Query rewriting plays a vital role in enhancing conversational search by transforming context-dependent user queries into standalone forms. Existing approaches primarily leverage human-rewritten queries as labels to train query rewriting models. However, human rewrites may lack sufficient information for optimal retrieval performance. To overcome this limitation, we propose utilizing large language models (LLMs) as query rewriters, enabling the generation of informative query rewrites through well-designed instructions. We define four essential properties for well-formed rewrites and incorporate all of them into the instruction. In addition, we introduce the role of rewrite editors for LLMs when initial query rewrites are available, forming a "rewrite-then-edit" process. Furthermore, we propose distilling the rewriting capabilities of LLMs into smaller models to reduce rewriting latency. Our experimental evaluation on the QReCC dataset demonstrates that informative query rewrites can yield substantially improved retrieval performance compared to human rewrites, especially with sparse retrievers.
Large language models (LMs) are currently trained to predict tokens given document prefixes, enabling them to directly perform long-form generation and prompting-style tasks which can be reduced to document completion. Existing pretraining pipelines train LMs by concatenating random sets of short documents to create input contexts but the prior documents provide no signal for predicting the next document. We instead present In-Context Pretraining, a new approach where language models are pretrained on a sequence of related documents, thereby explicitly encouraging them to read and reason across document boundaries. We can do In-Context Pretraining by simply changing the document ordering so that each context contains related documents, and directly applying existing pretraining pipelines. However, this document sorting problem is challenging. There are billions of documents and we would like the sort to maximize contextual similarity for every document without repeating any data. To do this, we introduce approximate algorithms for finding related documents with efficient nearest neighbor search and constructing coherent input contexts with a graph traversal algorithm. Our experiments show In-Context Pretraining offers a simple and scalable approach to significantly enhance LMs'performance: we see notable improvements in tasks that require more complex contextual reasoning, including in-context learning (+8%), reading comprehension (+15%), faithfulness to previous contexts (+16%), long-context reasoning (+5%), and retrieval augmentation (+9%).
The ability to automatically generate accurate protocols for scientific experiments would represent a major step towards the automation of science. Large Language Models (LLMs) have impressive capabilities on a wide range of tasks, such as question answering and the generation of coherent text and code. However, LLMs can struggle with multi-step problems and long-term planning, which are crucial for designing scientific experiments. Moreover, evaluation of the accuracy of scientific protocols is challenging, because experiments can be described correctly in many different ways, require expert knowledge to evaluate, and cannot usually be executed automatically. Here we present an automatic evaluation framework for the task of planning experimental protocols, and we introduce BioProt: a dataset of biology protocols with corresponding pseudocode representations. To measure performance on generating scientific protocols, we use an LLM to convert a natural language protocol into pseudocode, and then evaluate an LLM's ability to reconstruct the pseudocode from a high-level description and a list of admissible pseudocode functions. We evaluate GPT-3 and GPT-4 on this task and explore their robustness. We externally validate the utility of pseudocode representations of text by generating accurate novel protocols using retrieved pseudocode, and we run a generated protocol successfully in our biological laboratory. Our framework is extensible to the evaluation and improvement of language model planning abilities in other areas of science or other areas that lack automatic evaluation.
We present Llemma, a large language model for mathematics. We continue pretraining Code Llama on the Proof-Pile-2, a mixture of scientific papers, web data containing mathematics, and mathematical code, yielding Llemma. On the MATH benchmark Llemma outperforms all known open base models, as well as the unreleased Minerva model suite on an equi-parameter basis. Moreover, Llemma is capable of tool use and formal theorem proving without any further finetuning. We openly release all artifacts, including 7 billion and 34 billion parameter models, the Proof-Pile-2, and code to replicate our experiments.
The advent of artificial intelligence has led to a growing emphasis on data-driven modeling in macroeconomics, with agent-based modeling (ABM) emerging as a prominent bottom-up simulation paradigm. In ABM, agents (e.g., households, firms) interact within a macroeconomic environment, collectively generating market dynamics. Existing agent modeling typically employs predetermined rules or learning-based neural networks for decision-making. However, customizing each agent presents significant challenges, complicating the modeling of agent heterogeneity. Additionally, the influence of multi-period market dynamics and multifaceted macroeconomic factors are often overlooked in decision-making processes. In this work, we introduce EconAgent, a large language model-empowered agent with human-like characteristics for macroeconomic simulation. We first construct a simulation environment that incorporates various market dynamics driven by agents' decisions regarding work and consumption. Through the perception module, we create heterogeneous agents with distinct decision-making mechanisms. Furthermore, we model the impact of macroeconomic trends using a memory module, which allows agents to reflect on past individual experiences and market dynamics. Simulation experiments show that EconAgent can make realistic decisions, leading to more reasonable macroeconomic phenomena compared to existing rule-based or learning-based agents. Our codes are released at https://github.com/tsinghua-fib-lab/ACL24-EconAgent.
The large language model (LLM) has garnered significant attention due to its in-context learning mechanisms and emergent capabilities. The research community has conducted several pilot studies to apply LLMs to machine translation tasks and evaluate their performance from diverse perspectives. However, previous research has primarily focused on the LLM itself and has not explored human intervention in the inference process of LLM. The characteristics of LLM, such as in-context learning and prompt engineering, closely mirror human cognitive abilities in language tasks, offering an intuitive solution for human-in-the-loop generation. In this study, we propose a human-in-the-loop pipeline that guides LLMs to produce customized outputs with revision instructions. The pipeline initiates by prompting the LLM to produce a draft translation, followed by the utilization of automatic retrieval or human feedback as supervision signals to enhance the LLM's translation through in-context learning. The human-machine interactions generated in this pipeline are also stored in an external database to expand the in-context retrieval database, enabling us to leverage human supervision in an offline setting. We evaluate the proposed pipeline using GPT-3.5-turbo API on five domain-specific benchmarks for German-English translation. The results demonstrate the effectiveness of the pipeline in tailoring in-domain translations and improving translation performance compared to direct translation. Additionally, we discuss the results from the following perspectives: 1) the effectiveness of different in-context retrieval methods; 2) the construction of a retrieval database under low-resource scenarios; 3) the observed domains differences; 4) the quantitative analysis of linguistic statistics; and 5) the qualitative analysis of translation cases. The code and data are available at https://github.com/NLP2CT/HIL-MT/.
Natural language processing (NLP) has made significant progress for well-resourced languages such as English but lagged behind for low-resource languages like Setswana. This paper addresses this gap by presenting PuoBERTa, a customised masked language model trained specifically for Setswana. We cover how we collected, curated, and prepared diverse monolingual texts to generate a high-quality corpus for PuoBERTa's training. Building upon previous efforts in creating monolingual resources for Setswana, we evaluated PuoBERTa across several NLP tasks, including part-of-speech (POS) tagging, named entity recognition (NER), and news categorisation. Additionally, we introduced a new Setswana news categorisation dataset and provided the initial benchmarks using PuoBERTa. Our work demonstrates the efficacy of PuoBERTa in fostering NLP capabilities for understudied languages like Setswana and paves the way for future research directions.
Integrating large language models (LLMs) into healthcare holds great potential but faces challenges. Pre-training LLMs from scratch for domains like medicine is resource-heavy and often unfeasible. On the other hand, sole reliance on Supervised Fine-tuning (SFT) can result in overconfident predictions and may not tap into domain-specific insights. In response, we present a multi-stage training method combining Domain-specific Continued Pre-training (DCPT), SFT, and Direct Preference Optimization (DPO). In addition, we publish a 3Gb Chinese Medicine (ChiMed) dataset, encompassing medical question answering, plain texts, knowledge graphs, and dialogues, segmented into three training stages. The medical LLM trained with our pipeline, Qilin-Med, shows substantial performance improvement. In the CPT and SFT phases, Qilin-Med achieved 38.4% and 40.0% accuracy on the CMExam test set, respectively. It outperformed the basemodel Baichuan-7B (accuracy: 33.5%), by 7.5%. In the DPO phase, it scored 16.66 in BLEU-1 and 27.44 in ROUGE-1 on the Huatuo-26M test set, bringing further improvement to the SFT phase (12.69 in BLEU-1 and 24.21 in ROUGE-1). Additionally, we have further enhanced the model's performance through the Retrieval Augmented Generation (RAG) approach. Experiments demonstrate that Qilin-Med-RAG achieves an accuracy rate of 42.8% on CMExam. These results highlight the contribution of our novel training approach in building LLMs for medical applications.
Large Language Models (LLMs) excel in natural language understanding by capturing hidden semantics in vector space. This process enriches the value of text embeddings for various downstream tasks, thereby fostering the Embedding-as-a-Service (EaaS) business model. However, the risk of privacy leakage due to direct text transmission to servers remains a critical concern. To address this, we introduce Split-N-Denoise (SnD), an private inference framework that splits the model to execute the token embedding layer on the client side at minimal computational cost. This allows the client to introduce noise prior to transmitting the embeddings to the server, and subsequently receive and denoise the perturbed output embeddings for downstream tasks. Our approach is designed for the inference stage of LLMs and requires no modifications to the model parameters. Extensive experiments demonstrate SnD's effectiveness in optimizing the privacy-utility tradeoff across various LLM architectures and diverse downstream tasks. The results reveal an improvement in performance under the same privacy budget compared to the baselines by over 10\% on average, offering clients a privacy-preserving solution for local privacy protection.
The growing capabilities and increasingly widespread deployment of AI systems necessitate robust benchmarks for measuring their cooperative capabilities. Unfortunately, most multi-agent benchmarks are either zero-sum or purely cooperative, providing limited opportunities for such measurements. We introduce a general-sum variant of the zero-sum board game Diplomacy -- called Welfare Diplomacy -- in which players must balance investing in military conquest and domestic welfare. We argue that Welfare Diplomacy facilitates both a clearer assessment of and stronger training incentives for cooperative capabilities. Our contributions are: (1) proposing the Welfare Diplomacy rules and implementing them via an open-source Diplomacy engine; (2) constructing baseline agents using zero-shot prompted language models; and (3) conducting experiments where we find that baselines using state-of-the-art models attain high social welfare but are exploitable. Our work aims to promote societal safety by aiding researchers in developing and assessing multi-agent AI systems. Code to evaluate Welfare Diplomacy and reproduce our experiments is available at https://github.com/mukobi/welfare-diplomacy.
While multi-modal models have successfully integrated information from image, video, and audio modalities, integrating graph modality into large language models (LLMs) remains unexplored. This discrepancy largely stems from the inherent divergence between structured graph data and unstructured text data. Incorporating graph knowledge provides a reliable source of information, enabling potential solutions to address issues in text generation, e.g., hallucination, and lack of domain knowledge. To evaluate the integration of graph knowledge into language models, a dedicated dataset is needed. However, there is currently no benchmark dataset specifically designed for multimodal graph-language models. To address this gap, we propose GraphextQA, a question answering dataset with paired subgraphs, retrieved from Wikidata, to facilitate the evaluation and future development of graph-language models. Additionally, we introduce a baseline model called CrossGNN, which conditions answer generation on the paired graphs by cross-attending question-aware graph features at decoding. The proposed dataset is designed to evaluate graph-language models' ability to understand graphs and make use of it for answer generation. We perform experiments with language-only models and the proposed graph-language model to validate the usefulness of the paired graphs and to demonstrate the difficulty of the task.
In the large language model (LLM) revolution, embedding is a key component of various systems. For example, it is used to retrieve knowledge or memories for LLMs, to build content moderation filters, etc. As such cases span from English to other natural or programming languages, from retrieval to classification and beyond, it is desirable to build a unified embedding model rather than dedicated ones for each scenario. In this work, we make an initial step towards this goal, demonstrating that multiple languages (both natural and programming) pre-trained transformer decoders can embed universally when finetuned on limited English data. We provide a comprehensive practice with thorough evaluations. On English MTEB, our models achieve competitive performance on different embedding tasks by minimal training data. On other benchmarks, such as multilingual classification and code search, our models (without any supervision) perform comparably to, or even surpass heavily supervised baselines and/or APIs. These results provide evidence of a promising path towards building powerful unified embedders that can be applied across tasks and languages.
The quadratic complexity of the attention module makes it gradually become the bulk of compute in Transformer-based LLMs during generation. Moreover, the excessive key-value cache that arises when dealing with long inputs also brings severe issues on memory footprint and inference latency. In this work, we propose a plug-and-play approach that is able to incrementally compress the intermediate activation of a specified span of tokens into compact ones, thereby reducing both memory and computational cost when processing subsequent context. Experiments on both in-domain language modeling and zero-shot open-ended document generation demonstrate the advantage of our approach over sparse attention baselines in terms of fluency, n-gram matching, and semantic similarity. At last, we comprehensively profile the benefit of context compression on improving the system throughout. Code is available at https://github.com/DRSY/KV_Compression.
Nowadays, Question Answering (QA) tasks receive significant research focus, particularly with the development of Large Language Model (LLM) such as Chat GPT [1]. LLM can be applied to various domains, but it contradicts the principles of information transmission when applied to the Islamic domain. In Islam we strictly regulates the sources of information and who can give interpretations or tafseer for that sources [2]. The approach used by LLM to generate answers based on its own interpretation is similar to the concept of tafseer, LLM is neither an Islamic expert nor a human which is not permitted in Islam. Indonesia is the country with the largest Islamic believer population in the world [3]. With the high influence of LLM, we need to make evaluation of LLM in religious domain. Currently, there is only few religious QA dataset available and none of them using Sirah Nabawiyah especially in Indonesian Language. In this paper, we propose the Question Answering Sirah Nabawiyah (QASiNa) dataset, a novel dataset compiled from Sirah Nabawiyah literatures in Indonesian language. We demonstrate our dataset by using mBERT [4], XLM-R [5], and IndoBERT [6] which fine-tuned with Indonesian translation of SQuAD v2.0 [7]. XLM-R model returned the best performance on QASiNa with EM of 61.20, F1-Score of 75.94, and Substring Match of 70.00. We compare XLM-R performance with Chat GPT-3.5 and GPT-4 [1]. Both Chat GPT version returned lower EM and F1-Score with higher Substring Match, the gap of EM and Substring Match get wider in GPT-4. The experiment indicate that Chat GPT tends to give excessive interpretations as evidenced by its higher Substring Match scores compared to EM and F1-Score, even after providing instruction and context. This concludes Chat GPT is unsuitable for question answering task in religious domain especially for Islamic religion.
Semantic identifier (ID) is an important concept in information retrieval that aims to preserve the semantics of objects such as documents and items inside their IDs. Previous studies typically adopt a two-stage pipeline to learn semantic IDs by first procuring embeddings using off-the-shelf text encoders and then deriving IDs based on the embeddings. However, each step introduces potential information loss, and there is usually an inherent mismatch between the distribution of embeddings within the latent space produced by text encoders and the anticipated distribution required for semantic indexing. It is non-trivial to design a method that can learn the document's semantic representations and its hierarchical structure simultaneously, given that semantic IDs are discrete and sequentially structured, and the semantic supervision is deficient. In this paper, we introduce LMIndexer, a self-supervised framework to learn semantic IDs with a generative language model. We tackle the challenge of sequential discrete ID by introducing a semantic indexer capable of generating neural sequential discrete representations with progressive training and contrastive learning. In response to the semantic supervision deficiency, we propose to train the model with a self-supervised document reconstruction objective. We show the high quality of the learned IDs and demonstrate their effectiveness on three tasks including recommendation, product search, and document retrieval on five datasets from various domains. Code is available at https://github.com/PeterGriffinJin/LMIndexer.
Recently, using a powerful proprietary Large Language Model (LLM) (e.g., GPT-4) as an evaluator for long-form responses has become the de facto standard. However, for practitioners with large-scale evaluation tasks and custom criteria in consideration (e.g., child-readability), using proprietary LLMs as an evaluator is unreliable due to the closed-source nature, uncontrolled versioning, and prohibitive costs. In this work, we propose Prometheus, a fully open-source LLM that is on par with GPT-4's evaluation capabilities when the appropriate reference materials (reference answer, score rubric) are accompanied. We first construct the Feedback Collection, a new dataset that consists of 1K fine-grained score rubrics, 20K instructions, and 100K responses and language feedback generated by GPT-4. Using the Feedback Collection, we train Prometheus, a 13B evaluator LLM that can assess any given long-form text based on customized score rubric provided by the user. Experimental results show that Prometheus scores a Pearson correlation of 0.897 with human evaluators when evaluating with 45 customized score rubrics, which is on par with GPT-4 (0.882), and greatly outperforms ChatGPT (0.392). Furthermore, measuring correlation with GPT-4 with 1222 customized score rubrics across four benchmarks (MT Bench, Vicuna Bench, Feedback Bench, Flask Eval) shows similar trends, bolstering Prometheus's capability as an evaluator LLM. Lastly, Prometheus achieves the highest accuracy on two human preference benchmarks (HHH Alignment & MT Bench Human Judgment) compared to open-sourced reward models explicitly trained on human preference datasets, highlighting its potential as an universal reward model. We open-source our code, dataset, and model at https://kaistai.github.io/prometheus/.
Multimodal image-text memes are prevalent on the internet, serving as a unique form of communication that combines visual and textual elements to convey humor, ideas, or emotions. However, some memes take a malicious turn, promoting hateful content and perpetuating discrimination. Detecting hateful memes within this multimodal context is a challenging task that requires understanding the intertwined meaning of text and images. In this work, we address this issue by proposing a novel approach named ISSUES for multimodal hateful meme classification. ISSUES leverages a pre-trained CLIP vision-language model and the textual inversion technique to effectively capture the multimodal semantic content of the memes. The experiments show that our method achieves state-of-the-art results on the Hateful Memes Challenge and HarMeme datasets. The code and the pre-trained models are publicly available at https://github.com/miccunifi/ISSUES.
What is the relationship between model architecture and the ability to perform in-context learning? In this empirical study, we take the first steps toward answering this question. We evaluate thirteen model architectures capable of causal language modeling across a suite of synthetic in-context learning tasks. These selected architectures represent a broad range of paradigms, including recurrent and convolution-based neural networks, transformers, state space model inspired, and other emerging attention alternatives. We discover that all the considered architectures can perform in-context learning under a wider range of conditions than previously documented. Additionally, we observe stark differences in statistical efficiency and consistency by varying the number of in-context examples and task difficulty. We also measure each architecture's predisposition towards in-context learning when presented with the option to memorize rather than leverage in-context examples. Finally, and somewhat surprisingly, we find that several attention alternatives are sometimes competitive with or better in-context learners than transformers. However, no single architecture demonstrates consistency across all tasks, with performance either plateauing or declining when confronted with a significantly larger number of in-context examples than those encountered during gradient-based training.
Evaluating generalist agents presents significant challenges due to their wide-ranging abilities and the limitations of current benchmarks in assessing true generalization. We introduce the Minecraft Universe (MCU), a fully automated benchmarking framework set within the open-world game Minecraft. MCU dynamically generates and evaluates a broad spectrum of tasks, offering three core components: 1) a task generation mechanism that provides high degrees of freedom and variability, 2) an ever-expanding set of over 3K composable atomic tasks, and 3) a general evaluation framework that supports open-ended task assessment. By integrating large language models (LLMs), MCU dynamically creates diverse environments for each evaluation, fostering agent generalization. The framework uses a vision-language model (VLM) to automatically generate evaluation criteria, achieving over 90% agreement with human ratings across multi-dimensional assessments, which demonstrates that MCU is a scalable and explainable solution for evaluating generalist agents. Additionally, we show that while state-of-the-art foundational models perform well on specific tasks, they often struggle with increased task diversity and difficulty.
Knowledge base construction entails acquiring structured information to create a knowledge base of factual and relational data, facilitating question answering, information retrieval, and semantic understanding. The challenge called "Knowledge Base Construction from Pretrained Language Models" at International Semantic Web Conference 2023 defines tasks focused on constructing knowledge base using language model. Our focus was on Track 1 of the challenge, where the parameters are constrained to a maximum of 1 billion, and the inclusion of entity descriptions within the prompt is prohibited. Although the masked language model offers sufficient flexibility to extend its vocabulary, it is not inherently designed for multi-token prediction. To address this, we present Vocabulary Expandable BERT for knowledge base construction, which expand the language model's vocabulary while preserving semantic embeddings for newly added words. We adopt task-specific re-pre-training on masked language model to further enhance the language model. Through experimentation, the results show the effectiveness of our approaches. Our framework achieves F1 score of 0.323 on the hidden test set and 0.362 on the validation set, both data set is provided by the challenge. Notably, our framework adopts a lightweight language model (BERT-base, 0.13 billion parameters) and surpasses the model using prompts directly on large language model (Chatgpt-3, 175 billion parameters). Besides, Token-Recode achieves comparable performances as Re-pretrain. This research advances language understanding models by enabling the direct embedding of multi-token entities, signifying a substantial step forward in link prediction task in knowledge graph and metadata completion in data management.