Recently in natural language processing (NLP), there has been a paradigm shift from full language model finetuning to the optimization of a small subset of prompt tokens (Shin et al., 2020; Lester et al., 2021; Li and Liang, 2021; Zhong et al., 2021). As language models have dramatically increased in size and may contain billions of parameters (Brown et al., 2020), the strategy of freezing language models while optimizing the learnable prompt parameters becomes the most affordable and efficient alternative for downstream tasks. This technique, referred to as prompt tuning, has gained substantial recognition for its efficacy across a range of language models (Shin et al., 2020; Lester et al., 2021; Li and Liang, 2021; Zhong et al., 2021). Various prompt tuning methods have been explored, which can be generally categorized into discrete and continuous cases. Discrete prompt tuning, such as AutoPrompt (Shin et al., 2020), primarily focuses on the selection and optimization of a predetermined set of tokens within a language model's vocabulary.

The recent success of Large Language Models (LLMs) has gained significant attention in both academia and industry. Substantial efforts have been made to enhance the zero- and few-shot generalization capabilities of open-source LLMs through finetuning. Currently, the prevailing approach is instruction-tuning, which trains LLMs to complete real-world tasks by generating responses guided by natural language instructions. It is worth noticing that such an approach may underperform in sequence and token classification tasks. Unlike text generation tasks, classification tasks have a limited label space, where precise label prediction is more appreciated than generating diverse and human-like responses. Prior research has unveiled that instruction-tuned LLMs cannot outperform BERT, prompting us to explore the potential of leveraging latent representations from LLMs for supervised label prediction. In this paper, we introduce a label-supervised adaptation for LLMs, which aims to finetuning the model with discriminant labels. We evaluate this approach with Label Supervised LLaMA (LS-LLaMA), based on LLaMA-2-7B, a relatively small-scale LLM, and can be finetuned on a single GeForce RTX4090 GPU. We extract latent representations from the final LLaMA layer and project them into the label space to compute the cross-entropy loss. The model is finetuned by Low-Rank Adaptation (LoRA) to minimize this loss. Remarkably, without intricate prompt engineering or external knowledge, LS-LLaMA substantially outperforms LLMs ten times its size in scale and demonstrates consistent improvements compared to robust baselines like BERT-Large and RoBERTa-Large in text classification. Moreover, by removing the causal mask from decoders, LS-unLLaMA achieves the state-of-the-art performance in named entity recognition (NER). Our work will shed light on a novel approach to adapting LLMs for various downstream tasks.

Large Language Models (LLMs) have gained the ability to assimilate human knowledge and facilitate natural language interactions with both humans and other LLMs. However, despite their impressive achievements, LLMs have not made significant advancements in the realm of graph machine learning. This limitation arises because graphs encapsulate distinct relational data, making it challenging to transform them into natural language that LLMs understand. In this paper, we bridge this gap with a novel framework, GraphText, that translates graphs into natural language. GraphText derives a graph-syntax tree for each graph that encapsulates both the node attributes and inter-node relationships. Traversal of the tree yields a graph text sequence, which is then processed by an LLM to treat graph tasks as text generation tasks. Notably, GraphText offers multiple advantages. It introduces training-free graph reasoning: even without training on graph data, GraphText with ChatGPT can achieve on par with, or even surpassing, the performance of supervised-trained graph neural networks through in-context learning (ICL). Furthermore, GraphText paves the way for interactive graph reasoning, allowing both humans and LLMs to communicate with the model seamlessly using natural language. These capabilities underscore the vast, yet-to-be-explored potential of LLMs in the domain of graph machine learning.

What can be learned about causality and experimentation from passive data? This question is salient given recent successes of passively-trained language models in interactive domains such as tool use. Passive learning is inherently limited. However, we show that purely passive learning can in fact allow an agent to learn generalizable strategies for determining and using causal structures, as long as the agent can intervene at test time. We formally illustrate that learning a strategy of first experimenting, then seeking goals, can allow generalization from passive learning in principle. We then show empirically that agents trained via imitation on expert data can indeed generalize at test time to infer and use causal links which are never present in the training data; these agents can also generalize experimentation strategies to novel variable sets never observed in training. We then show that strategies for causal intervention and exploitation can be generalized from passive data even in a more complex environment with high-dimensional observations, with the support of natural language explanations. Explanations can even allow passive learners to generalize out-of-distribution from perfectly-confounded training data. Finally, we show that language models, trained only on passive next-word prediction, can generalize causal intervention strategies from a few-shot prompt containing examples of experimentation, together with explanations and reasoning. These results highlight the surprising power of passive learning of active causal strategies, and may help to understand the behaviors and capabilities of language models.

The evaluation of large language models is an essential task in the field of language understanding and generation. As language models continue to advance, the need for effective benchmarks to assess their performance has become imperative. In the context of Traditional Chinese, there is a scarcity of comprehensive and diverse benchmarks to evaluate the capabilities of language models, despite the existence of certain benchmarks such as DRCD, TTQA, CMDQA, and FGC dataset. To address this gap, we propose a novel set of benchmarks that leverage existing English datasets and are tailored to evaluate language models in Traditional Chinese. These benchmarks encompass a wide range of tasks, including contextual question-answering, summarization, classification, and table understanding. The proposed benchmarks offer a comprehensive evaluation framework, enabling the assessment of language models' capabilities across different tasks. In this paper, we evaluate the performance of GPT-3.5, Taiwan-LLaMa-v1.0, and Model 7-C, our proprietary model, on these benchmarks. The evaluation results highlight that our model, Model 7-C, achieves performance comparable to GPT-3.5 with respect to a part of the evaluated capabilities. In an effort to advance the evaluation of language models in Traditional Chinese and stimulate further research in this field, we have open-sourced our benchmark and opened the model for trial.

Quantifying the impact of training data points is crucial for understanding the outputs of machine learning models and for improving the transparency of the AI pipeline. The influence function is a principled and popular data attribution method, but its computational cost often makes it challenging to use. This issue becomes more pronounced in the setting of large language models and text-to-image models. In this work, we propose DataInf, an efficient influence approximation method that is practical for large-scale generative AI models. Leveraging an easy-to-compute closed-form expression, DataInf outperforms existing influence computation algorithms in terms of computational and memory efficiency. Our theoretical analysis shows that DataInf is particularly well-suited for parameter-efficient fine-tuning techniques such as LoRA. Through systematic empirical evaluations, we show that DataInf accurately approximates influence scores and is orders of magnitude faster than existing methods. In applications to RoBERTa-large, Llama-2-13B-chat, and stable-diffusion-v1.5 models, DataInf effectively identifies the most influential fine-tuning examples better than other approximate influence scores. Moreover, it can help to identify which data points are mislabeled.

Large Language Models (LLMs), including the LLaMA model, have exhibited their efficacy across various general-domain natural language processing (NLP) tasks. However, their performance in high-performance computing (HPC) domain tasks has been less than optimal due to the specialized expertise required to interpret the model responses. In response to this challenge, we propose HPC-GPT, a novel LLaMA-based model that has been supervised fine-tuning using generated QA (Question-Answer) instances for the HPC domain. To evaluate its effectiveness, we concentrate on two HPC tasks: managing AI models and datasets for HPC, and data race detection. By employing HPC-GPT, we demonstrate comparable performance with existing methods on both tasks, exemplifying its excellence in HPC-related scenarios. Our experiments on open-source benchmarks yield extensive results, underscoring HPC-GPT's potential to bridge the performance gap between LLMs and HPC-specific tasks. With HPC-GPT, we aim to pave the way for LLMs to excel in HPC domains, simplifying the utilization of language models in complex computing applications.

The transformer architecture has made breakthroughs in recent years on tasks which require modeling pairwise relationships between sequential elements, as is the case in natural language understanding. However, transformers struggle with long sequences due to the quadratic complexity of the attention operation, and previous research has aimed to lower the complexity by sparsifying or linearly approximating the attention matrix. Yet, these approaches cannot straightforwardly distill knowledge from a teacher's attention matrix, and often require complete retraining from scratch. Furthermore, previous sparse and linear approaches may also lose interpretability if they do not produce full quadratic attention matrices. To address these challenges, we propose SEA: Sparse linear attention with an Estimated Attention mask. SEA estimates the attention matrix with linear complexity via kernel-based linear attention, then creates a sparse approximation to the full attention matrix with a top-k selection to perform a sparse attention operation. For language modeling tasks (Wikitext2), previous linear and sparse attention methods show a roughly two-fold worse perplexity scores over the quadratic OPT-125M baseline, while SEA achieves an even better perplexity than OPT-125M, using roughly half as much memory as OPT-125M. Moreover, SEA maintains an interpretable attention matrix and can utilize knowledge distillation to lower the complexity of existing pretrained transformers. We believe that our work will have a large practical impact, as it opens the possibility of running large transformers on resource-limited devices with less memory.

Empirical software engineering research on production systems has brought forth a better understanding of the software engineering process for practitioners and researchers alike. However, only a small subset of production systems is studied, limiting the impact of this research. While software engineering practitioners benefit from replicating research on their own data, this poses its own set of challenges, since performing replications requires a deep understanding of research methodologies and subtle nuances in software engineering data. Given that large language models (LLMs), such as GPT-4, show promise in tackling both software engineering- and science-related tasks, these models could help democratize empirical software engineering research. In this paper, we examine LLMs' abilities to perform replications of empirical software engineering research on new data. We specifically study their ability to surface assumptions made in empirical software engineering research methodologies, as well as their ability to plan and generate code for analysis pipelines on seven empirical software engineering papers. We perform a user study with 14 participants with software engineering research expertise, who evaluate GPT-4-generated assumptions and analysis plans (i.e., a list of module specifications) from the papers. We find that GPT-4 is able to surface correct assumptions, but struggle to generate ones that reflect common knowledge about software engineering data. In a manual analysis of the generated code, we find that the GPT-4-generated code contains the correct high-level logic, given a subset of the methodology. However, the code contains many small implementation-level errors, reflecting a lack of software engineering knowledge. Our findings have implications for leveraging LLMs for software engineering research as well as practitioner data scientists in software teams.

Fault Localization (FL) aims to automatically localize buggy lines of code, a key first step in many manual and automatic debugging tasks. Previous FL techniques assume the provision of input tests, and often require extensive program analysis, program instrumentation, or data preprocessing. Prior work on deep learning for APR struggles to learn from small datasets and produces limited results on real-world programs. Inspired by the ability of large language models (LLMs) of code to adapt to new tasks based on very few examples, we investigate the applicability of LLMs to line level fault localization. Specifically, we propose to overcome the left-to-right nature of LLMs by fine-tuning a small set of bidirectional adapter layers on top of the representations learned by LLMs to produce LLMAO, the first language model based fault localization approach that locates buggy lines of code without any test coverage information. We fine-tune LLMs with 350 million, 6 billion, and 16 billion parameters on small, manually curated corpora of buggy programs such as the Defects4J corpus. We observe that our technique achieves substantially more confidence in fault localization when built on the larger models, with bug localization performance scaling consistently with the LLM size. Our empirical evaluation shows that LLMAO improves the Top-1 results over the state-of-the-art machine learning fault localization (MLFL) baselines by 2.3%-54.4%, and Top-5 results by 14.4%-35.6%. LLMAO is also the first FL technique trained using a language model architecture that can detect security vulnerabilities down to the code line level.