provide a coarse-grained approximation of semantic textual similarity but ignore specific aspects that make texts similar. Conversely, aspect-based sentence embeddings provide similarities between texts based on certain predefined aspects. Thus, similarity predictions of texts are more targeted to specific requirements and more easily explainable. In this paper, we present AspectCSE, an approach for aspect-based contrastive learning of sentence embeddings. Results indicate that AspectCSE achieves an average improvement of 3.97% on information retrieval tasks across multiple aspects compared to the previous best results. We also propose using Wikidata knowledge graph properties to train models of multi-aspect sentence embeddings in which multiple specific aspects are simultaneously considered during similarity predictions. We demonstrate that multi-aspect embeddings outperform single-aspect embeddings on aspect-specific information retrieval tasks. Finally, we examine the aspect-based sentence embedding space and demonstrate that embeddings of semantically similar aspect labels are often close, even without explicit similarity training between different aspect labels.
For example, in machine translation tasks, to achieve bidirectional translation between two languages, the source corpus is often used as the target corpus, which involves the training of two models with opposite directions. The question of which one can adapt most quickly to a domain shift is of significant importance in many fields. Specifically, consider an original distribution p that changes due to an unknown intervention, resulting in a modified distribution p*. In aligning p with p*, several factors can affect the adaptation rate, including the causal dependencies between variables in p. In real-life scenarios, however, we have to consider the fairness of the training process, and it is particularly crucial to involve a sensitive variable (bias) present between a cause and an effect variable. To explore this scenario, we examine a simple structural causal model (SCM) with a cause-bias-effect structure, where variable A acts as a sensitive variable between cause (X) and effect (Y). The two models, respectively, exhibit consistent and contrary cause-effect directions in the cause-bias-effect SCM. After conducting unknown interventions on variables within the SCM, we can simulate some kinds of domain shifts for analysis. We then compare the adaptation speeds of two models across four shift scenarios. Additionally, we prove the connection between the adaptation speeds of the two models across all interventions.
Text-to-SQL is a task that converts a natural language question into a structured query language (SQL) to retrieve information from a database. Large language models (LLMs) work well in natural language generation tasks, but they are not specifically pre-trained to understand the syntax and semantics of SQL commands. In this paper, we propose an LLM-based framework for Text-to-SQL which retrieves helpful demonstration examples to prompt LLMs. However, questions with different database schemes can vary widely, even if the intentions behind them are similar and the corresponding SQL queries exhibit similarities. Consequently, it becomes crucial to identify the appropriate SQL demonstrations that align with our requirements. We design a de-semanticization mechanism that extracts question skeletons, allowing us to retrieve similar examples based on their structural similarity. We also model the relationships between question tokens and database schema items (i.e., tables and columns) to filter out scheme-related information. Our framework adapts the range of the database schema in prompts to balance length and valuable information. A fallback mechanism allows for a more detailed schema to be provided if the generated SQL query fails. Ours outperforms state-of-the-art models and demonstrates strong generalization ability on three cross-domain Text-to-SQL benchmarks.
Non-autoregressive approaches aim to improve the inference speed of translation models, particularly those that generate output in a one-pass forward manner. However, these approaches often suffer from a significant drop in translation quality compared to autoregressive models. This paper introduces a series of innovative techniques to enhance the translation quality of Non-Autoregressive Translation (NAT) models while maintaining a substantial acceleration in inference speed. We propose fine-tuning Pretrained Multilingual Language Models (PMLMs) with the CTC loss to train NAT models effectively. Furthermore, we adopt the MASK insertion scheme for up-sampling instead of token duplication, and we present an embedding distillation method to further enhance performance. In our experiments, our model outperforms the baseline autoregressive model (Transformer \textit{base}) on multiple datasets, including WMT'14 DE$\leftrightarrow$EN, WMT'16 RO$\leftrightarrow$EN, and IWSLT'14 DE$\leftrightarrow$EN. Notably, our model achieves better performance than the baseline autoregressive model on the IWSLT'14 En$\leftrightarrow$De and WMT'16 En$\leftrightarrow$Ro datasets, even without using distillation data during training. It is worth highlighting that on the IWSLT'14 DE$\rightarrow$EN dataset, our model achieves an impressive BLEU score of 39.59, setting a new state-of-the-art performance. Additionally, our model exhibits a remarkable speed improvement of 16.35 times compared to the autoregressive model.
We present ToddlerBERTa, a BabyBERTa-like language model, exploring its capabilities through five different models with varied hyperparameters. Evaluating on BLiMP, SuperGLUE, MSGS, and a Supplement benchmark from the BabyLM challenge, we find that smaller models can excel in specific tasks, while larger models perform well with substantial data. Despite training on a smaller dataset, ToddlerBERTa demonstrates commendable performance, rivalling the state-of-the-art RoBERTa-base. The model showcases robust language understanding, even with single-sentence pretraining, and competes with baselines that leverage broader contextual information. Our work provides insights into hyperparameter choices, and data utilization, contributing to the advancement of language models.
Large language models (LMs) beyond a certain scale, demonstrate the emergent capability of generating free-text rationales for their predictions via chain-of-thought (CoT) prompting. While CoT can yield dramatically improved performance, such gains are only observed for sufficiently large LMs. Even more concerning, there is little guarantee that the generated rationales are consistent with LM's predictions or faithfully justify the decisions. In this work, we propose a faithful knowledge distillation method to learn a small, self-consistent CoT model from a teacher model that is orders of magnitude larger. To form better supervision, we elicit rationales supporting the gold answers from a large LM (teacher) by contrastive decoding, which encourages the teacher to generate tokens that become more plausible only when the answer is considered. To ensure faithful distillation, we use the teacher-generated rationales to learn a student LM with a counterfactual reasoning objective, which prevents the student from ignoring the rationales to make inconsistent predictions. Experiments show that, while yielding comparable end-task performance, our method can generate CoT rationales that are more faithful than baselines do. Further analysis suggests that such a model respects the rationales more when making decisions; thus, we can improve its performance more by refining its rationales.
The emergence of large language models (LLMs) has substantially influenced natural language processing, demonstrating exceptional results across various tasks. In this study, we employ ``Introspective Tips" to facilitate LLMs in self-optimizing their decision-making. By introspectively examining trajectories, LLM refines its policy by generating succinct and valuable tips. Our method enhances the agent's performance in both few-shot and zero-shot learning situations by considering three essential scenarios: learning from the agent's past experiences, integrating expert demonstrations, and generalizing across diverse games. Importantly, we accomplish these improvements without fine-tuning the LLM parameters; rather, we adjust the prompt to generalize insights from the three aforementioned situations. Our framework not only supports but also emphasizes the advantage of employing LLM in in-contxt decision-making. Experiments involving over 100 games in TextWorld illustrate the superior performance of our approach.
Transformer-based pretrained language models (T-PTLMs) have achieved great success in almost every NLP task. The evolution of these models started with GPT and BERT. These models are built on the top of transformers, self-supervised learning and transfer learning. Transformed-based PTLMs learn universal language representations from large volumes of text data using self-supervised learning and transfer this knowledge to downstream tasks. These models provide good background knowledge to downstream tasks which avoids training of downstream models from scratch. In this comprehensive survey paper, we initially give a brief overview of self-supervised learning. Next, we explain various core concepts like pretraining, pretraining methods, pretraining tasks, embeddings and downstream adaptation methods. Next, we present a new taxonomy of T-PTLMs and then give brief overview of various benchmarks including both intrinsic and extrinsic. We present a summary of various useful libraries to work with T-PTLMs. Finally, we highlight some of the future research directions which will further improve these models. We strongly believe that this comprehensive survey paper will serve as a good reference to learn the core concepts as well as to stay updated with the recent happenings in T-PTLMs.
For languages with no annotated resources, transferring knowledge from rich-resource languages is an effective solution for named entity recognition (NER). While all existing methods directly transfer from source-learned model to a target language, in this paper, we propose to fine-tune the learned model with a few similar examples given a test case, which could benefit the prediction by leveraging the structural and semantic information conveyed in such similar examples. To this end, we present a meta-learning algorithm to find a good model parameter initialization that could fast adapt to the given test case and propose to construct multiple pseudo-NER tasks for meta-training by computing sentence similarities. To further improve the model's generalization ability across different languages, we introduce a masking scheme and augment the loss function with an additional maximum term during meta-training. We conduct extensive experiments on cross-lingual named entity recognition with minimal resources over five target languages. The results show that our approach significantly outperforms existing state-of-the-art methods across the board.
Pre-trained deep neural network language models such as ELMo, GPT, BERT and XLNet have recently achieved state-of-the-art performance on a variety of language understanding tasks. However, their size makes them impractical for a number of scenarios, especially on mobile and edge devices. In particular, the input word embedding matrix accounts for a significant proportion of the model's memory footprint, due to the large input vocabulary and embedding dimensions. Knowledge distillation techniques have had success at compressing large neural network models, but they are ineffective at yielding student models with vocabularies different from the original teacher models. We introduce a novel knowledge distillation technique for training a student model with a significantly smaller vocabulary as well as lower embedding and hidden state dimensions. Specifically, we employ a dual-training mechanism that trains the teacher and student models simultaneously to obtain optimal word embeddings for the student vocabulary. We combine this approach with learning shared projection matrices that transfer layer-wise knowledge from the teacher model to the student model. Our method is able to compress the BERT_BASE model by more than 60x, with only a minor drop in downstream task metrics, resulting in a language model with a footprint of under 7MB. Experimental results also demonstrate higher compression efficiency and accuracy when compared with other state-of-the-art compression techniques.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191