Authors: Kavita Sharma, Ritu Patel, Sunita Iyer
In this paper, we propose a novel method to enhance sentiment analysis by addressing the challenge of context-specific word meanings. It combines the advantages of a bidirectional long short-term memory network (Bi-LSTM) with a knowledge graph's synonym data. This synergy leverages a dynamic attention mechanism to develop a knowledge-driven state vector. For classifying sentiments linked to specific aspects, the approach constructs a memory bank integrating positional data. This data is then analyzed using a multi-layer gated recurrent unit (GRU) to pinpoint sentiment characteristics related to specific aspect terms. Tests on three widely available datasets demonstrate this method's superior performance in sentiment classification.
Authors: Noam Benkler, Drisana Mosaphir, Scott Friedman, Andrew Smart, Sonja Schmer-Galunder
The fields of AI current lacks methods to quantitatively assess and potentially alter the moral values inherent in the output of large language models (LLMs). However, decades of social science research has developed and refined widely-accepted moral value surveys, such as the World Values Survey (WVS), eliciting value judgments from direct questions in various geographies. We have turned those questions into value statements and use NLP to compute to how well popular LLMs are aligned with moral values for various demographics and cultures. While the WVS is accepted as an explicit assessment of values, we lack methods for assessing implicit moral and cultural values in media, e.g., encountered in social media, political rhetoric, narratives, and generated by AI systems such as LLMs that are increasingly present in our daily lives. As we consume online content and utilize LLM outputs, we might ask, which moral values are being implicitly promoted or undercut, or -- in the case of LLMs -- if they are intending to represent a cultural identity, are they doing so consistently? In this paper we utilize a Recognizing Value Resonance (RVR) NLP model to identify WVS values that resonate and conflict with a given passage of output text. We apply RVR to the text generated by LLMs to characterize implicit moral values, allowing us to quantify the moral/cultural distance between LLMs and various demographics that have been surveyed using the WVS. In line with other work we find that LLMs exhibit several Western-centric value biases; they overestimate how conservative people in non-Western countries are, they are less accurate in representing gender for non-Western countries, and portray older populations as having more traditional values. Our results highlight value misalignment and age groups, and a need for social science informed technological solutions addressing value plurality in LLMs.
Authors: Yuyang Deng, Ni Zhao, Xin Huang
Since its launch, ChatGPT has achieved remarkable success as a versatile conversational AI platform, drawing millions of users worldwide and garnering widespread recognition across academic, industrial, and general communities. This paper aims to point a portrait of early GPT users and understand how they evolved. Specific questions include their topics of interest and their potential careers; and how this changes over time. We conduct a detailed analysis of real-world ChatGPT datasets with multi-turn conversations between users and ChatGPT. Through a multi-pronged approach, we quantify conversation dynamics by examining the number of turns, then gauge sentiment to understand user sentiment variations, and finally employ Latent Dirichlet Allocation (LDA) to discern overarching topics within the conversation. By understanding shifts in user demographics and interests, we aim to shed light on the changing nature of human-AI interaction and anticipate future trends in user engagement with language models.
Authors: Alexandre Variengien, Eric Winsor
When solving challenging problems, language models (LMs) are able to identify relevant information from long and complicated contexts. To study how LMs solve retrieval tasks in diverse situations, we introduce ORION, a collection of structured retrieval tasks spanning six domains, from text understanding to coding. Each task in ORION can be represented abstractly by a request (e.g. a question) that retrieves an attribute (e.g. the character name) from a context (e.g. a story). We apply causal analysis on 18 open-source language models with sizes ranging from 125 million to 70 billion parameters. We find that LMs internally decompose retrieval tasks in a modular way: middle layers at the last token position process the request, while late layers retrieve the correct entity from the context. After causally enforcing this decomposition, models are still able to solve the original task, preserving 70% of the original correct token probability in 98 of the 106 studied model-task pairs. We connect our macroscopic decomposition with a microscopic description by performing a fine-grained case study of a question-answering task on Pythia-2.8b. Building on our high-level understanding, we demonstrate a proof of concept application for scalable internal oversight of LMs to mitigate prompt-injection while requiring human supervision on only a single input. Our solution improves accuracy drastically (from 15.5% to 97.5% on Pythia-12b). This work presents evidence of a universal emergent modular processing of tasks across varied domains and models and is a pioneering effort in applying interpretability for scalable internal oversight of LMs.
Authors: Isidor Konrad Maier, Matthias Wolff
In this paper we present a novel numeral decomposer that is designed to revert Hurford's Packing Strategy. The Packing Strategy is a model on how numeral words are formed out of smaller numeral words by recursion. The decomposer does not simply check decimal digits but it also works for numerals formed on base 20 or any other base or even combinations of different bases. All assumptions that we use are justified with Hurford's Packing Strategy. The decomposer reads through the numeral. When it finds a sub-numeral, it checks arithmetic conditions to decide whether or not to unpack the sub-numeral. The goal is to unpack those numerals that can sensibly be substituted by similar numerals. E.g., in 'twenty-seven thousand and two hundred and six' it should unpack 'twenty-seven' and 'two hundred and six', as those could each be sensibly replaced by any numeral from 1 to 999. Our most used condition is: If S is a substitutable sub-numeral of a numeral N, then 2*value(S) < value(N). We have tested the decomposer on numeral systems in 254 different natural languages. We also developed a reinforcement learning algorithm based on the decomposer. Both algorithms' code and the results are open source on GitHub.
Authors: Douglas Schonholtz
As coding challenges become more complex, recent advancements in Large Language Models (LLMs) have led to notable successes, such as achieving a 94.6\% solve rate on the HumanEval benchmark. Concurrently, there is an increasing commercial push for repository-level inline code completion tools, such as GitHub Copilot and Tab Nine, aimed at enhancing developer productivity. This paper delves into the transition from individual coding problems to repository-scale solutions, presenting a thorough review of the current literature on effective LLM prompting for code generation at the repository level. We examine approaches that will work with black-box LLMs such that they will be useful and applicable to commercial use cases, and their applicability in interpreting code at a repository scale. We juxtapose the Repository-Level Prompt Generation technique with RepoCoder, an iterative retrieval and generation method, to highlight the trade-offs inherent in each approach and to establish best practices for their application in cutting-edge coding benchmarks. The interplay between iterative refinement of prompts and the development of advanced retrieval systems forms the core of our discussion, offering a pathway to significantly improve LLM performance in code generation tasks. Insights from this study not only guide the application of these methods but also chart a course for future research to integrate such techniques into broader software engineering contexts.
Authors: Yingzhe Peng, Xu Yang, Haoxuan Ma, Shuo Xu, Chi Zhang, Yucheng Han, Hanwang Zhang
This paper studies how to configure powerful In-Context Demonstration (ICD) sequences for a Large Vision-Language Model (LVLM) to solve Vision-Language tasks through In-Context Learning (ICL). After observing that configuring an ICD sequence is a mirror process of composing a sentence, i.e., just as a sentence can be composed word by word via a Language Model, an ICD sequence can also be configured one by one. Consequently, we introduce an ICD Language Model (ICD-LM) specifically designed to generate effective ICD sequences. This involves creating a dataset of hand-crafted ICD sequences for various query samples and using it to train the ICD-LM. Our approach, diverging from traditional methods in NLP that select and order ICDs separately, enables to simultaneously learn how to select and order ICDs, enhancing the effect of the sequences. Moreover, during data construction, we use the LVLM intended for ICL implementation to validate the strength of each ICD sequence, resulting in a model-specific dataset and the ICD-LM trained by this dataset is also model-specific. We validate our methodology through experiments in Visual Question Answering and Image Captioning, confirming the viability of using a Language Model for ICD configuration. Our comprehensive ablation studies further explore the impact of various dataset construction and ICD-LM development settings on the outcomes. The code is given in https://github.com/ForJadeForest/ICD-LM.
Authors: Sweta Agrawal, Marine Carpuat
Automatic text simplification (TS) aims to automate the process of rewriting text to make it easier for people to read. A pre-requisite for TS to be useful is that it should convey information that is consistent with the meaning of the original text. However, current TS evaluation protocols assess system outputs for simplicity and meaning preservation without regard for the document context in which output sentences occur and for how people understand them. In this work, we introduce a human evaluation framework to assess whether simplified texts preserve meaning using reading comprehension questions. With this framework, we conduct a thorough human evaluation of texts by humans and by nine automatic systems. Supervised systems that leverage pre-training knowledge achieve the highest scores on the reading comprehension (RC) tasks amongst the automatic controllable TS systems. However, even the best-performing supervised system struggles with at least 14% of the questions, marking them as "unanswerable'' based on simplified content. We further investigate how existing TS evaluation metrics and automatic question-answering systems approximate the human judgments we obtained.
Authors: Kung-Hsiang Huang, Mingyang Zhou, Hou Pong Chan, Yi R. Fung, Zhenhailong Wang, Lingyu Zhang, Shih-Fu Chang, Heng Ji
Recent advancements in large vision-language models (LVLMs) have led to significant progress in generating natural language descriptions for visual content and thus enhancing various applications. One issue with these powerful models is that they sometimes produce texts that are factually inconsistent with the visual input. While there has been some effort to mitigate such inconsistencies in natural image captioning, the factuality of generated captions for structured document images, such as charts, has not received as much scrutiny, posing a potential threat to information reliability in critical applications. This work delves into the factuality aspect by introducing a comprehensive typology of factual errors in generated chart captions. A large-scale human annotation effort provides insight into the error patterns and frequencies in captions crafted by various chart captioning models, ultimately forming the foundation of a novel dataset, CHOCOLATE. Our analysis reveals that even state-of-the-art models, including GPT-4V, frequently produce captions laced with factual inaccuracies. In response to this challenge, we establish the new task of Chart Caption Factual Error Correction and introduce CHARTVE, a model for visual entailment that outperforms proprietary and open-source LVLMs in evaluating factual consistency. Furthermore, we propose C2TFEC, an interpretable two-stage framework that excels at correcting factual errors. This work inaugurates a new domain in factual error correction for chart captions, presenting a novel evaluation mechanism, and demonstrating an effective approach to ensuring the factuality of generated chart captions.
Authors: Niels Dickson
An ability that underlies human syntactic knowledge is determining which words can appear in the similar structures (i.e. grouping words by their syntactic categories). These groupings enable humans to combine structures in order to communicate complex meanings. A foundational question is how do children acquire this ability underlying syntactic knowledge. In exploring this process, we will review various engineering approaches whose goal is similar to that of a child's -- without prior syntactic knowledge, correctly identify the parts of speech (POS) of the words in a sample of text. In reviewing these unsupervised tagging efforts, we will discuss common themes that support the advances in the models and their relevance for language acquisition. For example, we discuss how each model judges success (evaluation metrics), the "additional information" that constrains the POS learning (such as orthographic information), and the context used to determine POS (only previous word, words before and after the target, etc). The identified themes pave the way for future investigations into the cognitive processes that underpin the acquisition of syntactic categories and provide a useful layout of current state of the art unsupervised POS tagging models.
Authors: Jan Drchal, Herbert Ullrich, Tomáš Mlynář, Václav Moravec
This article presents a pipeline for automated fact-checking leveraging publicly available Language Models and data. The objective is to assess the accuracy of textual claims using evidence from a ground-truth evidence corpus. The pipeline consists of two main modules -- the evidence retrieval and the claim veracity evaluation. Our primary focus is on the ease of deployment in various languages that remain unexplored in the field of automated fact-checking. Unlike most similar pipelines, which work with evidence sentences, our pipeline processes data on a paragraph level, simplifying the overall architecture and data requirements. Given the high cost of annotating language-specific fact-checking training data, our solution builds on the Question Answering for Claim Generation (QACG) method, which we adapt and use to generate the data for all models of the pipeline. Our strategy enables the introduction of new languages through machine translation of only two fixed datasets of moderate size. Subsequently, any number of training samples can be generated based on an evidence corpus in the target language. We provide open access to all data and fine-tuned models for Czech, English, Polish, and Slovak pipelines, as well as to our codebase that may be used to reproduce the results.We comprehensively evaluate the pipelines for all four languages, including human annotations and per-sample difficulty assessment using Pointwise V-information. The presented experiments are based on full Wikipedia snapshots to promote reproducibility. To facilitate implementation and user interaction, we develop the FactSearch application featuring the proposed pipeline and the preliminary feedback on its performance.
Authors: Jiachen Zhao
Knowledge distillation (KD) has been widely employed to transfer knowledge from a large language model (LLM) to a specialized model in low-data regimes through pseudo label learning. However, pseudo labels generated by teacher models are usually noisy and may influence KD performance. This study delves into KD with noisy teachers and uncovers that the student model can already generate more accurate predictions than the teacher labels used to train it during KD, indicating its inherent ability to denoise noisy teacher labels. Motivated by this finding, we propose Peer-Advised KD to improve vanilla KD from noisy teachers. Experiments show that Peer-Advised KD can outperform LLM by approximately 5% with 50 human-labeled data, and even competitive to standard supervised finetuning with 750 human-labeled data.
Authors: Tuan Dung Le, Thanh Duong, Thanh Thieu
Objective: Function is increasingly recognized as an important indicator of whole-person health. This study evaluates the ability of publicly available large language models (LLMs) to accurately identify the presence of functioning information from clinical notes. We explore various strategies to improve the performance on this task. Materials and Methods: We collect a balanced binary classification dataset of 1000 sentences from the Mobility NER dataset, which was curated from n2c2 clinical notes. For evaluation, we construct zero-shot and few-shot prompts to query the LLMs whether a given sentence contains mobility functioning information. Two sampling techniques, random sampling and k-nearest neighbor (kNN)-based sampling, are used to select the few-shot examples. Furthermore, we apply a parameter-efficient prompt-based fine-tuning method to the LLMs and evaluate their performance under various training settings. Results: Flan-T5-xxl outperforms all other models in both zero-shot and few-shot settings, achieving a F1 score of 0.865 with a single demonstrative example selected by kNN sampling. In prompt-based fine-tuning experiments, this foundation model also demonstrates superior performance across all low-resource settings, particularly achieving an impressive F1 score of 0.922 using the full training dataset. The smaller model, Flan-T5-xl, requires fine-tuning with only 2.3M additional parameters to achieve comparable performance to the fully fine-tuned Gatortron-base model, both surpassing 0.9 F1 score. Conclusion: Open-source instruction-tuned LLMs demonstrate impressive in-context learning capability in the mobility functioning classification task. The performance of these models can be further improved by continuing fine-tuning on a task-specific dataset.
Authors: Liqiang Jing, Xuemeng Song, Xinxing Zu, Na Zheng, Zhongzhou Zhao, Liqiang Nie
Existing sign language translation methods follow a two-stage pipeline: first converting the sign language video to a gloss sequence (i.e. Sign2Gloss) and then translating the generated gloss sequence into a spoken language sentence (i.e. Gloss2Text). While previous studies have focused on boosting the performance of the Sign2Gloss stage, we emphasize the optimization of the Gloss2Text stage. However, this task is non-trivial due to two distinct features of Gloss2Text: (1) isolated gloss input and (2) low-capacity gloss vocabulary. To address these issues, we propose a vision and context knowledge enhanced Gloss2Text model, named VK-G2T, which leverages the visual content of the sign language video to learn the properties of the target sentence and exploit the context knowledge to facilitate the adaptive translation of gloss words. Extensive experiments conducted on a Chinese benchmark validate the superiority of our model.
Authors: Jiangyan Yi, Ye Bai, Jianhua Tao, Haoxin Ma, Zhengkun Tian, Chenglong Wang, Tao Wang, Ruibo Fu
Diverse promising datasets have been designed to hold back the development of fake audio detection, such as ASVspoof databases. However, previous datasets ignore an attacking situation, in which the hacker hides some small fake clips in real speech audio. This poses a serious threat since that it is difficult to distinguish the small fake clip from the whole speech utterance. Therefore, this paper develops such a dataset for half-truth audio detection (HAD). Partially fake audio in the HAD dataset involves only changing a few words in an utterance.The audio of the words is generated with the very latest state-of-the-art speech synthesis technology. We can not only detect fake uttrances but also localize manipulated regions in a speech using this dataset. Some benchmark results are presented on this dataset. The results show that partially fake audio presents much more challenging than fully fake audio for fake audio detection. The HAD dataset is publicly available: https://zenodo.org/records/10377492.
Authors: Kevin Denamganaï, Sondess Missaoui, James Alfred Walker
Human beings use compositionality to generalise from past experiences to novel experiences. We assume a separation of our experiences into fundamental atomic components that can be recombined in novel ways to support our ability to engage with novel experiences. We frame this as the ability to learn to generalise compositionally, and we will refer to behaviours making use of this ability as compositional learning behaviours (CLBs). A central problem to learning CLBs is the resolution of a binding problem (BP). While it is another feat of intelligence that human beings perform with ease, it is not the case for state-of-the-art artificial agents. Thus, in order to build artificial agents able to collaborate with human beings, we propose to develop a novel benchmark to investigate agents' abilities to exhibit CLBs by solving a domain-agnostic version of the BP. We take inspiration from the language emergence and grounding framework of referential games and propose a meta-learning extension of referential games, entitled Meta-Referential Games, and use this framework to build our benchmark, the Symbolic Behaviour Benchmark (S2B). We provide baseline results and error analysis showing that our benchmark is a compelling challenge that we hope will spur the research community towards developing more capable artificial agents.
Authors: Rong Tian, Zijing Zhao, Weijie Liu, Haoyan Liu, Weiquan Mao, Zhe Zhao, Kan Zhou
The latest industrial inference engines, such as FasterTransformer and TurboTransformers, have verified that half-precision floating point (FP16) and 8-bit integer (INT8) quantization can greatly improve model inference speed. However, the existing INT8 quantization methods are too complicated, and improper usage will lead to model performance damage greatly. In this paper, we develop a toolkit for users to easily quantize their models for inference, in which Self-Adaptive Mixed-Precision (SAMP) is proposed to automatically control quantization rate by a mixed-precision architecture to balance model accuracy and efficiency. Experimental results show that our SAMP toolkit has a higher speedup than PyTorch and FasterTransformer while ensuring the required accuracy. In addition, SAMP is based on a modular design, decoupling the tokenizer, embedding, encoder and target layers, which allows users to handle various downstream tasks and can be seamlessly integrated into PyTorch.
Authors: Daniel Gilo, Shaul Markovitch
One of the prominent methods for explaining the decision of a machine-learning classifier is by a counterfactual example. Most current algorithms for generating such examples in the textual domain are based on generative language models. Generative models, however, are trained to minimize a specific loss function in order to fulfill certain requirements for the generated texts. Any change in the requirements may necessitate costly retraining, thus potentially limiting their applicability. In this paper, we present a general search-based framework for generating counterfactual explanations in the textual domain. Our framework is model-agnostic, domain-agnostic, anytime, and does not require retraining in order to adapt to changes in the user requirements. We model the task as a search problem in a space where the initial state is the classified text, and the goal state is a text in a given target class. Our framework includes domain-independent modification operators, but can also exploit domain-specific knowledge through specialized operators. The search algorithm attempts to find a text from the target class with minimal user-specified distance from the original classified object.
Authors: Siyuan Cheng, Bozhong Tian, Xi Chen, Ningyu Zhang, Qingbing Liu, Huajun Chen
Recently decades have witnessed the empirical success of framing Knowledge Graph (KG) embeddings via language models. However, language model-based KG embeddings are usually deployed as static artifacts, making them difficult to modify post-deployment without re-training after deployment. To address this issue, we propose a new task of editing language model-based KG embeddings in this paper. This task is designed to facilitate rapid, data-efficient updates to KG embeddings without compromising the performance of other aspects. We build four new datasets: E-FB15k237, A-FB15k237, E-WN18RR, and A-WN18RR, and evaluate several knowledge editing baselines demonstrating the limited ability of previous models to handle the proposed challenging task. We further propose a simple yet strong baseline dubbed KGEditor, which utilizes additional parametric layers of the hypernetwork to edit/add facts. Our comprehensive experimental results reveal that KGEditor excels in updating specific facts without impacting the overall performance, even when faced with limited training resources. Code and datasets are available in https://github.com/zjunlp/PromptKG/tree/main/deltaKG.
Authors: Mingyang Chen, Wen Zhang, Yuxia Geng, Zezhong Xu, Jeff Z. Pan, Huajun Chen
Knowledge graphs (KGs) have become valuable knowledge resources in various applications, and knowledge graph embedding (KGE) methods have garnered increasing attention in recent years. However, conventional KGE methods still face challenges when it comes to handling unseen entities or relations during model testing. To address this issue, much effort has been devoted to various fields of KGs. In this paper, we use a set of general terminologies to unify these methods and refer to them collectively as Knowledge Extrapolation. We comprehensively summarize these methods, classified by our proposed taxonomy, and describe their interrelationships. Additionally, we introduce benchmarks and provide comparisons of these methods based on aspects that are not captured by the taxonomy. Finally, we suggest potential directions for future research.
Authors: Neil Shah, Saiteja Kosgi, Vishal Tambrahalli, Neha Sahipjohn, Niranjan Pedanekar, Vineet Gandhi
We present ParrotTTS, a modularized text-to-speech synthesis model leveraging disentangled self-supervised speech representations. It can train a multi-speaker variant effectively using transcripts from a single speaker. ParrotTTS adapts to a new language in low resource setup and generalizes to languages not seen while training the self-supervised backbone. Moreover, without training on bilingual or parallel examples, ParrotTTS can transfer voices across languages while preserving the speaker specific characteristics, e.g., synthesizing fluent Hindi speech using a French speaker's voice and accent. We present extensive results in monolingual and multi-lingual scenarios. ParrotTTS outperforms state-of-the-art multi-lingual TTS models using only a fraction of paired data as latter.
Authors: Daixuan Cheng, Shaohan Huang, Junyu Bi, Yuefeng Zhan, Jianfeng Liu, Yujing Wang, Hao Sun, Furu Wei, Denvy Deng, Qi Zhang
Large Language Models (LLMs) are popular for their impressive abilities, but the need for model-specific fine-tuning or task-specific prompt engineering can hinder their generalization. We propose UPRISE (Universal Prompt Retrieval for Improving zero-Shot Evaluation), which tunes a lightweight and versatile retriever that automatically retrieves prompts for a given zero-shot task input. Specifically, we demonstrate universality in a cross-task and cross-model scenario: the retriever is tuned on a diverse set of tasks, but tested on unseen task types; we use a small frozen LLM, GPT-Neo-2.7B, for tuning the retriever, but test the retriever on different LLMs of much larger scales, such as BLOOM-7.1B, OPT-66B and GPT3-175B. Additionally, we show that UPRISE mitigates the hallucination problem in our experiments with ChatGPT, suggesting its potential to improve even the strongest LLMs. Our model and code are available at https://github.com/microsoft/LMOps.
Authors: Lei Wang, Yi Hu, Jiabang He, Xing Xu, Ning Liu, Hui Liu, Heng Tao Shen
Large Language Models (LLMs) have recently demonstrated exceptional performance in various Natural Language Processing (NLP) tasks. They have also shown the ability to perform chain-of-thought (CoT) reasoning to solve complex problems. Recent studies have explored CoT reasoning in complex multimodal scenarios, such as the science question answering task, by fine-tuning multimodal models with high-quality human-annotated CoT rationales. However, collecting high-quality COT rationales is usually time-consuming and costly. Besides, the annotated rationales are hardly accurate due to the external essential information missed. To address these issues, we propose a novel method termed T-SciQ that aims at teaching science question answering with LLM signals. The T-SciQ approach generates high-quality CoT rationales as teaching signals and is advanced to train much smaller models to perform CoT reasoning in complex modalities. Additionally, we introduce a novel data mixing strategy to produce more effective teaching data samples for simple and complex science question answer problems. Extensive experimental results show that our T-SciQ method achieves a new state-of-the-art performance on the ScienceQA benchmark, with an accuracy of 96.18%. Moreover, our approach outperforms the most powerful fine-tuned baseline by 4.5%. The code is publicly available at https://github.com/T-SciQ/T-SciQ.
Authors: Prateek Chhikara, Jiarui Zhang, Filip Ilievski, Jonathan Francis, Kaixin Ma
Communication via natural language is a key aspect of machine intelligence, and it requires computational models to learn and reason about world concepts, with varying levels of supervision. Significant progress has been made on fully-supervised non-interactive tasks, such as question-answering and procedural text understanding. Yet, various sequential interactive tasks, as in text-based games, have revealed limitations of existing approaches in terms of coherence, contextual awareness, and their ability to learn effectively from the environment. In this paper, we propose a knowledge-injection framework for improved functional grounding of agents in text-based games. Specifically, we consider two forms of domain knowledge that we inject into learning-based agents: memory of previous correct actions and affordances of relevant objects in the environment. Our framework supports two representative model classes: reinforcement learning agents and language model agents. Furthermore, we devise multiple injection strategies for the above domain knowledge types and agent architectures, including injection via knowledge graphs and augmentation of the existing input encoding strategies. We experiment with four models on the 10 tasks in the ScienceWorld text-based game environment, to illustrate the impact of knowledge injection on various model configurations and challenging task settings. Our findings provide crucial insights into the interplay between task properties, model architectures, and domain knowledge for interactive contexts.
Authors: Zhengqing Yuan, Xinyi Wang, Kun Wang, Lichao Sun, Yanyang Ye
In recent years, advancements in large language models have been remarkable, with models such as ChatGPT demonstrating exceptional proficiency in diverse linguistic tasks. The pre-training of large models with billions of parameters, poses a formidable challenge, primarily due to the scarcity of datasets of a commensurate scale for effective training. Nevertheless, innovative strategies have emerged, including methods to fine-tune these pre-trained models using fewer parameters set, as evidenced by models like MiniGPT-4 and LLaVA. Despite their potential in various domains, these models remain limited in their understanding of artistic imagery. They have yet to fully grasp the intricate nuances of art images or to provide an objective articulation of the emotions they evoke, in a manner akin to human perception. This work introduces ArtGPT-4, a pioneering large vision-language model tailored to address the deficiencies of contemporary models in artistic comprehension. ArtGPT-4 underwent training on image-text pairs utilizing a Tesla A100 device in a mere 2 hours, with a dataset comprising approximately 0.52M entries. Impressively, the model can render images with an artistic-understanding and convey the emotions they inspire, mirroring human interpretation. Additionally, this work presents a unique dataset designed to evaluate the efficacy of vision-language models. In subsequent evaluations, ArtGPT-4 not only achieved state-of-the-art performance on the ArtEmis and ArtEmis-v2.0 datasets but also exceeded the established benchmarks introduced in This study, lagging behind professional artists' descriptions by a negligible 0.15 points on a 6-point scale. The code and the pre-trained model are accessible in https://huggingface.co/Tyrannosaurus/ArtGPT-4.
Authors: Viktor Moskvoretskii, Anton Frolov, Denis Kuznetsov
Currently, dialogue systems have achieved high performance in processing text-based communication. However, they have not yet effectively incorporated visual information, which poses a significant challenge. Furthermore, existing models that incorporate images in dialogue generation focus on discussing the image itself. Our proposed approach presents a novel perspective on multi-modal dialogue systems, which interprets the image in the context of the dialogue. By doing so, we aim to expand the capabilities of current dialogue systems and transition them from single modality (text) to multi-modality. However, there is a lack of validated English datasets that contain both images and dialogue contexts for this task. Thus, we propose a two-stage approach to automatically construct a multi-modal dialogue dataset. In the first stage, we utilize text-to-image similarity and sentence similarity to identify which utterances could be replaced with an image. In the second stage, we replace those utterances by selecting a subset of relevant images and filtering them with a visual question answering model. We used this approach, along with additional labeling, to create the IMage Augmented multi-modal Dialogue dataset (IMAD), which can serve as a validated dataset for this task. Furthermore, we propose a baseline model trained on this dataset, which outperforms model trained on the same data without images and BlenderBot.
Authors: Yaobo Liang, Quanzhi Zhu, Junhe Zhao, Nan Duan
There are two primary approaches to addressing cross-lingual transfer: multilingual pre-training, which implicitly aligns the hidden representations of various languages, and translate-test, which explicitly translates different languages into an intermediate language, such as English. Translate-test offers better interpretability compared to multilingual pre-training. However, it has lower performance than multilingual pre-training(Conneau and Lample, 2019; Conneau et al, 2020) and struggles with word-level tasks due to translation altering word order. As a result, we propose a new Machine-created Universal Language (MUL) as an alternative intermediate language. MUL comprises a set of discrete symbols forming a universal vocabulary and a natural language to MUL translator for converting multiple natural languages to MUL. MUL unifies shared concepts from various languages into a single universal word, enhancing cross-language transfer. Additionally, MUL retains language-specific words and word order, allowing the model to be easily applied to word-level tasks. Our experiments demonstrate that translating into MUL yields improved performance compared to multilingual pre-training, and our analysis indicates that MUL possesses strong interpretability. The code is at: https://github.com/microsoft/Unicoder/tree/master/MCUL.
Authors: Uri Shaham, Maor Ivgi, Avia Efrat, Jonathan Berant, Omer Levy
We introduce ZeroSCROLLS, a zero-shot benchmark for natural language understanding over long texts, which contains only test and small validation sets, without training data. We adapt six tasks from the SCROLLS benchmark, and add four new datasets, including two novel information fusing tasks, such as aggregating the percentage of positive reviews. Using ZeroSCROLLS, we conduct a comprehensive evaluation of both open-source and closed large language models, finding that Claude outperforms ChatGPT, and that GPT-4 achieves the highest average score. However, there is still room for improvement on multiple open challenges in ZeroSCROLLS, such as aggregation tasks, where models struggle to pass the naive baseline. As the state of the art is a moving target, we invite researchers to evaluate their ideas on the live ZeroSCROLLS leaderboard.
Authors: Ziyi Ni, Minglun Han, Feilong Chen, Linghui Meng, Jing Shi, Pin Lv, Bo Xu
Enhancing automatic speech recognition (ASR) performance by leveraging additional multimodal information has shown promising results in previous studies. However, most of these works have primarily focused on utilizing visual cues derived from human lip motions. In fact, context-dependent visual and linguistic cues can also benefit in many scenarios. In this paper, we first propose ViLaS (Vision and Language into Automatic Speech Recognition), a novel multimodal ASR model based on the continuous integrate-and-fire (CIF) mechanism, which can integrate visual and textual context simultaneously or separately, to facilitate speech recognition. Next, we introduce an effective training strategy that improves performance in modal-incomplete test scenarios. Then, to explore the effects of integrating vision and language, we create VSDial, a multimodal ASR dataset with multimodal context cues in both Chinese and English versions. Finally, empirical results are reported on the public Flickr8K and self-constructed VSDial datasets. We explore various cross-modal fusion schemes, analyze fine-grained crossmodal alignment on VSDial, and provide insights into the effects of integrating multimodal information on speech recognition.
Authors: Yongqi Li, Nan Yang, Liang Wang, Furu Wei, Wenjie Li
Generative retrieval stands out as a promising new paradigm in text retrieval that aims to generate identifier strings of relevant passages as the retrieval target. This generative paradigm taps into powerful generative language models, distinct from traditional sparse or dense retrieval methods. However, only learning to generate is insufficient for generative retrieval. Generative retrieval learns to generate identifiers of relevant passages as an intermediate goal and then converts predicted identifiers into the final passage rank list. The disconnect between the learning objective of autoregressive models and the desired passage ranking target leads to a learning gap. To bridge this gap, we propose a learning-to-rank framework for generative retrieval, dubbed LTRGR. LTRGR enables generative retrieval to learn to rank passages directly, optimizing the autoregressive model toward the final passage ranking target via a rank loss. This framework only requires an additional learning-to-rank training phase to enhance current generative retrieval systems and does not add any burden to the inference stage. We conducted experiments on three public benchmarks, and the results demonstrate that LTRGR achieves state-of-the-art performance among generative retrieval methods. The code and checkpoints are released at https://github.com/liyongqi67/LTRGR.
Authors: Delong Chen, Jianfeng Liu, Wenliang Dai, Baoyuan Wang
Recent research has demonstrated that the multi-task fine-tuning of multi-modal Large Language Models (LLMs) using an assortment of annotated downstream vision-language datasets significantly enhances their performance. Yet, during this process, a side effect, which we termed as the "multi-modal alignment tax", surfaces. This side effect negatively impacts the model's ability to format responses appropriately -- for instance, its "politeness" -- due to the overly succinct and unformatted nature of raw annotations, resulting in reduced human preference. In this paper, we introduce Polite Flamingo, a multi-modal response rewriter that transforms raw annotations into a more appealing, "polite" format. Polite Flamingo is trained to reconstruct high-quality responses from their automatically distorted counterparts and is subsequently applied to a vast array of vision-language datasets for response rewriting. After rigorous filtering, we generate the PF-1M dataset and further validate its value by fine-tuning a multi-modal LLM with it. Combined with novel methodologies including U-shaped multi-stage tuning and multi-turn augmentation, the resulting model, Clever Flamingo, demonstrates its advantages in both multi-modal understanding and response politeness according to automated and human evaluations.
Authors: Chunyu Qiang, Hao Li, Hao Ni, He Qu, Ruibo Fu, Tao Wang, Longbiao Wang, Jianwu Dang
Recently, there has been a growing interest in text-to-speech (TTS) methods that can be trained with minimal supervision by combining two types of discrete speech representations and using two sequence-to-sequence tasks to decouple TTS. However, existing methods suffer from three problems: the high dimensionality and waveform distortion of discrete speech representations, the prosodic averaging problem caused by the duration prediction model in non-autoregressive frameworks, and the information redundancy and dimension explosion problems of existing semantic encoding methods. To address these problems, three progressive methods are proposed. First, we propose Diff-LM-Speech, an autoregressive structure consisting of a language model and diffusion models, which models the semantic embedding into the mel-spectrogram based on a diffusion model to achieve higher audio quality. We also introduce a prompt encoder structure based on a variational autoencoder and a prosody bottleneck to improve prompt representation ability. Second, we propose Tetra-Diff-Speech, a non-autoregressive structure consisting of four diffusion model-based modules that design a duration diffusion model to achieve diverse prosodic expressions. Finally, we propose Tri-Diff-Speech, a non-autoregressive structure consisting of three diffusion model-based modules that verify the non-necessity of existing semantic encoding models and achieve the best results. Experimental results show that our proposed methods outperform baseline methods. We provide a website with audio samples.
Authors: Kevin Denamganaï, Daniel Hernandez, Ozan Vardal, Sondess Missaoui, James Alfred Walker
Natural language instruction following is paramount to enable collaboration between artificial agents and human beings. Natural language-conditioned reinforcement learning (RL) agents have shown how natural languages' properties, such as compositionality, can provide a strong inductive bias to learn complex policies. Previous architectures like HIGhER combine the benefit of language-conditioning with Hindsight Experience Replay (HER) to deal with sparse rewards environments. Yet, like HER, HIGhER relies on an oracle predicate function to provide a feedback signal highlighting which linguistic description is valid for which state. This reliance on an oracle limits its application. Additionally, HIGhER only leverages the linguistic information contained in successful RL trajectories, thus hurting its final performance and data-efficiency. Without early successful trajectories, HIGhER is no better than DQN upon which it is built. In this paper, we propose the Emergent Textual Hindsight Experience Replay (ETHER) agent, which builds on HIGhER and addresses both of its limitations by means of (i) a discriminative visual referential game, commonly studied in the subfield of Emergent Communication (EC), used here as an unsupervised auxiliary task and (ii) a semantic grounding scheme to align the emergent language with the natural language of the instruction-following benchmark. We show that the referential game's agents make an artificial language emerge that is aligned with the natural-like language used to describe goals in the BabyAI benchmark and that it is expressive enough so as to also describe unsuccessful RL trajectories and thus provide feedback to the RL agent to leverage the linguistic, structured information contained in all trajectories. Our work shows that EC is a viable unsupervised auxiliary task for RL and provides missing pieces to make HER more widely applicable.
Authors: Andrea Santilli, Emanuele Rodolà
In recent years Large Language Models (LLMs) have increased the state of the art on several natural language processing tasks. However, their accessibility is often limited to paid API services, posing challenges for researchers in conducting extensive investigations. On the other hand, while some open-source models have been proposed by the community, they are typically English-centric or multilingual without a specific adaptation for the Italian language. In an effort to democratize the available and open resources for the Italian language, in this paper we introduce Camoscio: a language model specifically tuned to follow users' prompts in Italian. Specifically, we finetuned the smallest variant of LLaMA (7b) with LoRA on a corpus of instruction prompts translated to Italian via ChatGPT. Results indicate that the model's zero-shot performance on various downstream tasks in Italian competes favorably with existing models specifically finetuned for those tasks. All the artifacts (code, dataset, model) are released to the community at the following url: https://github.com/teelinsan/camoscio
Authors: Marija Šakota, Maxime Peyrard, Robert West
Generative language models (LMs) have become omnipresent across data science. For a wide variety of tasks, inputs can be phrased as natural language prompts for an LM, from whose output the solution can then be extracted. LM performance has consistently been increasing with model size - but so has the monetary cost of querying the ever larger models. Importantly, however, not all inputs are equally hard: some require larger LMs for obtaining a satisfactory solution, whereas for others smaller LMs suffice. Based on this fact, we design a framework for cost-effective language model choice, called "Fly-swat or cannon" (FORC). Given a set of inputs and a set of candidate LMs, FORC judiciously assigns each input to an LM predicted to do well on the input according to a so-called meta-model, aiming to achieve high overall performance at low cost. The cost-performance tradeoff can be flexibly tuned by the user. Options include, among others, maximizing total expected performance (or the number of processed inputs) while staying within a given cost budget, or minimizing total cost while processing all inputs. We evaluate FORC on 14 datasets covering five natural language tasks, using four candidate LMs of vastly different size and cost. With FORC, we match the performance of the largest available LM while achieving a cost reduction of 63%. Via our publicly available library, researchers as well as practitioners can thus save large amounts of money without sacrificing performance.
Authors: Linfeng Liu, Hongqiu Wu, Hai Zhao
This paper studies Chinese Spelling Correction (CSC), which aims to detect and correct the potential spelling errors in a given sentence. Current state-of-the-art methods regard CSC as a sequence tagging task and fine-tune BERT-based models on sentence pairs. However, we note a critical flaw in the process of tagging one character to another, that the correction is excessively conditioned on the error. This is opposite from human mindset, where individuals rephrase the complete sentence based on its semantics, rather than solely on the error patterns memorized before. Such a counter-intuitive learning process results in the bottleneck of generalizability and transferability of machine spelling correction. To address this, we propose Rephrasing Language Model (ReLM), where the model is trained to rephrase the entire sentence by infilling additional slots, instead of character-to-character tagging. This novel training paradigm achieves the new state-of-the-art results across fine-tuned and zero-shot CSC benchmarks, outperforming previous counterparts by a large margin. Our method also learns transferable language representation when CSC is jointly trained with other tasks.
Authors: Stefano Nolfi
Large Language Models (LLMs) are capable of displaying a wide range of abilities that are not directly connected with the task for which they are trained: predicting the next words of human-written texts. In this article, I review recent research investigating the cognitive abilities developed by LLMs and their relation to human cognition. I discuss the nature of the indirect process that leads to the acquisition of these cognitive abilities, their relation to other indirect processes, and the implications for the acquisition of integrated abilities. Moreover, I propose the factors that enable the development of abilities that are related only very indirectly to the proximal objective of the training task. Finally, I discuss whether the full set of capabilities that LLMs could possibly develop is predictable.
Authors: Wenbo Hu, Yifan Xu, Yi Li, Weiyue Li, Zeyuan Chen, Zhuowen Tu
Vision Language Models (VLMs), which extend Large Language Models (LLM) by incorporating visual understanding capability, have demonstrated significant advancements in addressing open-ended visual question-answering (VQA) tasks. However, these models cannot accurately interpret images infused with text, a common occurrence in real-world scenarios. Standard procedures for extracting information from images often involve learning a fixed set of query embeddings. These embeddings are designed to encapsulate image contexts and are later used as soft prompt inputs in LLMs. Yet, this process is limited to the token count, potentially curtailing the recognition of scenes with text-rich context. To improve upon them, the present study introduces BLIVA: an augmented version of InstructBLIP with Visual Assistant. BLIVA incorporates the query embeddings from InstructBLIP and also directly projects encoded patch embeddings into the LLM, a technique inspired by LLaVA. This approach assists the model to capture intricate details potentially missed during the query decoding process. Empirical evidence demonstrates that our model, BLIVA, significantly enhances performance in processing text-rich VQA benchmarks (up to 17.76% in OCR-VQA benchmark) and in undertaking general (not particularly text-rich) VQA benchmarks (up to 7.9% in Visual Spatial Reasoning benchmark), and achieved 17.72% overall improvement in a comprehensive multimodal LLM benchmark (MME), comparing to our baseline InstructBLIP. BLIVA demonstrates significant capability in decoding real-world images, irrespective of text presence. To demonstrate the broad industry applications enabled by BLIVA, we evaluate the model using a new dataset comprising YouTube thumbnails paired with question-answer sets across 11 diverse categories. Our code and models are freely accessible at https://github.com/mlpc-ucsd/BLIVA.
Authors: Andrew Zhao, Daniel Huang, Quentin Xu, Matthieu Lin, Yong-Jin Liu, Gao Huang
The recent surge in research interest in applying large language models (LLMs) to decision-making tasks has flourished by leveraging the extensive world knowledge embedded in LLMs. While there is a growing demand to tailor LLMs for custom decision-making tasks, finetuning them for specific tasks is resource-intensive and may diminish the model's generalization capabilities. Moreover, state-of-the-art language models like GPT-4 and Claude are primarily accessible through API calls, with their parametric weights remaining proprietary and unavailable to the public. This scenario emphasizes the growing need for new methodologies that allow learning from agent experiences without requiring parametric updates. To address these problems, we introduce the Experiential Learning (ExpeL) agent. Our agent autonomously gathers experiences and extracts knowledge using natural language from a collection of training tasks. At inference, the agent recalls its extracted insights and past experiences to make informed decisions. Our empirical results highlight the robust learning efficacy of the ExpeL agent, indicating a consistent enhancement in its performance as it accumulates experiences. We further explore the emerging capabilities and transfer learning potential of the ExpeL agent through qualitative observations and additional experiments.
Authors: Yu Wang, Nedim Lipka, Ryan A. Rossi, Alexa Siu, Ruiyi Zhang, Tyler Derr
The `pre-train, prompt, predict' paradigm of large language models (LLMs) has achieved remarkable success in open-domain question answering (OD-QA). However, few works explore this paradigm in the scenario of multi-document question answering (MD-QA), a task demanding a thorough understanding of the logical associations among the contents and structures of different documents. To fill this crucial gap, we propose a Knowledge Graph Prompting (KGP) method to formulate the right context in prompting LLMs for MD-QA, which consists of a graph construction module and a graph traversal module. For graph construction, we create a knowledge graph (KG) over multiple documents with nodes symbolizing passages or document structures (e.g., pages/tables), and edges denoting the semantic/lexical similarity between passages or intra-document structural relations. For graph traversal, we design an LLM-based graph traversal agent that navigates across nodes and gathers supporting passages assisting LLMs in MD-QA. The constructed graph serves as the global ruler that regulates the transitional space among passages and reduces retrieval latency. Concurrently, the graph traversal agent acts as a local navigator that gathers pertinent context to progressively approach the question and guarantee retrieval quality. Extensive experiments underscore the efficacy of KGP for MD-QA, signifying the potential of leveraging graphs in enhancing the prompt design for LLMs. Our code: https://github.com/YuWVandy/KG-LLM-MDQA.
Authors: Zhifang Guo, Jianguo Mao, Rui Tao, Long Yan, Kazushige Ouchi, Hong Liu, Xiangdong Wang
Text-based audio generation models have limitations as they cannot encompass all the information in audio, leading to restricted controllability when relying solely on text. To address this issue, we propose a novel model that enhances the controllability of existing pre-trained text-to-audio models by incorporating additional conditions including content (timestamp) and style (pitch contour and energy contour) as supplements to the text. This approach achieves fine-grained control over the temporal order, pitch, and energy of generated audio. To preserve the diversity of generation, we employ a trainable control condition encoder that is enhanced by a large language model and a trainable Fusion-Net to encode and fuse the additional conditions while keeping the weights of the pre-trained text-to-audio model frozen. Due to the lack of suitable datasets and evaluation metrics, we consolidate existing datasets into a new dataset comprising the audio and corresponding conditions and use a series of evaluation metrics to evaluate the controllability performance. Experimental results demonstrate that our model successfully achieves fine-grained control to accomplish controllable audio generation. Audio samples and our dataset are publicly available at https://conditionaudiogen.github.io/conditionaudiogen/
Authors: Yiyang Yao, Peng Liu, Tiancheng Zhao, Qianqian Zhang, Jiajia Liao, Chunxin Fang, Kyusong Lee, Qing Wang
Object detection (OD) in computer vision has made significant progress in recent years, transitioning from closed-set labels to open-vocabulary detection (OVD) based on large-scale vision-language pre-training (VLP). However, current evaluation methods and datasets are limited to testing generalization over object types and referral expressions, which do not provide a systematic, fine-grained, and accurate benchmark of OVD models' abilities. In this paper, we propose a new benchmark named OVDEval, which includes 9 sub-tasks and introduces evaluations on commonsense knowledge, attribute understanding, position understanding, object relation comprehension, and more. The dataset is meticulously created to provide hard negatives that challenge models' true understanding of visual and linguistic input. Additionally, we identify a problem with the popular Average Precision (AP) metric when benchmarking models on these fine-grained label datasets and propose a new metric called Non-Maximum Suppression Average Precision (NMS-AP) to address this issue. Extensive experimental results show that existing top OVD models all fail on the new tasks except for simple object types, demonstrating the value of the proposed dataset in pinpointing the weakness of current OVD models and guiding future research. Furthermore, the proposed NMS-AP metric is verified by experiments to provide a much more truthful evaluation of OVD models, whereas traditional AP metrics yield deceptive results. Data is available at \url{https://github.com/om-ai-lab/OVDEval}
Authors: Zhen Bi, Ningyu Zhang, Yinuo Jiang, Shumin Deng, Guozhou Zheng, Huajun Chen
In the realm of embodied artificial intelligence, the reasoning capabilities of Large Language Models (LLMs) play a pivotal role. Although there are effective methods like program-of-thought prompting for LLMs which uses programming language to tackle complex reasoning tasks, the specific impact of code data on the improvement of reasoning capabilities remains under-explored. To address this gap, we propose complexity-impacted reasoning score (CIRS), which combines structural and logical attributes, to measure the correlation between code and reasoning abilities. Specifically, we use the abstract syntax tree to encode the structural information and calculate logical complexity by considering the difficulty and the cyclomatic complexity. Through an empirical analysis, we find not all code data of complexity can be learned or understood by LLMs. Optimal level of complexity is critical to the improvement of reasoning abilities by program-aided prompting. Then we design an auto-synthesizing and stratifying algorithm, and apply it to instruction generation for mathematical reasoning and code data filtering for code generation tasks. Extensive results demonstrates the effectiveness of our proposed approach. Code will be integrated into the EasyInstruct framework at https://github.com/zjunlp/EasyInstruct.
Authors: Chunyu Qiang, Hao Li, Yixin Tian, Ruibo Fu, Tao Wang, Longbiao Wang, Jianwu Dang
For fine-grained generation and recognition tasks such as minimally-supervised text-to-speech (TTS), voice conversion (VC), and automatic speech recognition (ASR), the intermediate representations extracted from speech should serve as a "bridge" between text and acoustic information, containing information from both modalities. The semantic content is emphasized, while the paralinguistic information such as speaker identity and acoustic details should be de-emphasized. However, existing methods for extracting fine-grained intermediate representations from speech suffer from issues of excessive redundancy and dimension explosion. Contrastive learning is a good method for modeling intermediate representations from two modalities. However, existing contrastive learning methods in the audio field focus on extracting global descriptive information for downstream audio classification tasks, making them unsuitable for TTS, VC, and ASR tasks. To address these issues, we propose a method named "Contrastive Token-Acoustic Pretraining (CTAP)", which uses two encoders to bring phoneme and speech into a joint multimodal space, learning how to connect phoneme and speech at the frame level. The CTAP model is trained on 210k speech and phoneme pairs, achieving minimally-supervised TTS, VC, and ASR. The proposed CTAP method offers a promising solution for fine-grained generation and recognition downstream tasks in speech processing. We provide a website with audio samples.
Authors: Zhengxiang Shi, Aldo Lipani
Prompt tuning (PT), where a small amount of trainable soft (continuous) prompt vectors is affixed to the input of language models (LM), has shown promising results across various tasks and models for parameter-efficient fine-tuning (PEFT). PT stands out from other PEFT approaches because it maintains competitive performance with fewer trainable parameters and does not drastically scale up its parameters as the model size expands. However, PT introduces additional soft prompt tokens, leading to longer input sequences, which significantly impacts training and inference time and memory usage due to the Transformer's quadratic complexity. Particularly concerning for Large Language Models (LLMs) that face heavy daily querying. To address this issue, we propose Decomposed Prompt Tuning (DePT), which decomposes the soft prompt into a shorter soft prompt and a pair of low-rank matrices that are then optimised with two different learning rates. This allows DePT to achieve better performance while saving over 20% memory and time costs compared to vanilla PT and its variants, without changing trainable parameter sizes. Through extensive experiments on 23 natural language processing (NLP) and vision-language (VL) tasks, we demonstrate that DePT outperforms state-of-the-art PEFT approaches, including the full fine-tuning baseline in some scenarios. Additionally, we empirically show that DEPT grows more efficient as the model size increases. Our further study reveals that DePT integrates seamlessly with parameter-efficient transfer learning in the few-shot learning setting and highlights its adaptability to various model architectures and sizes.
Authors: Chen Jiang, Hong Liu, Xuzheng Yu, Qing Wang, Yuan Cheng, Jia Xu, Zhongyi Liu, Qingpei Guo, Wei Chu, Ming Yang, Yuan Qi
In recent years, the explosion of web videos makes text-video retrieval increasingly essential and popular for video filtering, recommendation, and search. Text-video retrieval aims to rank relevant text/video higher than irrelevant ones. The core of this task is to precisely measure the cross-modal similarity between texts and videos. Recently, contrastive learning methods have shown promising results for text-video retrieval, most of which focus on the construction of positive and negative pairs to learn text and video representations. Nevertheless, they do not pay enough attention to hard negative pairs and lack the ability to model different levels of semantic similarity. To address these two issues, this paper improves contrastive learning using two novel techniques. First, to exploit hard examples for robust discriminative power, we propose a novel Dual-Modal Attention-Enhanced Module (DMAE) to mine hard negative pairs from textual and visual clues. By further introducing a Negative-aware InfoNCE (NegNCE) loss, we are able to adaptively identify all these hard negatives and explicitly highlight their impacts in the training loss. Second, our work argues that triplet samples can better model fine-grained semantic similarity compared to pairwise samples. We thereby present a new Triplet Partial Margin Contrastive Learning (TPM-CL) module to construct partial order triplet samples by automatically generating fine-grained hard negatives for matched text-video pairs. The proposed TPM-CL designs an adaptive token masking strategy with cross-modal interaction to model subtle semantic differences. Extensive experiments demonstrate that the proposed approach outperforms existing methods on four widely-used text-video retrieval datasets, including MSR-VTT, MSVD, DiDeMo and ActivityNet.
Authors: Chunyu Qiang, Hao Li, Yixin Tian, Yi Zhao, Ying Zhang, Longbiao Wang, Jianwu Dang
Text-to-speech (TTS) methods have shown promising results in voice cloning, but they require a large number of labeled text-speech pairs. Minimally-supervised speech synthesis decouples TTS by combining two types of discrete speech representations(semantic \& acoustic) and using two sequence-to-sequence tasks to enable training with minimal supervision. However, existing methods suffer from information redundancy and dimension explosion in semantic representation, and high-frequency waveform distortion in discrete acoustic representation. Autoregressive frameworks exhibit typical instability and uncontrollability issues. And non-autoregressive frameworks suffer from prosodic averaging caused by duration prediction models. To address these issues, we propose a minimally-supervised high-fidelity speech synthesis method, where all modules are constructed based on the diffusion models. The non-autoregressive framework enhances controllability, and the duration diffusion model enables diversified prosodic expression. Contrastive Token-Acoustic Pretraining (CTAP) is used as an intermediate semantic representation to solve the problems of information redundancy and dimension explosion in existing semantic coding methods. Mel-spectrogram is used as the acoustic representation. Both semantic and acoustic representations are predicted by continuous variable regression tasks to solve the problem of high-frequency fine-grained waveform distortion. Experimental results show that our proposed method outperforms the baseline method. We provide audio samples on our website.
Authors: Wanpeng Zhang, Zongqing Lu
Large Language Models (LLMs) have demonstrated significant success across various domains. However, their application in complex decision-making tasks frequently necessitates intricate prompt engineering or fine-tuning, leading to challenges in unseen downstream tasks and heavy demands on computational resources. Meanwhile, Reinforcement Learning (RL) has been recognized as effective in decision-making problems but struggles in environments with sparse rewards, such as open-world games. To overcome these challenges, we introduce AdaRefiner, a novel framework designed to enhance the synergy between LLMs and RL feedback. The key component of AdaRefiner is a lightweight Adapter Language Model (LM), which automatically refines task comprehension based on feedback from RL agents. This method mitigates the need for intricate prompt engineering and intensive LLM fine-tuning while maintaining the LLMs' generalization abilities and enhancing their decision-making capabilities in downstream tasks. Empirical evaluations of AdaRefiner on 22 diverse tasks within the open-world game Crafter have demonstrated its superior effectiveness, especially in guiding agents towards higher-level and common-sense skills. Our work makes contributions to the automatic self-refinement of LLMs with RL feedback, offering a more adaptable and efficient solution for complex decision-making problems.
Authors: Kuan-Po Huang, Chih-Kai Yang, Yu-Kuan Fu, Ewan Dunbar, Hung-yi Lee
We introduce a new zero resource code-switched speech benchmark designed to directly assess the code-switching capabilities of self-supervised speech encoders. We showcase a baseline system of language modeling on discrete units to demonstrate how the code-switching abilities of speech encoders can be assessed in a zero-resource manner. Our experiments encompass a variety of well-known speech encoders, including Wav2vec 2.0, HuBERT, XLSR, etc. We examine the impact of pre-training languages and model size on benchmark performance. Notably, though our results demonstrate that speech encoders with multilingual pre-training, exemplified by XLSR, outperform monolingual variants (Wav2vec 2.0, HuBERT) in code-switching scenarios, there is still substantial room for improvement in their code-switching linguistic abilities.
Authors: Anej Svete, Ryan Cotterell
Understanding neural network architectures with formal models of computation promises to spark a better understanding of the network's capabilities and limitations. A long line of work has described recurrent neural networks (RNN) in terms of their connection to the well-understood finite-state automata (FSAs), whose sequential nature provides a useful analogy to how RNNs function. Minsky's [1954] construction first showed how RNNs can simulate FSAs and provided a way of understanding RNNs as FSAs. This paper presents a comprehensive review of this construction along with two additional classical results showcasing the relationship between RNNs and FSAs: The constructions due to Dewdney [1977] and Indyk [1995]. We are not only interested in \emph{whether} an RNN can simulate an FSA, but also in the space requirements to do so: Whereas Minsky [1954] shows that an RNN can simulate an FSA with $N$ states using $\mathcal{O}\left(N\right)$ neurons, Dewdney [1977] improved this to $\mathcal{O}\left(N^\frac{3}{4}\right)$ and Indyk [1995] further to $\mathcal{O}\left(\sqrt{N}\right)$, which he also showed to be optimal. We discuss the constructions, emphasizing their commonalities, and put them into the context of more modern research, focusing on the representational capacity of neural language models.
Authors: Cunxiang Wang, Xiaoze Liu, Yuanhao Yue, Xiangru Tang, Tianhang Zhang, Cheng Jiayang, Yunzhi Yao, Wenyang Gao, Xuming Hu, Zehan Qi, Yidong Wang, Linyi Yang, Jindong Wang, Xing Xie, Zheng Zhang, Yue Zhang
This survey addresses the crucial issue of factuality in Large Language Models (LLMs). As LLMs find applications across diverse domains, the reliability and accuracy of their outputs become vital. We define the Factuality Issue as the probability of LLMs to produce content inconsistent with established facts. We first delve into the implications of these inaccuracies, highlighting the potential consequences and challenges posed by factual errors in LLM outputs. Subsequently, we analyze the mechanisms through which LLMs store and process facts, seeking the primary causes of factual errors. Our discussion then transitions to methodologies for evaluating LLM factuality, emphasizing key metrics, benchmarks, and studies. We further explore strategies for enhancing LLM factuality, including approaches tailored for specific domains. We focus two primary LLM configurations standalone LLMs and Retrieval-Augmented LLMs that utilizes external data, we detail their unique challenges and potential enhancements. Our survey offers a structured guide for researchers aiming to fortify the factual reliability of LLMs.
Authors: Bo Ni, David L. Kaplan, Markus J. Buehler
Through evolution, nature has presented a set of remarkable protein materials, including elastins, silks, keratins and collagens with superior mechanical performances that play crucial roles in mechanobiology. However, going beyond natural designs to discover proteins that meet specified mechanical properties remains challenging. Here we report a generative model that predicts protein designs to meet complex nonlinear mechanical property-design objectives. Our model leverages deep knowledge on protein sequences from a pre-trained protein language model and maps mechanical unfolding responses to create novel proteins. Via full-atom molecular simulations for direct validation, we demonstrate that the designed proteins are novel, and fulfill the targeted mechanical properties, including unfolding energy and mechanical strength, as well as the detailed unfolding force-separation curves. Our model offers rapid pathways to explore the enormous mechanobiological protein sequence space unconstrained by biological synthesis, using mechanical features as target to enable the discovery of protein materials with superior mechanical properties.
Authors: Yazhou Zhang, Mengyao Wang, Youxi Wu, Prayag Tiwari, Qiuchi Li, Benyou Wang, Jing Qin
Large language models (LLMs) and their variants have shown extraordinary efficacy across numerous downstream natural language processing (NLP) tasks, which has presented a new vision for the development of NLP. Despite their remarkable performance in natural language generating (NLG), LLMs lack a distinct focus on the emotion understanding domain. As a result, using LLMs for emotion recognition may lead to suboptimal and inadequate precision. Another limitation of LLMs is that they are typical trained without leveraging multi-modal information. To overcome these limitations, we propose DialogueLLM, a context and emotion knowledge tuned LLM that is obtained by fine-tuning LLaMA models with 13,638 multi-modal (i.e., texts and videos) emotional dialogues. The visual information is considered as the supplementary knowledge to construct high-quality instructions. We offer a comprehensive evaluation of our proposed model on three benchmarking emotion recognition in conversations (ERC) datasets and compare the results against the SOTA baselines and other SOTA LLMs. Additionally, DialogueLLM-7B can be easily trained using LoRA on a 40GB A100 GPU in 5 hours, facilitating reproducibility for other researchers.
Authors: Hongwei Yao, Jian Lou, Zhan Qin
Prompts have significantly improved the performance of pretrained Large Language Models (LLMs) on various downstream tasks recently, making them increasingly indispensable for a diverse range of LLM application scenarios. However, the backdoor vulnerability, a serious security threat that can maliciously alter the victim model's normal predictions, has not been sufficiently explored for prompt-based LLMs. In this paper, we present POISONPROMPT, a novel backdoor attack capable of successfully compromising both hard and soft prompt-based LLMs. We evaluate the effectiveness, fidelity, and robustness of POISONPROMPT through extensive experiments on three popular prompt methods, using six datasets and three widely used LLMs. Our findings highlight the potential security threats posed by backdoor attacks on prompt-based LLMs and emphasize the need for further research in this area.
Authors: Danis Alukaev, Semen Kiselev, Ilya Pershin, Bulat Ibragimov, Vladimir Ivanov, Alexey Kornaev, Ivan Titov
Concept Bottleneck Models (CBMs) assume that training examples (e.g., x-ray images) are annotated with high-level concepts (e.g., types of abnormalities), and perform classification by first predicting the concepts, followed by predicting the label relying on these concepts. The main difficulty in using CBMs comes from having to choose concepts that are predictive of the label and then having to label training examples with these concepts. In our approach, we adopt a more moderate assumption and instead use text descriptions (e.g., radiology reports), accompanying the images in training, to guide the induction of concepts. Our cross-modal approach treats concepts as discrete latent variables and promotes concepts that (1) are predictive of the label, and (2) can be predicted reliably from both the image and text. Through experiments conducted on datasets ranging from synthetic datasets (e.g., synthetic images with generated descriptions) to realistic medical imaging datasets, we demonstrate that cross-modal learning encourages the induction of interpretable concepts while also facilitating disentanglement. Our results also suggest that this guidance leads to increased robustness by suppressing the reliance on shortcut features.
Authors: Haneen Liqreina, Mustafa Jarrar, Mohammed Khalilia, Ahmed Oumar El-Shangiti, Muhammad Abdul-Mageed
Traditional NER systems are typically trained to recognize coarse-grained entities, and less attention is given to classifying entities into a hierarchy of fine-grained lower-level subtypes. This article aims to advance Arabic NER with fine-grained entities. We chose to extend Wojood (an open-source Nested Arabic Named Entity Corpus) with subtypes. In particular, four main entity types in Wojood, geopolitical entity (GPE), location (LOC), organization (ORG), and facility (FAC), are extended with 31 subtypes. To do this, we first revised Wojood's annotations of GPE, LOC, ORG, and FAC to be compatible with the LDC's ACE guidelines, which yielded 5, 614 changes. Second, all mentions of GPE, LOC, ORG, and FAC (~44K) in Wojood are manually annotated with the LDC's ACE sub-types. We refer to this extended version of Wojood as WojoodF ine. To evaluate our annotations, we measured the inter-annotator agreement (IAA) using both Cohen's Kappa and F1 score, resulting in 0.9861 and 0.9889, respectively. To compute the baselines of WojoodF ine, we fine-tune three pre-trained Arabic BERT encoders in three settings: flat NER, nested NER and nested NER with subtypes and achieved F1 score of 0.920, 0.866, and 0.885, respectively. Our corpus and models are open-source and available at https://sina.birzeit.edu/wojood/.
Authors: Yucheng Li
Data contamination in language model evaluation is increasingly prevalent as the popularity of large language models. It allows models to "cheat" via memorisation instead of displaying true capabilities. Therefore, contamination analysis has became an crucial part of reliable model evaluation to validate results. However, existing contamination analysis is usually conducted internally by LLM developers and often lacks transparency and completeness. This paper present an open source data contamination reports for the Llama series models. We analyse six popular multi-choice QA benchmarks and quantify their overlapping with the training set of Llama. Various levels of contamination ranging from 1\% to 8.7\% are found across benchmarks. Our comparison also reveals that Llama models can gain over 5\% higher accuracy on contaminated subsets versus clean subsets. Data and code are available at: https://github.com/liyucheng09/Contamination_Detector.
Authors: Ao Zhang, Yuan Yao, Wei Ji, Zhiyuan Liu, Tat-Seng Chua
The development of large language models (LLMs) has greatly advanced the field of multimodal understanding, leading to the emergence of large multimodal models (LMMs). In order to enhance the level of visual comprehension, recent studies have equipped LMMs with region-level understanding capabilities by representing object bounding box coordinates as a series of text sequences (pix2seq). In this paper, we introduce a novel paradigm for object location modeling called pix2emb method, where we ask the LMM to output the location embeddings and then decode them with different decoders. This paradigm allows us to use different location formats (such as bounding boxes and masks) in multimodal conversations. Leveraging the proposed pix2emb method, we train an LMM named NExT-Chat and demonstrate its capability of handling multiple tasks like visual grounding, region captioning, and grounded reasoning. Comprehensive experiments show the effectiveness of our NExT-Chat on various tasks, e.g., NExT-Chat (87.7) vs. Shikra (86.9) on POPE-Random, NExT-Chat (68.9) vs. LISA (67.9) on referring expression segmentation task, and NExT-Chat (79.6) vs. Kosmos-2 (62.3) on region caption task. The code and model are released at https://github.com/NExT-ChatV/NExT-Chat.
Authors: Khushi Bhardwaj, Raj Sanjay Shah, Sashank Varma
Pre-trained Large Language Models (LLMs) have shown success in a diverse set of language inference and understanding tasks. The pre-training stage of LLMs looks at a large corpus of raw textual data. The BabyLM shared task compares LLM pre-training to human language acquisition, where the number of tokens seen by 13-year-old kids is magnitudes smaller than the number of tokens seen by LLMs. In this work, we pre-train and evaluate LLMs on their ability to learn contextual word representations using roughly the same number of tokens as seen by children. We provide a strong set of baselines; with different architectures, evaluation of changes in performance across epochs, and reported pre-training metrics for the strict small and strict tracks of the task. We also try to loosely replicate the RoBERTa baseline given by the task organizers to observe the training robustness to hyperparameter selection and replicability. We provide the submission details to the strict and strict-small tracks in this report.
Authors: Joshua Clymer, Garrett Baker, Rohan Subramani, Sam Wang
As AI systems become more intelligent and their behavior becomes more challenging to assess, they may learn to game the flaws of human feedback instead of genuinely striving to follow instructions; however, this risk can be mitigated by controlling how LLMs generalize human feedback to situations where it is unreliable. To better understand how reward models generalize, we craft 69 distribution shifts spanning 8 categories. We find that reward models do not learn to evaluate `instruction-following' by default and instead favor personas that resemble internet text. Techniques for interpreting reward models' internal representations achieve better generalization than standard fine-tuning, but still frequently fail to distinguish instruction-following from conflated behaviors. We consolidate the 15 most challenging distribution shifts into the GENeralization analogIES (GENIES) benchmark, which we hope will enable progress toward controlling reward model generalization.
Authors: Zhenran Xu, Senbao Shi, Baotian Hu, Jindi Yu, Dongfang Li, Min Zhang, Yuxiang Wu
Large Language Models (LLMs) have shown remarkable capabilities in general natural language processing tasks but often fall short in complex reasoning tasks. Recent studies have explored human-like problem-solving strategies, such as self-correct, to push further the boundary of single-model reasoning ability. In this work, we let a single model "step outside the box" by engaging multiple models to correct each other. We introduce a multi-agent collaboration strategy that emulates the academic peer review process. Each agent independently constructs its own solution, provides reviews on the solutions of others, and assigns confidence levels to its reviews. Upon receiving peer reviews, agents revise their initial solutions. Extensive experiments on three different types of reasoning tasks show that our collaboration approach delivers superior accuracy across all ten datasets compared to existing methods. Further study underscores the effectiveness of integrating confidence in reviews, demonstrates the superiority of feedback exchange over mere solution sharing, and highlights the role of capability and diversity in fostering successful collaboration.
Authors: Yuhan Li, Jian Wu, Zhiwei Yu, Börje F. Karlsson, Wei Shen, Manabu Okumura, Chin-Yew Lin
Extracting key information from scientific papers has the potential to help researchers work more efficiently and accelerate the pace of scientific progress. Over the last few years, research on Scientific Information Extraction (SciIE) witnessed the release of several new systems and benchmarks. However, existing paper-focused datasets mostly focus only on specific parts of a manuscript (e.g., abstracts) and are single-modality (i.e., text- or table-only), due to complex processing and expensive annotations. Moreover, core information can be present in either text or tables or across both. To close this gap in data availability and enable cross-modality IE, while alleviating labeling costs, we propose a semi-supervised pipeline for annotating entities in text, as well as entities and relations in tables, in an iterative procedure. Based on this pipeline, we release novel resources for the scientific community, including a high-quality benchmark, a large-scale corpus, and a semi-supervised annotation pipeline. We further report the performance of state-of-the-art IE models on the proposed benchmark dataset, as a baseline. Lastly, we explore the potential capability of large language models such as ChatGPT for the current task. Our new dataset, results, and analysis validate the effectiveness and efficiency of our semi-supervised pipeline, and we discuss its remaining limitations.
Authors: Shaoxiong Ji, Tianlin Zhang, Kailai Yang, Sophia Ananiadou, Erik Cambria
Large Language Models (LLMs) have become valuable assets in mental health, showing promise in both classification tasks and counseling applications. This paper offers a perspective on using LLMs in mental health applications. It discusses the instability of generative models for prediction and the potential for generating hallucinatory outputs, underscoring the need for ongoing audits and evaluations to maintain their reliability and dependability. The paper also distinguishes between the often interchangeable terms ``explainability'' and ``interpretability'', advocating for developing inherently interpretable methods instead of relying on potentially hallucinated self-explanations generated by LLMs. Despite the advancements in LLMs, human counselors' empathetic understanding, nuanced interpretation, and contextual awareness remain irreplaceable in the sensitive and complex realm of mental health counseling. The use of LLMs should be approached with a judicious and considerate mindset, viewing them as tools that complement human expertise rather than seeking to replace it.
Authors: Shi Yin Hong, Susan Gauch
Reliable automatic hate speech (HS) detection systems must adapt to the in-flow of diverse new data to curtail hate speech. However, hate speech detection systems commonly lack generalizability in identifying hate speech dissimilar to data used in training, impeding their robustness in real-world deployments. In this work, we propose a hate speech generalization framework that leverages emotion knowledge in a multitask architecture to improve the generalizability of hate speech detection in a cross-domain setting. We investigate emotion corpora with varying emotion categorical scopes to determine the best corpus scope for supplying emotion knowledge to foster generalized hate speech detection. We further assess the relationship between using pretrained Transformers models adapted for hate speech and its effect on our emotion-enriched hate speech generalization model. We perform extensive experiments on six publicly available datasets sourced from different online domains and show that our emotion-enriched HS detection generalization method demonstrates consistent generalization improvement in cross-domain evaluation, increasing generalization performance up to 18.1% and average cross-domain performance up to 8.5%, according to the F1 measure.
Authors: Shaohua Wu, Xudong Zhao, Shenling Wang, Jiangang Luo, Lingjun Li, Xi Chen, Bing Zhao, Wei Wang, Tong Yu, Rongguo Zhang, Jiahua Zhang, Chao Wang
In this work, we develop and release Yuan 2.0, a series of large language models with parameters ranging from 2.1 billion to 102.6 billion. The Localized Filtering-based Attention (LFA) is introduced to incorporate prior knowledge of local dependencies of natural language into Attention. A data filtering and generating system is presented to build pre-training and fine-tuning dataset in high quality. A distributed training method with non-uniform pipeline parallel, data parallel, and optimizer parallel is proposed, which greatly reduces the bandwidth requirements of intra-node communication, and achieves good performance in large-scale distributed training. Yuan 2.0 models display impressive ability in code generation, math problem-solving, and chatting compared with existing models. The latest version of YUAN 2.0, including model weights and source code, is accessible at Github.
Authors: Xiao Wang, Yaoyu Li, Tian Gan, Zheng Zhang, Jingjing Lv, Liqiang Nie
Recent advancements in video-language understanding have been established on the foundation of image-text models, resulting in promising outcomes due to the shared knowledge between images and videos. However, video-language understanding presents unique challenges due to the inclusion of highly complex semantic details, which result in information redundancy, temporal dependency, and scene complexity. Current techniques have only partially tackled these issues, and our quantitative analysis indicates that some of these methods are complementary. In light of this, we propose a novel framework called RTQ (Refine, Temporal model, and Query), which addresses these challenges simultaneously. The approach involves refining redundant information within frames, modeling temporal relations among frames, and querying task-specific information from the videos. Remarkably, our model demonstrates outstanding performance even in the absence of video-language pre-training, and the results are comparable with or superior to those achieved by state-of-the-art pre-training methods. Code is available at https://github.com/SCZwangxiao/RTQ-MM2023.
Authors: Qiang Li, Xiaoyan Yang, Haowen Wang, Qin Wang, Lei Liu, Junjie Wang, Yang Zhang, Mingyuan Chu, Sen Hu, Yicheng Chen, Yue Shen, Cong Fan, Wangshu Zhang, Teng Xu, Jinjie Gu, Jing Zheng, Guannan Zhang Ant Group
Recently, large language model (LLM) based artificial intelligence (AI) systems have demonstrated remarkable capabilities in natural language understanding and generation. However, these models face a significant challenge when it comes to sensitive applications, such as reasoning over medical knowledge and answering medical questions in a physician-like manner. Prior studies attempted to overcome this challenge by increasing the model size (>100B) to learn more general medical knowledge, while there is still room for improvement in LLMs with smaller-scale model sizes (<100B). In this work, we start from a pre-trained general LLM model (AntGLM-10B) and fine-tune it from a medical beginner towards a medical expert (called AntGLM-Med-10B), which leverages a 3-stage optimization procedure, i.e., general medical knowledge injection, medical domain instruction tuning, and specific medical task adaptation. Our contributions are threefold: (1) We specifically investigate how to adapt a pre-trained general LLM in medical domain, especially for a specific medical task. (2) We collect and construct large-scale medical datasets for each stage of the optimization process. These datasets encompass various data types and tasks, such as question-answering, medical reasoning, multi-choice questions, and medical conversations. (3) Specifically for multi-choice questions in the medical domain, we propose a novel Verification-of-Choice approach for prompting engineering, which significantly enhances the reasoning ability of LLMs. Remarkably, by combining the above approaches, our AntGLM-Med-10B model can outperform the most of LLMs on PubMedQA, including both general and medical LLMs, even when these LLMs have larger model size.
Authors: Y.Sun, J.Zhao, C.Yu, W.Wang, X.Zhou
The large language models represented by ChatGPT have a disruptive impact on the field of artificial intelligence. But it mainly focuses on natural language processing, speech recognition, machine learning and natural language understanding. This paper innovatively applies the large language model to the field of intelligent decision-making, places the large language model in the decision-making center, and constructs an agent architecture with the large language model as the core. Based on this, it further proposes a two-layer agent task planning, issues and executes decision commands through the interaction of natural language, and carries out simulation verification through the wargame simulation environment. Through the game confrontation simulation experiment, it is found that the intelligent decision-making ability of the large language model is significantly stronger than the commonly used reinforcement learning AI and rule AI, and the intelligence, understandability and generalization are all better. And through experiments, it was found that the intelligence of the large language model is closely related to prompt. This work also extends the large language model from previous human-computer interaction to the field of intelligent decision-making, which has important reference value and significance for the development of intelligent decision-making.
Authors: Kan Hatakeyama-Sato, Yasuhiko Igarashi, Shun Katakami, Yuta Nabae, Teruaki Hayakawa
Through additional training, we explore embedding specialized scientific knowledge into the Llama 2 Large Language Model (LLM). Key findings reveal that effective knowledge integration requires reading texts from multiple perspectives, especially in instructional formats. We utilize text augmentation to tackle the scarcity of specialized texts, including style conversions and translations. Hyperparameter optimization proves crucial, with different size models (7b, 13b, and 70b) reasonably undergoing additional training. Validating our methods, we construct a dataset of 65,000 scientific papers. Although we have succeeded in partially embedding knowledge, the study highlights the complexities and limitations of incorporating specialized information into LLMs, suggesting areas for further improvement.
Authors: Yue Zhang, Ming Zhang, Haipeng Yuan, Shichun Liu, Yongyao Shi, Tao Gui, Qi Zhang, Xuanjing Huang
Recently, the evaluation of Large Language Models has emerged as a popular area of research. The three crucial questions for LLM evaluation are ``what, where, and how to evaluate''. However, the existing research mainly focuses on the first two questions, which are basically what tasks to give the LLM during testing and what kind of knowledge it should deal with. As for the third question, which is about what standards to use, the types of evaluators, how to score, and how to rank, there hasn't been much discussion. In this paper, we analyze evaluation methods by comparing various criteria with both manual and automatic evaluation, utilizing onsite, crowd-sourcing, public annotators and GPT-4, with different scoring methods and ranking systems. We propose a new dataset, LLMEval and conduct evaluations on 20 LLMs. A total of 2,186 individuals participated, leading to the generation of 243,337 manual annotations and 57,511 automatic evaluation results. We perform comparisons and analyses of different settings and conduct 10 conclusions that can provide some insights for evaluating LLM in the future. The dataset and the results are publicly available at https://github.com/llmeval .
Authors: Manish Nagireddy, Lamogha Chiazor, Moninder Singh, Ioana Baldini
Current datasets for unwanted social bias auditing are limited to studying protected demographic features such as race and gender. In this work, we introduce a comprehensive benchmark that is meant to capture the amplification of social bias, via stigmas, in generative language models. Taking inspiration from social science research, we start with a documented list of 93 US-centric stigmas and curate a question-answering (QA) dataset which involves simple social situations. Our benchmark, SocialStigmaQA, contains roughly 10K prompts, with a variety of prompt styles, carefully constructed to systematically test for both social bias and model robustness. We present results for SocialStigmaQA with two open source generative language models and we find that the proportion of socially biased output ranges from 45% to 59% across a variety of decoding strategies and prompting styles. We demonstrate that the deliberate design of the templates in our benchmark (e.g., adding biasing text to the prompt or using different verbs that change the answer that indicates bias) impacts the model tendencies to generate socially biased output. Additionally, through manual evaluation, we discover problematic patterns in the generated chain-of-thought output that range from subtle bias to lack of reasoning.
Warning: This paper contains examples of text which are toxic, biased, and potentially harmful.
Authors: Aiwei Liu, Leyi Pan, Yijian Lu, Jingjing Li, Xuming Hu, Lijie Wen, Irwin King, Philip S. Yu
In recent years, significant advancements have been made in the text generation capabilities of Large Language Models (LLMs), demonstrating exceptional performance in downstream tasks such as abstract summarization, dialogue generation, and data-to-text conversion. However, their generative abilities also pose risks such as the rapid spread of fake news, infringement of datasets/LLM copyrights, and challenges to academic integrity. Text watermarking technology emerges as a potential solution. By embedding invisible yet detectable patterns in generated texts, it helps in tracking and verifying text origins, thus preventing misuse and piracy.
This survey aims to comprehensively summarize current text watermarking technologies, covering three main aspects: (1) an overview and comparison of different text watermarking techniques; (2) evaluation methods for text watermarking algorithms, including their success rate, impact on text quality, robustness, and unforgeability; (3) potential applications of text watermarking technologies. This survey aims to help researchers thoroughly understanding the text watermarking technologies, thereby fostering further development.
Authors: Golnaz Shapurian, Michael J Kurtz, Alberto Accomazzi
The automatic identification of planetary feature names in astronomy publications presents numerous challenges. These features include craters, defined as roughly circular depressions resulting from impact or volcanic activity; dorsas, which are elongate raised structures or wrinkle ridges; and lacus, small irregular patches of dark, smooth material on the Moon, referred to as "lake" (Planetary Names Working Group, n.d.). Many feature names overlap with places or people's names that they are named after, for example, Syria, Tempe, Einstein, and Sagan, to name a few (U.S. Geological Survey, n.d.). Some feature names have been used in many contexts, for instance, Apollo, which can refer to mission, program, sample, astronaut, seismic, seismometers, core, era, data, collection, instrument, and station, in addition to the crater on the Moon. Some feature names can appear in the text as adjectives, like the lunar craters Black, Green, and White. Some feature names in other contexts serve as directions, like craters West and South on the Moon. Additionally, some features share identical names across different celestial bodies, requiring disambiguation, such as the Adams crater, which exists on both the Moon and Mars. We present a multi-step pipeline combining rule-based filtering, statistical relevance analysis, part-of-speech (POS) tagging, named entity recognition (NER) model, hybrid keyword harvesting, knowledge graph (KG) matching, and inference with a locally installed large language model (LLM) to reliably identify planetary names despite these challenges. When evaluated on a dataset of astronomy papers from the Astrophysics Data System (ADS), this methodology achieves an F1-score over 0.97 in disambiguating planetary feature names.
Authors: Xiaoxia Wu, Haojun Xia, Stephen Youn, Zhen Zheng, Shiyang Chen, Arash Bakhtiari, Michael Wyatt, Reza Yazdani Aminabadi, Yuxiong He, Olatunji Ruwase, Leon Song, Zhewei Yao
This study examines 4-bit quantization methods like GPTQ in large language models (LLMs), highlighting GPTQ's overfitting and limited enhancement in Zero-Shot tasks. While prior works merely focusing on zero-shot measurement, we extend task scope to more generative categories such as code generation and abstractive summarization, in which we found that INT4 quantization can significantly underperform. However, simply shifting to higher precision formats like FP6 has been particularly challenging, thus overlooked, due to poor performance caused by the lack of sophisticated integration and system acceleration strategies on current AI hardware. Our results show that FP6, even with a coarse-grain quantization scheme, performs robustly across various algorithms and tasks, demonstrating its superiority in accuracy and versatility. Notably, with the FP6 quantization, \codestar-15B model performs comparably to its FP16 counterpart in code generation, and for smaller models like the 406M it closely matches their baselines in summarization. Neither can be achieved by INT4. To better accommodate various AI hardware and achieve the best system performance, we propose a novel 4+2 design for FP6 to achieve similar latency to the state-of-the-art INT4 fine-grain quantization. With our design, FP6 can become a promising solution to the current 4-bit quantization methods used in LLMs.
Authors: Bo Li, Wei Ye, Quansen Wang, Wen Zhao, Shikun Zhang
Textual label names (descriptions) are typically semantically rich in many natural language understanding (NLU) tasks. In this paper, we incorporate the prompting methodology, which is widely used to enrich model input, into the label side for the first time. Specifically, we propose a Mask Matching method, which equips an input with a prompt and its label with another, and then makes predictions by matching their mask representations. We evaluate our method extensively on 8 NLU tasks with 14 datasets. The experimental results show that Mask Matching significantly outperforms its counterparts of fine-tuning and conventional prompt-tuning, setting up state-of-the-art performances in several datasets. Mask Matching is particularly good at handling NLU tasks with large label counts and informative label names. As pioneering efforts that investigate the label-side prompt, we also discuss open issues for future study.
Authors: Thinh Pham, Dat Quoc Nguyen
Profile-based intent detection and slot filling are important tasks aimed at reducing the ambiguity in user utterances by leveraging user-specific supporting profile information. However, research in these two tasks has not been extensively explored. To fill this gap, we propose a joint model, namely JPIS, designed to enhance profile-based intent detection and slot filling. JPIS incorporates the supporting profile information into its encoder and introduces a slot-to-intent attention mechanism to transfer slot information representations to intent detection. Experimental results show that our JPIS substantially outperforms previous profile-based models, establishing a new state-of-the-art performance in overall accuracy on the Chinese benchmark dataset ProSLU.
Authors: Tony T. Wang, Miles Wang, Kaivalya Hariharan, Nir Shavit
LLMs often face competing pressures (for example helpfulness vs. harmlessness). To understand how models resolve such conflicts, we study Llama-2-chat models on the forbidden fact task. Specifically, we instruct Llama-2 to truthfully complete a factual recall statement while forbidding it from saying the correct answer. This often makes the model give incorrect answers. We decompose Llama-2 into 1000+ components, and rank each one with respect to how useful it is for forbidding the correct answer. We find that in aggregate, around 35 components are enough to reliably implement the full suppression behavior. However, these components are fairly heterogeneous and many operate using faulty heuristics. We discover that one of these heuristics can be exploited via a manually designed adversarial attack which we call The California Attack. Our results highlight some roadblocks standing in the way of being able to successfully interpret advanced ML systems. Project website available at https://forbiddenfacts.github.io .
Authors: Zichao Wei, Yewei Qin
Linguistic landscape is an important field in sociolinguistic research. Eye tracking technology is a common technology in psychological research. There are few cases of using eye movement to study linguistic landscape. This paper uses eye tracking technology to study the actual fixation of the linguistic landscape and finds that in the two dimensions of fixation time and fixation times, the fixation of native Chinese speakers to the linguistic landscape is higher than that of the general landscape. This paper argues that this phenomenon is due to the higher information density of linguistic landscapes. At the same time, the article also discusses other possible reasons for this phenomenon.
Authors: Haoran Liao, Qinyi Du, Shaohua Hu, Hao He, Yanyan Xu, Jidong Tian, Yaohui Jin
Large language models (LLMs) face challenges in solving complex mathematical problems that require comprehensive capacities to parse the statements, associate domain knowledge, perform compound logical reasoning, and integrate the intermediate rationales. Tackling all these problems once could be arduous for LLMs, thus leading to confusion in generation. In this work, we explore the potential of enhancing LLMs with agents by meticulous decomposition and modeling of mathematical reasoning process. Specifically, we propose a formal description of the mathematical solving and extend LLMs with an agent-based zero-shot framework named $\bf{P}$lanner-$\bf{R}$easoner-$\bf{E}$xecutor-$\bf{R}$eflector (PRER). We further provide and implement two MathAgents that define the logical forms and inherent relations via a pool of actions in different grains and orientations: MathAgent-M adapts its actions to LLMs, while MathAgent-H aligns with humankind. Experiments on miniF2F and MATH have demonstrated the effectiveness of PRER and proposed MathAgents, achieving an increase of $12.3\%$($53.9\%\xrightarrow{}66.2\%$) on the MiniF2F, $9.2\%$ ($49.8\%\xrightarrow{}59.0\%$) on MATH, and $13.2\%$($23.2\%\xrightarrow{}35.4\%$) for level-5 problems of MATH against GPT-4. Further analytical results provide more insightful perspectives on exploiting the behaviors of LLMs as agents.
Authors: Shengyao Zhang, Mi Zhang, Xudong Pan, Min Yang
To reduce the computation cost and the energy consumption in large language models (LLM), skimming-based acceleration dynamically drops unimportant tokens of the input sequence progressively along layers of the LLM while preserving the tokens of semantic importance. However, our work for the first time reveals the acceleration may be vulnerable to Denial-of-Service (DoS) attacks. In this paper, we propose No-Skim, a general framework to help the owners of skimming-based LLM to understand and measure the robustness of their acceleration scheme. Specifically, our framework searches minimal and unnoticeable perturbations at character-level and token-level to generate adversarial inputs that sufficiently increase the remaining token ratio, thus increasing the computation cost and energy consumption. We systematically evaluate the vulnerability of the skimming acceleration in various LLM architectures including BERT and RoBERTa on the GLUE benchmark. In the worst case, the perturbation found by No-Skim substantially increases the running cost of LLM by over 145% on average. Moreover, No-Skim extends the evaluation framework to various scenarios, making the evaluation conductible with different level of knowledge.
Authors: Shihan Dou, Enyu Zhou, Yan Liu, Songyang Gao, Jun Zhao, Wei Shen, Yuhao Zhou, Zhiheng Xi, Xiao Wang, Xiaoran Fan, Shiliang Pu, Jiang Zhu, Rui Zheng, Tao Gui, Qi Zhang, Xuanjing Huang
Supervised fine-tuning (SFT) is a crucial step for large language models (LLMs), enabling them to align with human instructions and enhance their capabilities in downstream tasks. When the models are required to align with a broader range of downstream tasks, or there is a desire to notably improve the performance on a specific task, a substantial increase in fine-tuning data often emerges as the solution. However, we find that large-scale increases in instruction data can disrupt the world knowledge previously stored in the LLMs, i.e., world knowledge forgetting. In this paper, we introduce LoRAMoE to address the above challenge. The LoRAMoE is a plugin version of Mixture of Experts (MoE). The plugin form ensures the integrity of world knowledge by freezing the backbone model during the training phase. We then propose the use of localized balancing constraints to coordinate parts of experts for task utilization, meanwhile enabling other experts to fully leverage the world knowledge stored in the models. Experimental results demonstrate that LoRAMoE can reasonably coordinate experts based on data type during inference, and even dramatically increasing instruction data does not result in knowledge forgetting. Moreover, LoRAMoE provides additional benefits for the performance of downstream tasks, indicating the potential of our approach for multi-task learning.
Authors: Zengqing Wu, Run Peng, Xu Han, Shuyuan Zheng, Yixin Zhang, Chuan Xiao
Computer simulations offer a robust toolset for exploring complex systems across various disciplines. A particularly impactful approach within this realm is Agent-Based Modeling (ABM), which harnesses the interactions of individual agents to emulate intricate system dynamics. ABM's strength lies in its bottom-up methodology, illuminating emergent phenomena by modeling the behaviors of individual components of a system. Yet, ABM has its own set of challenges, notably its struggle with modeling natural language instructions and common sense in mathematical equations or rules. This paper seeks to transcend these boundaries by integrating Large Language Models (LLMs) like GPT into ABM. This amalgamation gives birth to a novel framework, Smart Agent-Based Modeling (SABM). Building upon the concept of smart agents -- entities characterized by their intelligence, adaptability, and computation ability -- we explore in the direction of utilizing LLM-powered agents to simulate real-world scenarios with increased nuance and realism. In this comprehensive exploration, we elucidate the state of the art of ABM, introduce SABM's potential and methodology, and present three case studies (source codes available at https://github.com/Roihn/SABM), demonstrating the SABM methodology and validating its effectiveness in modeling real-world systems. Furthermore, we cast a vision towards several aspects of the future of SABM, anticipating a broader horizon for its applications. Through this endeavor, we aspire to redefine the boundaries of computer simulations, enabling a more profound understanding of complex systems.