As large language models (LLMs) are adopted as a fundamental component of language technologies, it is crucial to accurately characterize their performance. Because choices in prompt design can strongly influence model behavior, this design process is critical in effectively using any modern pre-trained generative language model. In this work, we focus on LLM sensitivity to a quintessential class of meaning-preserving design choices: prompt formatting. We find that several widely used open-source LLMs are extremely sensitive to subtle changes in prompt formatting in few-shot settings, with performance differences of up to 76 accuracy points when evaluated using LLaMA-2-13B. Sensitivity remains even when increasing model size, the number of few-shot examples, or performing instruction tuning. Our analysis suggests that work evaluating LLMs with prompting-based methods would benefit from reporting a range of performance across plausible prompt formats, instead of the currently-standard practice of reporting performance on a single format. We also show that format performance only weakly correlates between models, which puts into question the methodological validity of comparing models with an arbitrarily chosen, fixed prompt format. To facilitate systematic analysis we propose FormatSpread, an algorithm that rapidly evaluates a sampled set of plausible prompt formats for a given task, and reports the interval of expected performance without accessing model weights. Furthermore, we present a suite of analyses that characterize the nature of this sensitivity, including exploring the influence of particular atomic perturbations and the internal representation of particular formats.
相關內容
Large language models (LLMs) have demonstrated impressive capabilities in storing and recalling factual knowledge, but also in adapting to novel in-context information. Yet, the mechanisms underlying their in-context grounding remain unknown, especially in situations where in-context information contradicts factual knowledge embedded in the parameters. This is critical for retrieval-augmented generation methods, which enrich the context with up-to-date information, hoping that grounding can rectify the outdated parametric knowledge. In this study, we introduce Fakepedia, a counterfactual dataset designed to evaluate grounding abilities when the parametric knowledge clashes with the in-context information. We benchmark various LLMs with Fakepedia and discover that GPT-4-turbo has a strong preference for its parametric knowledge. Mistral-7B, on the contrary, is the model that most robustly chooses the grounded answer. Then, we conduct causal mediation analysis on LLM components when answering Fakepedia queries. We demonstrate that inspection of the computational graph alone can predict LLM grounding with 92.8% accuracy, especially because few MLPs in the Transformer can predict non-grounded behavior. Our results, together with existing findings about factual recall mechanisms, provide a coherent narrative of how grounding and factual recall mechanisms interact within LLMs.
Pre-trained large language models have significantly improved code generation. As these models scale up, there is an increasing need for the output to handle more intricate tasks and to be appropriately specialized to particular domains. Here, we target bioinformatics due to the amount of specialized domain knowledge, algorithms, and data operations this discipline requires. We present BioCoder, a benchmark developed to evaluate large language models (LLMs) in generating bioinformatics-specific code. BioCoder spans a broad spectrum of the field and covers cross-file dependencies, class declarations, and global variables. It incorporates 1026 Python functions and 1243 Java methods extracted from GitHub, along with 253 examples from the Rosalind Project, all pertaining to bioinformatics. Using topic modeling we show that overall coverage of the included code is representative of the full spectrum of bioinformatics calculations. BioCoder incorporates a fuzz-testing framework for evaluation. We have applied it to evaluate many models including InCoder, CodeGen, CodeGen2, SantaCoder, StarCoder, StarCoder+, InstructCodeT5+, GPT-3.5, and GPT-4. Furthermore, we finetuned StarCoder, demonstrating how our dataset can effectively enhance the performance of LLMs on our benchmark (by >15% in terms of Pass@K in certain prompt configurations and always >3%). The results highlight two key aspects of successful models: (1) Successful models accommodate a long prompt (> ~2600 tokens) with full context, for functional dependencies. (2) They contain specific domain knowledge of bioinformatics, beyond just general coding knowledge. This is evident from the performance gain of GPT-3.5/4 compared to the smaller models on the benchmark (50% vs up to ~25%). Our dataset, benchmark, Docker images, and scripts required for testing are all available at //github.com/gersteinlab/biocoder.
The rapid development of large language models (LLMs), such as ChatGPT, has revolutionized the efficiency of creating programming tutorials. LLMs can be instructed with text prompts to generate comprehensive text descriptions of code snippets. However, the lack of transparency in the end-to-end generation process has hindered the understanding of model behavior and limited user control over the generated results. To tackle this challenge, we introduce a novel approach that breaks down the programming tutorial creation task into actionable steps. By employing the tree-of-thought method, LLMs engage in an exploratory process to generate diverse and faithful programming tutorials. We then present SPROUT, an authoring tool equipped with a series of interactive visualizations that empower users to have greater control and understanding of the programming tutorial creation process. A formal user study demonstrated the effectiveness of SPROUT, showing that our tool assists users to actively participate in the programming tutorial creation process, leading to more reliable and customizable results. By providing users with greater control and understanding, SPROUT enhances the user experience and improves the overall quality of programming tutorial. A free copy of this paper and all supplemental materials are available at //osf.io/uez2t/?view_only=5102e958802341daa4f86.
Large language model (LLM) powered chatbots are primarily text-based today, and impose a large interactional cognitive load, especially for exploratory or sensemaking tasks such as planning a trip or learning about a new city. Because the interaction is textual, users have little scaffolding in the way of structure, informational "scent", or ability to specify high-level preferences or goals. We introduce ExploreLLM that allows users to structure thoughts, help explore different options, navigate through the choices and recommendations, and to more easily steer models to generate more personalized responses. We conduct a user study and show that users find it helpful to use ExploreLLM for exploratory or planning tasks, because it provides a useful schema-like structure to the task, and guides users in planning. The study also suggests that users can more easily personalize responses with high-level preferences with ExploreLLM. Together, ExploreLLM points to a future where users interact with LLMs beyond the form of chatbots, and instead designed to support complex user tasks with a tighter integration between natural language and graphical user interfaces.
Large language models of code (Code-LLMs) have recently brought tremendous advances to code completion, a fundamental feature of programming assistance and code intelligence. However, most existing works ignore the possible presence of bugs in the code context for generation, which are inevitable in software development. Therefore, we introduce and study the buggy-code completion problem, inspired by the realistic scenario of real-time code suggestion where the code context contains potential bugs -- anti-patterns that can become bugs in the completed program. To systematically study the task, we introduce two datasets: one with synthetic bugs derived from semantics-altering operator changes (buggy-HumanEval) and one with realistic bugs derived from user submissions to coding problems (buggy-FixEval). We find that the presence of potential bugs significantly degrades the generation performance of the high-performing Code-LLMs. For instance, the passing rates of CODEGEN-2B-MONO on test cases of buggy-HumanEval drop more than 50% given a single potential bug in the context. Finally, we investigate several post-hoc methods for mitigating the adverse effect of potential bugs and find that there remains a significant gap in post-mitigation performance.
Adversarial testing of large language models (LLMs) is crucial for their safe and responsible deployment. We introduce a novel approach for automated generation of adversarial evaluation datasets to test the safety of LLM generations on new downstream applications. We call it AI-assisted Red-Teaming (AART) - an automated alternative to current manual red-teaming efforts. AART offers a data generation and augmentation pipeline of reusable and customizable recipes that reduce human effort significantly and enable integration of adversarial testing earlier in new product development. AART generates evaluation datasets with high diversity of content characteristics critical for effective adversarial testing (e.g. sensitive and harmful concepts, specific to a wide range of cultural and geographic regions and application scenarios). The data generation is steered by AI-assisted recipes to define, scope and prioritize diversity within the application context. This feeds into a structured LLM-generation process that scales up evaluation priorities. Compared to some state-of-the-art tools, AART shows promising results in terms of concept coverage and data quality.
Although large language models (LLMs) are impressive in solving various tasks, they can quickly be outdated after deployment. Maintaining their up-to-date status is a pressing concern in the current era. This paper provides a comprehensive review of recent advances in aligning LLMs with the ever-changing world knowledge without re-training from scratch. We categorize research works systemically and provide in-depth comparisons and discussion. We also discuss existing challenges and highlight future directions to facilitate research in this field. We release the paper list at //github.com/hyintell/awesome-refreshing-llms
In pace with developments in the research field of artificial intelligence, knowledge graphs (KGs) have attracted a surge of interest from both academia and industry. As a representation of semantic relations between entities, KGs have proven to be particularly relevant for natural language processing (NLP), experiencing a rapid spread and wide adoption within recent years. Given the increasing amount of research work in this area, several KG-related approaches have been surveyed in the NLP research community. However, a comprehensive study that categorizes established topics and reviews the maturity of individual research streams remains absent to this day. Contributing to closing this gap, we systematically analyzed 507 papers from the literature on KGs in NLP. Our survey encompasses a multifaceted review of tasks, research types, and contributions. As a result, we present a structured overview of the research landscape, provide a taxonomy of tasks, summarize our findings, and highlight directions for future work.
Knowledge enhanced pre-trained language models (K-PLMs) are shown to be effective for many public tasks in the literature but few of them have been successfully applied in practice. To address this problem, we propose K-AID, a systematic approach that includes a low-cost knowledge acquisition process for acquiring domain knowledge, an effective knowledge infusion module for improving model performance, and a knowledge distillation component for reducing the model size and deploying K-PLMs on resource-restricted devices (e.g., CPU) for real-world application. Importantly, instead of capturing entity knowledge like the majority of existing K-PLMs, our approach captures relational knowledge, which contributes to better-improving sentence-level text classification and text matching tasks that play a key role in question answering (QA). We conducted a set of experiments on five text classification tasks and three text matching tasks from three domains, namely E-commerce, Government, and Film&TV, and performed online A/B tests in E-commerce. Experimental results show that our approach is able to achieve substantial improvement on sentence-level question answering tasks and bring beneficial business value in industrial settings.
Many tasks in natural language processing can be viewed as multi-label classification problems. However, most of the existing models are trained with the standard cross-entropy loss function and use a fixed prediction policy (e.g., a threshold of 0.5) for all the labels, which completely ignores the complexity and dependencies among different labels. In this paper, we propose a meta-learning method to capture these complex label dependencies. More specifically, our method utilizes a meta-learner to jointly learn the training policies and prediction policies for different labels. The training policies are then used to train the classifier with the cross-entropy loss function, and the prediction policies are further implemented for prediction. Experimental results on fine-grained entity typing and text classification demonstrate that our proposed method can obtain more accurate multi-label classification results.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191