Many approaches to Natural Language Processing tasks often treat them as single-step problems, where an agent receives an instruction, executes it, and is evaluated based on the final outcome. However, language is inherently interactive, as evidenced by the back-and-forth nature of human conversations. In light of this, we posit that human-AI collaboration should also be interactive, with humans monitoring the work of AI agents and providing feedback that the agent can understand and utilize. Further, the AI agent should be able to detect when it needs additional information and proactively ask for help. Enabling this scenario would lead to more natural, efficient, and engaging human-AI collaboration.In this paper, we investigate these directions using the challenging task established by the IGLU competition, an interactive grounded language understanding task in a MineCraft-like world. We delve into multiple types of help players can give to the AI to guide it and analyze the impact of this help on behavior, resulting in performance improvements and an end-to-end interactive system.

The remarkable abilities of large language models (LLMs) like ChatGPT and GPT-4 partially stem from the post-training processes involving human preferences encoded within a reward model as part of a Reinforcement Learning from Human Feedback (RLHF) regimen. These reward models (RMs) often lack direct knowledge of why, or under what principles, the preferences annotations were made. In this study, we identify principles that guide RMs to better align with human preferences, and then develop an axiomatic framework to generate a rich variety of preference signals to uphold them. We use these axiomatic signals to train a model for the scoring answers to longform questions. Our approach yields a Preference Model with only about 220M parameters that agrees with gold human-annotated preference labels more often than GPT-4. The contributions of this work include: training a standalone preference model that can score human- and LLM-generated answers on the same scale; developing an axiomatic framework for generating training data pairs tailored to certain principles; and showing that a small amount of axiomatic signals can help small models outperform GPT-4 in preference scoring. We intend to release our axiomatic data and model.

Although adapting pre-trained language models with few examples has shown promising performance on text classification, there is a lack of understanding of where the performance gain comes from. In this work, we propose to answer this question by interpreting the adaptation behavior using post-hoc explanations from model predictions. By modeling feature statistics of explanations, we discover that (1) without fine-tuning, pre-trained models (e.g. BERT and RoBERTa) show strong prediction bias across labels; (2) although few-shot fine-tuning can mitigate the prediction bias and demonstrate promising prediction performance, our analysis shows models gain performance improvement by capturing non-task-related features (e.g. stop words) or shallow data patterns (e.g. lexical overlaps). These observations alert that pursuing model performance with fewer examples may incur pathological prediction behavior, which requires further sanity check on model predictions and careful design in model evaluations in few-shot fine-tuning.

Text summarization helps readers capture salient information from documents, news, interviews, and meetings. However, most state-of-the-art pretrained language models (LM) are unable to efficiently process long text for many summarization tasks. In this paper, we propose Summ^N, a simple, flexible, and effective multi-stage framework for input texts that are longer than the maximum context length of typical pretrained LMs. Summ^N first splits the data samples and generates a coarse summary in multiple stages and then produces the final fine-grained summary based on it. Our framework can process input text of arbitrary length by adjusting the number of stages while keeping the LM input size fixed. Moreover, it can deal with both single-source documents and dialogues, and it can be used on top of different backbone abstractive summarization models. To the best of our knowledge, Summ^N is the first multi-stage split-then-summarize framework for long input summarization. Our experiments demonstrate that Summ^N outperforms previous state-of-the-art methods by improving ROUGE scores on three long meeting summarization datasets AMI, ICSI, and QMSum, two long TV series datasets from SummScreen, and a long document summarization dataset GovReport. Our data and code are available at https://github.com/psunlpgroup/Summ-N.

Transformer-based models have achieved state-of-the-art performance on short-input summarization. However, they still struggle with summarizing longer text. In this paper, we present DYLE, a novel dynamic latent extraction approach for abstractive long-input summarization. DYLE jointly trains an extractor and a generator and treats the extracted text snippets as the latent variable, allowing dynamic snippet-level attention weights during decoding. To provide adequate supervision, we propose simple yet effective heuristics for oracle extraction as well as a consistency loss term, which encourages the extractor to approximate the averaged dynamic weights predicted by the generator. We evaluate our method on different long-document and long-dialogue summarization tasks: GovReport, QMSum, and arXiv. Experiment results show that DYLE outperforms all existing methods on GovReport and QMSum, with gains up to 6.1 ROUGE, while yielding strong results on arXiv. Further analysis shows that the proposed dynamic weights provide interpretability of our generation process.

We present a new method LiST for efficient fine-tuning of large pre-trained language models (PLMs) in few-shot learning settings. LiST improves over recent methods that adopt prompt-based fine-tuning (FN) using two key techniques. The first is the use of self-training to leverage large amounts of unlabeled data for prompt-based FN in few-shot settings. We use self-training in conjunction with meta-learning for re-weighting noisy pseudo-prompt labels. Traditionally, self-training is expensive as it requires updating all the model parameters repetitively. Therefore, we use a second technique for light-weight fine-tuning where we introduce a small number of task-specific parameters that are fine-tuned during self-training while keeping the PLM encoder frozen. Our experiments show that LiST can effectively leverage unlabeled data to improve the model performance for few-shot learning. Additionally, the finetuning process is efficient as it only updates a small percentage of the parameters and the overall model footprint is reduced since several tasks can share a common PLM encoder as backbone. We present a comprehensive study on six NLU tasks to validate the effectiveness of LiST. The results show that LiST improves by 35% over classic fine-tuning methods and 6% over prompt-based FN with 96% reduction in number of trainable parameters when fine-tuned with no more than 30 labeled examples from each task. With only 14M tunable parameters, LiST outperforms GPT-3 in-context learning by 33% on few-shot NLU tasks

Building machine learning models for natural language understanding (NLU) tasks relies heavily on labeled data. Weak supervision has been proven valuable when large amount of labeled data is unavailable or expensive to obtain. Existing works studying weak supervision for NLU either mostly focus on a specific task or simulate weak supervision signals from ground-truth labels. It is thus hard to compare different approaches and evaluate the benefit of weak supervision without access to a unified and systematic benchmark with diverse tasks and real-world weak labeling rules. In this paper, we propose such a benchmark, named WALNUT, to advocate and facilitate research on weak supervision for NLU. WALNUT consists of NLU tasks with different types, including document-level and token-level prediction tasks. WALNUT is the first semi-weakly supervised learning benchmark for NLU, where each task contains weak labels generated by multiple real-world weak sources, together with a small set of clean labels. We conduct baseline evaluations on WALNUT to systematically evaluate the effectiveness of various weak supervision methods and model architectures. Our results demonstrate the benefit of weak supervision for low-resource NLU tasks and highlight interesting patterns across tasks. We expect WALNUT to stimulate further research on methodologies to leverage weak supervision more effectively. The benchmark and code for baselines are available at aka.ms/walnut_benchmark.