GSM-Plus: A Comprehensive Benchmark for Evaluating the Robustness of LLMs as Mathematical Problem Solvers

Large language models (LLMs) have achieved impressive performance across various mathematical reasoning benchmarks. However, there are increasing debates regarding whether these models truly understand and apply mathematical knowledge or merely rely on shortcuts for mathematical reasoning. One essential and frequently occurring evidence is that when the math questions are slightly changed, LLMs can behave incorrectly. This motivates us to evaluate the robustness of LLMs' math reasoning capability by testing a wide range of question variations. We introduce the adversarial grade school math (\datasetname) dataset, an extension of GSM8K augmented with various mathematical perturbations. Our experiments on 25 LLMs and 4 prompting techniques show that while LLMs exhibit different levels of math reasoning abilities, their performances are far from robust. In particular, even for problems that have been solved in GSM8K, LLMs can make mistakes when new statements are added or the question targets are altered. We also explore whether more robust performance can be achieved by composing existing prompting methods, in which we try an iterative method that generates and verifies each intermediate thought based on its reasoning goal and calculation result. Code and data are available at \url{https://github.com/qtli/GSM-Plus}.

BBA: Bi-Modal Behavioral Alignment for Reasoning with Large Vision-Language Models

Multimodal reasoning stands as a pivotal capability for large vision-language models (LVLMs). The integration with Domain-Specific Languages (DSL), offering precise visual representations, equips these models with the opportunity to execute more accurate reasoning in complex and professional domains. However, the vanilla Chain-of-Thought (CoT) prompting method faces challenges in effectively leveraging the unique strengths of visual and DSL representations, primarily due to their differing reasoning mechanisms. Additionally, it often falls short in addressing critical steps in multi-step reasoning tasks. To mitigate these challenges, we introduce the \underline{B}i-Modal \underline{B}ehavioral \underline{A}lignment (BBA) prompting method, designed to maximize the potential of DSL in augmenting complex multi-modal reasoning tasks. This method initiates by guiding LVLMs to create separate reasoning chains for visual and DSL representations. Subsequently, it aligns these chains by addressing any inconsistencies, thus achieving a cohesive integration of behaviors from different modalities. Our experiments demonstrate that BBA substantially improves the performance of GPT-4V(ision) on geometry problem solving ($28.34\% \to 34.22\%$), chess positional advantage prediction ($42.08\% \to 46.99\%$) and molecular property prediction ($77.47\% \to 83.52\%$).

Collaborative Evaluation: Exploring the Synergy of Large Language Models and Humans for Open-ended Generation Evaluation

Humans are widely involved in the evaluation of open-ended natural language generation tasks (NLG) that demand creativity, as automatic metrics often exhibit weak correlations with human judgments. Large language models (LLMs) recently have emerged as a scalable and cost-effective alternative to human evaluations. However, both humans and LLMs have limitations, i.e., inherent subjectivity and unreliable judgments, particularly for open-ended tasks that require adaptable metrics tailored to diverse task requirements. To explore the synergy between humans and LLM-based evaluators and address the challenges of existing inconsistent evaluation criteria in open-ended NLG tasks, we propose a Collaborative Evaluation pipeline CoEval, involving the design of a checklist of task-specific criteria and the detailed evaluation of texts, in which LLM generates initial ideation, and then humans engage in scrutiny. We conducted a series of experiments to investigate the mutual effects between LLMs and humans in CoEval. Results show that, by utilizing LLMs, CoEval effectively evaluates lengthy texts, saving significant time and reducing human evaluation outliers. Human scrutiny still plays a role, revising around 20% of LLM evaluation scores for ultimate reliability.

Deepfake Text Detection in the Wild

Recent advances in large language models have enabled them to reach a level of text generation comparable to that of humans. These models show powerful capabilities across a wide range of content, including news article writing, story generation, and scientific writing. Such capability further narrows the gap between human-authored and machine-generated texts, highlighting the importance of deepfake text detection to avoid potential risks such as fake news propagation and plagiarism. However, previous work has been limited in that they testify methods on testbed of specific domains or certain language models. In practical scenarios, the detector faces texts from various domains or LLMs without knowing their sources. To this end, we build a wild testbed by gathering texts from various human writings and deepfake texts generated by different LLMs. Human annotators are only slightly better than random guessing at identifying machine-generated texts. Empirical results on automatic detection methods further showcase the challenges of deepfake text detection in a wild testbed. In addition, out-of-distribution poses a greater challenge for a detector to be employed in realistic application scenarios. We release our resources at https://github.com/yafuly/DeepfakeTextDetect.

A Cognitive Stimulation Dialogue System with Multi-source Knowledge Fusion for Elders with Cognitive Impairment

When communicating with elders with cognitive impairment, cognitive stimulation (CS) help to maintain the cognitive health of elders. Data sparsity is the main challenge in building CS-based dialogue systems, particularly in the Chinese language. To fill this gap, we construct a Chinese CS conversation (CSConv) dataset, which contains about 2.6K groups of dialogues with CS principles and emotional support strategy labels. Making chit chat while providing emotional support is overlooked by the majority of existing cognitive dialogue systems. In this paper, we propose a multi-source knowledge fusion method for CS dialogue (CSD), to generate open-ended responses guided by the CS principle and emotional support strategy. We first use a progressive mask method based on external knowledge to learn encoders as effective classifiers, which is the prerequisite to predict the CS principle and emotional support strategy of the target response. Then a decoder interacts with the perceived CS principle and emotional support strategy to generate responses. Extensive experiments conducted on the CSConv dataset demonstrate the effectiveness of the proposed method, while there is still a large space for improvement compared to human performance.

Retrieved Sequence Augmentation for Protein Representation Learning

Protein language models have excelled in a variety of tasks, ranging from structure prediction to protein engineering. However, proteins are highly diverse in functions and structures, and current state-of-the-art models including the latest version of AlphaFold rely on Multiple Sequence Alignments (MSA) to feed in the evolutionary knowledge. Despite their success, heavy computational overheads, as well as the de novo and orphan proteins remain great challenges in protein representation learning. In this work, we show that MSAaugmented models inherently belong to retrievalaugmented methods. Motivated by this finding, we introduce Retrieved Sequence Augmentation(RSA) for protein representation learning without additional alignment or pre-processing. RSA links query protein sequences to a set of sequences with similar structures or properties in the database and combines these sequences for downstream prediction. We show that protein language models benefit from the retrieval enhancement on both structure prediction and property prediction tasks, with a 5% improvement on MSA Transformer on average while being 373 times faster. In addition, we show that our model can transfer to new protein domains better and outperforms MSA Transformer on de novo protein prediction. Our study fills a much-encountered gap in protein prediction and brings us a step closer to demystifying the domain knowledge needed to understand protein sequences. Code is available on https://github.com/HKUNLP/RSA.

Explanation Regeneration via Information Bottleneck

Explaining the black-box predictions of NLP models naturally and accurately is an important open problem in natural language generation. These free-text explanations are expected to contain sufficient and carefully-selected evidence to form supportive arguments for predictions. Due to the superior generative capacity of large pretrained language models, recent work built on prompt engineering enables explanation generation without specific training. However, explanation generated through single-pass prompting often lacks sufficiency and conciseness. To address this problem, we develop an information bottleneck method EIB to produce refined explanations that are sufficient and concise. Our approach regenerates the free-text explanation by polishing the single-pass output from the pretrained language model but retaining the information that supports the contents being explained. Experiments on two out-of-domain tasks verify the effectiveness of EIB through automatic evaluation and thoroughly-conducted human evaluation.

Event Transition Planning for Open-ended Text Generation

Open-ended text generation tasks, such as dialogue generation and story completion, require models to generate a coherent continuation given limited preceding context. The open-ended nature of these tasks brings new challenges to the neural auto-regressive text generators nowadays. Despite these neural models are good at producing human-like text, it is difficult for them to arrange causalities and relations between given facts and possible ensuing events. To bridge this gap, we propose a novel two-stage method which explicitly arranges the ensuing events in open-ended text generation. Our approach can be understood as a specially-trained coarse-to-fine algorithm, where an event transition planner provides a "coarse" plot skeleton and a text generator in the second stage refines the skeleton. Experiments on two open-ended text generation tasks demonstrate that our proposed method effectively improves the quality of the generated text, especially in coherence and diversity. The code is available at: \url{https://github.com/qtli/EventPlanforTextGen}.

ZeroGen: Efficient Zero-shot Learning via Dataset Generation

There is a growing interest in dataset generation recently due to the superior generative capacity of large pre-trained language models (PLMs). In this paper, we study a flexible and efficient zero-short learning method, ZeroGen. Given a zero-shot task, we first generate a dataset from scratch using PLMs in an unsupervised manner. Then, we train a tiny task model (e.g., LSTM) under the supervision of the synthesized dataset. This approach allows highly efficient inference as the final task model only has orders of magnitude fewer parameters comparing to PLMs (e.g., GPT2-XL). Apart from being annotation-free and efficient, we argue that ZeroGen can also provide useful insights from the perspective of data-free model-agnostic knowledge distillation, and unreferenced text generation evaluation. Experiments and analysis on different NLP tasks, namely, text classification, question answering, and natural language inference), show the effectiveness of ZeroGen.

Abstractive Opinion Tagging

In e-commerce, opinion tags refer to a ranked list of tags provided by the e-commerce platform that reflect characteristics of reviews of an item. To assist consumers to quickly grasp a large number of reviews about an item, opinion tags are increasingly being applied by e-commerce platforms. Current mechanisms for generating opinion tags rely on either manual labelling or heuristic methods, which is time-consuming and ineffective. In this paper, we propose the abstractive opinion tagging task, where systems have to automatically generate a ranked list of opinion tags that are based on, but need not occur in, a given set of user-generated reviews. The abstractive opinion tagging task comes with three main challenges: (1) the noisy nature of reviews; (2) the formal nature of opinion tags vs. the colloquial language usage in reviews; and (3) the need to distinguish between different items with very similar aspects. To address these challenges, we propose an abstractive opinion tagging framework, named AOT-Net, to generate a ranked list of opinion tags given a large number of reviews. First, a sentence-level salience estimation component estimates each review's salience score. Next, a review clustering and ranking component ranks reviews in two steps: first, reviews are grouped into clusters and ranked by cluster size; then, reviews within each cluster are ranked by their distance to the cluster center. Finally, given the ranked reviews, a rank-aware opinion tagging component incorporates an alignment feature and alignment loss to generate a ranked list of opinion tags. To facilitate the study of this task, we create and release a large-scale dataset, called eComTag, crawled from real-world e-commerce websites. Extensive experiments conducted on the eComTag dataset verify the effectiveness of the proposed AOT-Net in terms of various evaluation metrics.