ACL 2024 论文列表

Detecting whether copyright holders’ works were used in LLM pretraining is poised to be an important problem. This work proposes using data watermarks to enable principled detection with only black-box model access, provided that the rightholder contributed multiple training documents and watermarked them before public release. By applying a randomly sampled data watermark, detection can be framed as hypothesis testing, which provides guarantees on the false detection rate. We study two watermarks: one that inserts random sequences, and another that randomly substitutes characters with Unicode lookalikes. We first show how three aspects of watermark design - watermark length, number of duplications, and interference - affect the power of the hypothesis test. Next, we study how a watermark’s detection strength changes under model and dataset scaling: while increasing the dataset size decreases the strength of the watermark, watermarks remain strong if the model size also increases. Finally, we view SHA hashes as natural watermarks and show that we can robustly detect hashes from BLOOM-176B’s training data, as long as they occurred at least 90 times. Together, our results point towards a promising future for data watermarks in real world use.

The reward model for Reinforcement Learning from Human Feedback (RLHF) has proven effective in fine-tuning Large Language Models (LLMs). Notably, collecting human feedback for RLHF can be resource-intensive and lead to scalability issues for LLMs and complex tasks. Our proposed framework Proto-RM leverages prototypical networks to enhance reward models under limited human feedback. By enabling stable and reliable structural learning from fewer samples, Proto-RM significantly enhances LLMs' adaptability and accuracy in interpreting human preferences. Extensive experiments on various datasets demonstrate that Proto-RM significantly improves the performance of reward models and LLMs in human feedback tasks, achieving comparable and usually better results than traditional methods, while requiring significantly less data in data-limited scenarios. This research offers a promising direction for enhancing the efficiency of reward models and optimizing the fine-tuning of language models under restricted feedback conditions.

Language Models (LMs) excel in understanding textual descriptions of proteins, as evident in biomedical question-answering tasks. However, their capability falters with raw protein data, such as amino acid sequences, due to a deficit in pretraining on such data. Conversely, Protein Language Models (PLMs) can understand and convert protein data into high-quality representations, but struggle to process texts. To address their limitations, we introduce ProtT3, a framework for Protein-to-Text Generation for Text-based Protein Understanding. ProtT3 empowers an LM to understand protein sequences of amino acids by incorporating a PLM as its protein understanding module, enabling effective protein-to-text generation. This collaboration between PLM and LM is facilitated by a cross-modal projector (i.e., Q-Former) that bridges the modality gap between the PLM’s representation space and the LM’s input space. Unlike previous studies focusing on protein property prediction and protein-text retrieval, we delve into the largely unexplored field of protein-to-text generation. To facilitate comprehensive benchmarks and promote future research, we establish quantitative evaluations for protein-text modeling tasks, including protein captioning, protein question-answering, and protein-text retrieval. Our experiments show that ProtT3 substantially surpasses current baselines, with ablation studies further highlighting the efficacy of its core components. Our code is available at https://github.com/acharkq/ProtT3.

We propose ProtLLM, a versatile cross-modal large language model (LLM) for both protein-centric and protein-language tasks. ProtLLM features a unique dynamic protein mounting mechanism, enabling it to handle complex inputs where the natural language text is interspersed with an arbitrary number of proteins. Besides, we propose the protein-as-word language modeling approach to train ProtLLM. By developing a specialized protein vocabulary, we equip the model with the capability to predict not just natural language but also proteins from a vast pool of candidates. Additionally, we construct a large-scale interleaved protein-text dataset, named InterPT, for pre-training. This dataset comprehensively encompasses both (1) structured data sources like protein annotations and (2) unstructured data sources like biological research papers, thereby endowing ProtLLM with crucial knowledge for understanding proteins. We evaluate ProtLLM on classic supervised protein-centric tasks and explore its novel protein-language applications. Experimental results demonstrate that ProtLLM not only achieves superior performance against protein-specialized baselines on protein-centric tasks but also induces zero-shot and in-context learning capabilities on protein-language tasks.

Current knowledge editing approaches struggle to effectively propagate updates to interconnected facts.In this work, we delve into the barriers that hinder the appropriate propagation of updated knowledge within these models for accurate reasoning. To support our analysis, we introduce a novel reasoning-based benchmark, ReCoE (Reasoning-based Counterfactual Editing dataset), which covers six common reasoning schemes in the real world. We conduct an extensive analysis of existing knowledge editing techniques, including input-augmentation, finetuning, and locate-and-edit methods. We found that all model editing methods exhibit notably low performance on this dataset, especially within certain reasoning schemes. Our analysis of the chain-of-thought responses from edited models indicate that, while the models effectively update individual facts, they struggle to recall these facts in reasoning tasks. Moreover, locate-and-edit methods severely deteriorate the models’ language modeling capabilities, leading to poor perplexity and logical coherence in their outputs.

The impressive capabilities in Large Language Models (LLMs) provide a powerful approach to reimagine users’ typing experience. This paper demonstrates the Proofread feature in Gboard, a virtual keyboard running on mobile phones. Proofread enables seamless sentence-level and paragraph-level corrections with a single tap. We describe the complete system in this paper, from data generation, metrics design to model tuning and deployment. To obtain models with sufficient quality, we implement a careful data synthetic pipeline tailored to online use cases, design multifaceted metrics, employ a two-stage tuning approach to acquire the dedicated LLM for the feature: the Supervised Fine Tuning (SFT) for foundational quality, followed by the Reinforcement Learning (RL) tuning approach for targeted refinement. Specifically, we find sequential tuning on Rewrite and proofread tasks yields the best quality in SFT stage, and propose global and direct rewards in the RL tuning stage to seek further improvement. Extensive experiments on a human-labeled golden set showed our tuned PaLM2-XS model achieved 85.56% good ratio. We launched the feature to Pixel 8 devices by serving the model on TPU v5 in Google Cloud, with thousands of daily active users. Serving latency was significantly reduced by quantization, bucket inference, text segmentation, and speculative decoding. Our demo could be seen in Youtube.

Thanks to recent advances in generative AI, we are able to prompt large language models (LLMs) to produce texts which are fluent and grammatical. In addition, it has been shown that we can elicit attempts at grammatical error correction (GEC) from LLMs when prompted with ungrammatical input sentences. We evaluate how well LLMs can perform at GEC by measuring their performance on established benchmark datasets. We go beyond previous studies, which only examined GPT* models on a selection of English GEC datasets, by evaluating seven open-source and three commercial LLMs on four established GEC benchmarks. We investigate model performance and report results against individual error types. Our results indicate that LLMs do not always outperform supervised English GEC models except in specific contexts – namely commercial LLMs on benchmarks annotated with fluency corrections as opposed to minimal edits. We find that several open-source models outperform commercial ones on minimal edit benchmarks, and that in some settings zero-shot prompting is just as competitive as few-shot prompting.

Key Point Analysis (KPA) aims for quantitative summarization that provides key points (KPs) as succinct textual summaries and quantities measuring their prevalence. KPA studies for arguments and reviews have been reported in the literature. A majority of KPA studies for reviews adopt supervised learning to extract short sentences as KPs before matching KPs to review comments for quantification of KP prevalence. Recent abstractive approaches still generate KPs based on sentences, often leading to KPs with overlapping and hallucinated opinions, and inaccurate quantification. In this paper, we propose Prompted Aspect Key Point Analysis (PAKPA) for quantitative review summarization. PAKPA employs aspect sentiment analysis and prompted in-context learning with Large Language Models (LLMs) to generate and quantify KPs grounded in aspects for business entities, which achieves faithful KPs with accurate quantification, and removes the need for large amounts of annotated data for supervised training. Experiments on the popular review dataset Yelp and the aspect-oriented review summarization dataset SPACE show that our framework achieves state-of-the-art performance. Source code and data are available at: https://github.com/antangrocket1312/PAKPA

Large language models (LLMs) have attracted great attention given their strong performance on a wide range of NLP tasks. In practice, users often expect generated texts to fall within a specific length range, making length controlled generation an important topic, especially for GPT-style models. Existing length control methods mostly focus on a simple control type of “equal to” a target length. Different from them, we propose a prompt-based method to achieve length controlled generation under different control types with high accuracy. In particular, we adopt reinforcement learning (RL) and sample filtering with the reward signal given by rule-based reward models, which enhances the length control ability of models by rewarding outputs that follow certain control instructions. In addition, we introduce a standard prompt extractor to parse arbitrary users’ input into standard control instructions. Experiments show that our method significantly improves the accuracy of prompt-based length control on popular summarization datasets like CNNDM and NYT under multiple control types. Moreover, both the standard prompt extractor and RL-tuned model show strong generalization to unseen control prompt templates.

For text-to-image generation, automatically refining user-provided natural language prompts into the keyword-enriched prompts favored by systems is essential for the user experience. Such a prompt refinement process is analogous to translating the prompt from “user languages” into “system languages”. However, the scarcity of such parallel corpora makes it difficult to train a prompt refinement model. Inspired by zero-shot machine translation techniques, we introduce Prompt Refinement with Image Pivot (PRIP). PRIP innovatively uses the latent representation of a user-preferred image as an intermediary “pivot” between the user and system languages. It decomposes the refinement process into two data-rich tasks: inferring representations of user-preferred images from user languages and subsequently translating image representations into system languages. Thus, it can leverage abundant data for training. Extensive experiments show that PRIP substantially outperforms a wide range of baselines and effectively transfers to unseen systems in a zero-shot manner.

We propose a new method, Adversarial In-Context Learning (adv-ICL), to optimize prompts for in-context learning (ICL). Inspired by adversarial learning, adv-ICL is implemented as a two-player game between a generator and discriminator, with LLMs acting as both. In each round, given an input prefixed by task instructions and several exemplars, the generator produces an output. The discriminator then classifies the generator’s input-output pair as model-generated or real data. Based on the discriminator’s loss, a prompt modifier LLM proposes possible edits to the generator and discriminator prompts, and the edits that most improve the adversarial loss are selected. We show that applying adv-ICL results in significant improvements over state-of-the-art prompt optimization techniques for both open and closed-source models on 13 generation and classification tasks including summarization, arithmetic reasoning, machine translation, data-to-text generation, and the MMLU and big-bench hard benchmarks. In addition, our method is computationally efficient, easily extensible to other LLMs and tasks, and effective in low-resource settings.

Text-to-image generation models are powerful but difficult to use. Users craft specific prompts to get better images, though the images can be repetitive. This paper proposes the Prompt Expansion framework that helps users generate high-quality, diverse images with less effort. The Prompt Expansion model takes a text query as input and outputs a set of expanded text prompts that are optimized such that when passed to a text-to-image model, they generate a wider variety of appealing images. We conduct a human evaluation study that shows that images generated through Prompt Expansion are more aesthetically pleasing and diverse than those generated by baseline methods. Overall, this paper presents a novel and effective approach to improving the text-to-image generation experience.

Prompt engineering is a challenging yet crucial task for optimizing the performance of large language models on customized tasks. It requires complex reasoning to examine the model’s errors, hypothesize what is missing or misleading in the current prompt, and communicate the task with clarity. While recent works indicate that large language models can be meta-prompted to perform automatic prompt engineering, we argue that their potential is limited due to insufficient guidance for complex reasoning in the meta-prompt. We fill this gap by infusing into the meta-prompt three key components: detailed descriptions, context specification, and a step-by-step reasoning template. The resulting method, named PE2, showcases remarkable versatility across diverse language tasks. It finds prompts that outperform “let’s think step by step” by 6.3% on MultiArith and 3.1% on GSM8K, and outperforms competitive baselines on counterfactual tasks by 6.9%. Further, we show that PE2 can make targeted prompt edits, rectify erroneous prompts, and induce multi-step plans for complex tasks.

Assessing long-form responses generated by Vision-Language Models (VLMs) is challenging. It not only requires checking whether the VLM follows the given instruction but also verifying whether the text output is properly grounded on the given image. Inspired by the recent approach of evaluating LMs with LMs, in this work, we propose to evaluate VLMs with VLMs. For this purpose, we present a new feedback dataset called the Perception Collection, encompassing 15K customized score rubrics that users might care about during assessment. Using the Perception Collection, we train Prometheus-Vision, the first open-source VLM evaluator model that can understand the user-defined score criteria during evaluation. Prometheus-Vision shows the highest Pearson correlation with human evaluators and GPT-4V among open-source models, showing its effectiveness for transparent and accessible evaluation of VLMs. We open-source our code, dataset, and model.

Multi-Modal Entity Alignment aims to discover identical entities across heterogeneous knowledge graphs. While recent studies have delved into fusion paradigms to represent entities holistically, the elimination of features irrelevant to alignment and modal inconsistencies is overlooked, which are caused by inherent differences in multi-modal features. To address these challenges, we propose a novel strategy of progressive modality freezing, called PMF, that focuses on alignment-relevant features and enhances multi-modal feature fusion. Notably, our approach introduces a pioneering cross-modal association loss to foster modal consistency.Empirical evaluations across nine datasets confirm PMF’s superiority, demonstrating state-of-the-art performance and the rationale for freezing modalities. Our code is available at https://github.com/ninibymilk/PMF-MMEA.

Understanding sentiment is arguably an advanced and important capability of AI agents in the physical world. In previous works, many efforts have been devoted to individual sentiment subtasks, without considering interrelated sentiment knowledge among these subtasks. Although some recent works model multiple sentiment subtasks in a unified manner, they merely simply combine these subtasks without deeply exploring the hierarchical relationships among subtasks. In this paper, we introduce GSA-7B, an open-source large language model specific to the sentiment domain. Specifically, we deeply explore the hierarchical relationships between sentiment subtasks, proposing progressive sentiment reasoning benchmark and progressive task instructions. Subsequently, we use Llama2-7B as the backbone model and propose parameter-efficient progressive tuning paradigm which is implemented by constructing chain of LoRA, resulting in the creation of GSA-7B. Experimental results show that GSA-7B as a unified model performs well across all datasets in the progressive sentiment reasoning benchmark. Additionally, under the few-shot setting, GSA-7B also exhibits good generalization ability for sentiment subtasks and datasets that were not encountered during its training phase.

Although Large Language Models (LLMs) exhibit remarkable adaptability across domains, these models often fall short in structured knowledge extraction tasks such as named entity recognition (NER). This paper explores an innovative, cost-efficient strategy to harness LLMs with modest NER capabilities for producing superior NER datasets. Our approach diverges from the basic class-conditional prompts by instructing LLMs to self-reflect on the specific domain, thereby generating domain-relevant attributes (such as category and emotions for movie reviews), which are utilized for creating attribute-rich training data. Furthermore, we preemptively generate entity terms and then develop NER context data around these entities, effectively bypassing the LLMs’ challenges with complex structures. Our experiments across both general and niche domains reveal significant performance enhancements over conventional data generation methods while being more cost-effective than existing alternatives.

The proliferation of Conversational AI agents (CAAs) has emphasised the need to distinguish between human and machine-generated texts, with implications spanning digital forensics and cybersecurity. While prior research primarily focussed on distinguishing human from machine-generated text, our study takes a more refined approach by analysing different CAAs. We construct linguistic profiles for five CAAs, aiming to identify Uniquely Identifiable Linguistic Patterns (UILPs) for each model using authorship attribution techniques. Authorship attribution (AA) is the task of identifying the author of an unknown text from a pool of known authors. Our research seeks to answer crucial questions about the existence of UILPs in CAAs, the linguistic overlap between various text types generated by these models, and the feasibility of Authorship Attribution (AA) for CAAs based on UILPs. Promisingly, we are able to attribute CAAs based on their original texts with a weighted F1-score of 96.94%. Further, we are able to attribute CAAs according to their writing style (as specified by prompts), yielding a weighted F1-score of 95.84%, which sets the baseline for this task. By employing principal component analysis (PCA), we identify the top 100 most informative linguistic features for each CAA, achieving a weighted F1-score ranging from 86.04% to 97.93%, and an overall weighted F1-score of 93.86%.

Large language models (LLMs) are effective at answering questions that are clearly asked. However, when faced with ambiguous queries they can act unpredictably and produce incorrect outputs. This underscores the need for the development of intelligent agents capable of asking clarification questions to resolve ambiguities effectively. This capability requires complex understanding, state tracking, reasoning and planning over multiple conversational turns. However, directly measuring this can be challenging.In this paper, we offer a surrogate problem which assesses an LLMs’s capability to deduce an entity unknown to itself, but revealed to a judge, by asking the judge a series of queries. This entity-deducing game can serve as an evaluation framework to probe the conversational reasoning and planning capabilities of language models.We systematically evaluate various LLMs and discover significant differences in their performance on this task. We find that strong LLMs like GPT-4 outperform human players by a large margin. We further employ Behavior Cloning (BC) to examine whether a weaker model is capable of imitating a stronger model and generalizing to data or domains, using only the demonstrations from a stronger model. We finally propose to use Reinforcement Learning to enhance reasoning and planning capacity of Vicuna models through episodes of game playing, which lead to significant performance improvement. We hope that this problem offers insights into how autonomous agents could be trained to behave more intelligently in ambiguous circumstances.