Model distillation – using outputs from a large teacher model to teach a small student model – is a practical means of creating efficient models for a particular task. We ask: Can we identify a students’ teacher based on its outputs? Such “footprints” left by teacher LLMs would be interesting artifacts. Beyond this, reliable teacher inference may have practical implications as actors seek to distill specific capabilities of massive proprietary LLMs into deployed smaller LMs, potentially violating terms of service. We consider practical task distillation targets including summarization, question answering, and instruction-following. We assume a finite set of candidate teacher models, which we treat as blackboxes. We design discriminative models that operate over lexical features. We find that n-gram similarity alone is unreliable for identifying teachers, but part-of-speech (PoS) templates preferred by student models mimic those of their teachers.

Fine-grained personas have recently been used for generating ‘diverse’ synthetic data for pre-training and supervised fine-tuning of Large Language Models (LLMs). In this work, we measure the diversity of persona-driven synthetically generated prompts and responses with a suite of lexical diversity and redundancy metrics. First, we find that synthetic prompts/instructions are significantly less diverse than human-written ones. Next, we sample responses from LLMs of different sizes with fine-grained and coarse persona descriptions to investigate how much fine-grained detail in persona descriptions contribute to generated text diversity. Our results indicate that persona prompting produces higher lexical diversity than prompting without personas, particularly in larger models. In contrast, adding fine-grained persona details yields minimal gains in diversity compared to simply specifying a length cutoff in the prompt.

The diversity of text can be measured beyond word-level features, however existing diversity evaluation focuses primarily on word-level features. Here we propose a method for evaluating diversity over syntactic features to characterize general repetition in models, beyond frequent n-grams. Specifically, we define syntactic templates (e.g., strings comprising parts-of-speech) and show that models tend to produce templated text in downstream tasks at a higher rate than what is found in human-reference textsWe find that most (76%) templates in model-generated text can be found in pre-training data (compared to only 35% of human-authored text), and are not overwritten during fine-tuning or alignment processes such as RLHF. The connection between templates in generated text and the pre-training data allows us to analyze syntactic templates in models where we do not have the pre-training data.We also find that templates as features are able to differentiate between models, tasks, and domains, and are useful for qualitatively evaluating common model constructions.Finally, we demonstrate the use of templates as a useful tool for analyzing style memorization of training data in LLMs.

Modern instruction-tuned models have become highly capable in text generation tasks such as summarization, and are expected to be released at a steady pace. In practice one may now wish to choose confidently, but with minimal effort, the best performing summarization model when applied to a new domain or purpose. In this work, we empirically investigate the test sample size necessary to select a preferred model in the context of news summarization. Empirical results reveal that comparative evaluation converges quickly for both automatic and human evaluation, with clear preferences for a system emerging from under 100 examples. The human preference data allows us to quantify how well automatic scores can reproduce preference rankings across a variety of downstream summarization tasks. We find that, while automatic metrics are stable at smaller sample sizes, only some automatic metrics are able to moderately predict model win rates according to human preference.

Large language models, particularly GPT-3, are able to produce high quality summaries ofgeneral domain news articles in few- and zero-shot settings. However, it is unclear if such models are similarly capable in more specialized domains such as biomedicine. In this paper we enlist domain experts (individuals with medical training) to evaluate summaries of biomedical articles generated by GPT-3, given no supervision. We consider bothsingle- and multi-document settings. In the former, GPT-3 is tasked with generating regular and plain-language summaries of articles describing randomized controlled trials; in thelatter, we assess the degree to which GPT-3 is able to synthesize evidence reported acrossa collection of articles. We design an annotation scheme for evaluating model outputs, withan emphasis on assessing the factual accuracy of generated summaries. We find that whileGPT-3 is able to summarize and simplify single biomedical articles faithfully, it strugglesto provide accurate aggregations of findings over multiple documents. We release all data,code, and annotations used in this work.

Clinical prediction models often use structured variables and provide outcomes that are not readily interpretable by clinicians. Further, free-text medical notes may contain information not immediately available in structured variables. We propose a hierarchical CNN-transformer model with explicit attention as an interpretable, multi-task clinical language model, which achieves an AUROC of 0.75 and 0.78 on sepsis and mortality prediction, respectively. We also explore the relationships between learned features from structured and unstructured variables using projection-weighted canonical correlation analysis. Finally, we outline a protocol to evaluate model usability in a clinical decision support context. From domain-expert evaluations, our model generates informative rationales that have promising real-life applications.