When prompting language models to complete a task, users often leave important aspects unsaid. While asking questions could resolve this ambiguity (GATE; Li et al., 2023), models often struggle to ask good questions. We explore a language model's ability to self-improve (STaR; Zelikman et al., 2022) by rewarding the model for generating useful questions-a simple method we dub STaR-GATE. We generate a synthetic dataset of 25,500 unique persona-task prompts to simulate conversations between a pretrained language model-the Questioner-and a Roleplayer whose preferences are unknown to the Questioner. By asking questions, the Questioner elicits preferences from the Roleplayer. The Questioner is iteratively finetuned on questions that increase the probability of high-quality responses to the task, which are generated by an Oracle with access to the Roleplayer's latent preferences. After two iterations of self-improvement, the Questioner asks better questions, allowing it to generate responses that are preferred over responses from the initial model on 72% of tasks. Our results indicate that teaching a language model to ask better questions leads to better personalized responses.

This paper presents the results of the SHROOM, a shared task focused on detecting hallucinations: outputs from natural language generation (NLG) systems that are fluent, yet inaccurate. Such cases of overgeneration put in jeopardy many NLG applications, where correctness is often mission-critical. The shared task was conducted with a newly constructed dataset of 4000 model outputs labeled by 5 annotators each, spanning 3 NLP tasks: machine translation, paraphrase generation and definition modeling. The shared task was tackled by a total of 58 different users grouped in 42 teams, out of which 27 elected to write a system description paper; collectively, they submitted over 300 prediction sets on both tracks of the shared task. We observe a number of key trends in how this approach was tackled -- many participants rely on a handful of model, and often rely either on synthetic data for fine-tuning or zero-shot prompting strategies. While a majority of the teams did outperform our proposed baseline system, the performances of top-scoring systems are still consistent with a random handling of the more challenging items.

In real-world applications, users express different behaviors when they interact with different items, including implicit click/like interactions, and explicit comments/reviews interactions. Nevertheless, almost all recommender works are focused on how to describe user preferences by the implicit click/like interactions, to find the synergy of people. For the content-based explicit comments/reviews interactions, some works attempt to utilize them to mine the semantic knowledge to enhance recommender models. However, they still neglect the following two points: (1) The content semantic is a universal world knowledge; how do we extract the multi-aspect semantic information to empower different domains? (2) The user/item ID feature is a fundamental element for recommender models; how do we align the ID and content semantic feature space? In this paper, we propose a `plugin' semantic knowledge transferring method \textbf{LoID}, which includes two major components: (1) LoRA-based large language model pretraining to extract multi-aspect semantic information; (2) ID-based contrastive objective to align their feature spaces. We conduct extensive experiments with SOTA baselines on real-world datasets, the detailed results demonstrating significant improvements of our method LoID.

Multilingual Language Models (MLLMs) exhibit robust cross-lingual transfer capabilities, or the ability to leverage information acquired in a source language and apply it to a target language. These capabilities find practical applications in well-established Natural Language Processing (NLP) tasks such as Named Entity Recognition (NER). This study aims to investigate the effectiveness of a source language when applied to a target language, particularly in the context of perturbing the input test set. We evaluate on 13 pairs of languages, each including one high-resource language (HRL) and one low-resource language (LRL) with a geographic, genetic, or borrowing relationship. We evaluate two well-known MLLMs--MBERT and XLM-R--on these pairs, in native LRL and cross-lingual transfer settings, in two tasks, under a set of different perturbations. Our findings indicate that NER cross-lingual transfer depends largely on the overlap of entity chunks. If a source and target language have more entities in common, the transfer ability is stronger. Models using cross-lingual transfer also appear to be somewhat more robust to certain perturbations of the input, perhaps indicating an ability to leverage stronger representations derived from the HRL. Our research provides valuable insights into cross-lingual transfer and its implications for NLP applications, and underscores the need to consider linguistic nuances and potential limitations when employing MLLMs across distinct languages.

Explainability algorithms aimed at interpreting decision-making AI systems usually consider balancing two critical dimensions: 1) \textit{faithfulness}, where explanations accurately reflect the model's inference process. 2) \textit{plausibility}, where explanations are consistent with domain experts. However, the question arises: do faithfulness and plausibility inherently conflict? In this study, through a comprehensive quantitative comparison between the explanations from the selected explainability methods and expert-level interpretations across three NLP tasks: sentiment analysis, intent detection, and topic labeling, we demonstrate that traditional perturbation-based methods Shapley value and LIME could attain greater faithfulness and plausibility. Our findings suggest that rather than optimizing for one dimension at the expense of the other, we could seek to optimize explainability algorithms with dual objectives to achieve high levels of accuracy and user accessibility in their explanations.

Text embedding models represent natural language as dense vectors, positioning semantically similar text near each other within the embedding space (Gao et al., 2021; Le and Mikolov, 2014; Reimers and Gurevych, 2019). These embeddings are commonly used for a wide range of downstream tasks including document retrieval, sentence similarity, classification, and clustering (Muennighoff et al., 2023). Instead of building separate embedding models for each downstream task, recent efforts seek to create a single embedding model supporting many tasks. The recent development of general-purpose text embedding models presents a challenge: these models require large amounts of training data to comprehensively cover desired domains and skills. Recent embedding efforts have focused on using extensive collections of training examples (Li et al., 2023; Wang et al., 2022).

Recently, various parameter-efficient fine-tuning (PEFT) strategies for application to language models have been proposed and successfully implemented. However, this raises the question of whether PEFT, which only updates a limited set of model parameters, constitutes security vulnerabilities when confronted with weight-poisoning backdoor attacks. In this study, we show that PEFT is more susceptible to weight-poisoning backdoor attacks compared to the full-parameter fine-tuning method, with pre-defined triggers remaining exploitable and pre-defined targets maintaining high confidence, even after fine-tuning. Motivated by this insight, we developed a Poisoned Sample Identification Module (PSIM) leveraging PEFT, which identifies poisoned samples through confidence, providing robust defense against weight-poisoning backdoor attacks. Specifically, we leverage PEFT to train the PSIM with randomly reset sample labels. During the inference process, extreme confidence serves as an indicator for poisoned samples, while others are clean. We conduct experiments on text classification tasks, five fine-tuning strategies, and three weight-poisoning backdoor attack methods. Experiments show near 100% success rates for weight-poisoning backdoor attacks when utilizing PEFT. Furthermore, our defensive approach exhibits overall competitive performance in mitigating weight-poisoning backdoor attacks.

People from different social and demographic groups express diverse perspectives and conflicting opinions on a broad set of topics such as product reviews, healthcare, law, and politics. A fair summary should provide a comprehensive coverage of diverse perspectives without underrepresenting certain groups. However, current work in summarization metrics and Large Language Models (LLMs) evaluation has not explored fair abstractive summarization. In this paper, we systematically investigate fair abstractive summarization for user-generated data. We first formally define fairness in abstractive summarization as not underrepresenting perspectives of any groups of people, and we propose four reference-free automatic metrics by measuring the differences between target and source perspectives. We evaluate nine LLMs, including three GPT models, four LLaMA models, PaLM 2, and Claude, on six datasets collected from social media, online reviews, and recorded transcripts. Experiments show that both the model-generated and the human-written reference summaries suffer from low fairness. We conduct a comprehensive analysis of the common factors influencing fairness and propose three simple but effective methods to alleviate unfair summarization. Our dataset and code are available at https://github.com/psunlpgroup/FairSumm.

Research on Large Language Models (LLMs) has recently witnessed an increasing interest in extending models' context size to better capture dependencies within long documents. While benchmarks have been proposed to assess long-range abilities, existing efforts primarily considered generic tasks that are not necessarily aligned with real-world applications. In contrast, our work proposes a new benchmark for long-context LLMs focused on a practical meeting assistant scenario. In this scenario, the long contexts consist of transcripts obtained by automatic speech recognition, presenting unique challenges for LLMs due to the inherent noisiness and oral nature of such data. Our benchmark, named ELITR-Bench, augments the existing ELITR corpus' transcripts with 271 manually crafted questions and their ground-truth answers. Our experiments with recent long-context LLMs on ELITR-Bench highlight a gap between open-source and proprietary models, especially when questions are asked sequentially within a conversation. We also provide a thorough analysis of our GPT-4-based evaluation method, encompassing insights from a crowdsourcing study. Our findings suggest that while GPT-4's evaluation scores are correlated with human judges', its ability to differentiate among more than three score levels may be limited.

Interactive Artificial Intelligent(AI) assistant systems are designed to offer timely guidance to help human users to complete a variety tasks. One of the remaining challenges is to understand user's mental states during the task for more personalized guidance. In this work, we analyze users' mental states during task executions and investigate the capabilities and challenges for large language models to interpret user profiles for more personalized user guidance. In the digital age, there is immense potential for artificial intelligent (AI) assistant to guides users through complex tasks, from changing laptop batteries to piping frosting on a cake. One of the main challenges, however, lies in creating an interactive system that can not only understand which step the user is at, but can also detect user's mental states, such as frustration, familiarity with the task, detail-orientation, etc.