Concept reasoning is an important capability for models to understand the world. However, the existing datasets, such as concept extraction and concept generation, suffer from modeledge leakage and context leakage. To address these limitations, we construct a dataset of concept reasoning for large language models (CR-LLM) with modeledge leakage prevention and context leakage prevention, which consists of 2,167 samples and covers different concept types. In addition, we propose a hybrid reasoning method, consisting of inductive reasoning, deductive reasoning and a controller. This method allows large language models to adaptively select the optimal reasoning method for each input sample. Finally, we conduct extensive experiments on CR-LLM using different models and methods. The results show that existing large language models and reasoning methods perform sub-optimally in the concept reasoning task. In contrast, our proposed method significantly improves the capabilities, achieving a 7% increase in accuracy compared to CoT and demonstrating better granularity. We release CR-LLM and code at https://github.com/Nianqi-Li/Concept-Reasoning-for-LLMs.
论文检索
输入标题、作者或关键词，从 1,915 篇学术成果中精准定位
ROSE Doesn’t Do That: Boosting the Safety of Instruction-Tuned Large Language Models with Reverse Prompt Contrastive Decoding
PDF ↗With the development of instruction-tuned large language models (LLMs), improving the safety of LLMs has become more critical. However, the current approaches for aligning the LLMs output with expected safety usually require substantial training efforts, e.g., high-quality safety data and expensive computational resources, which are costly and inefficient. To this end, we present reverse prompt contrastive decoding (ROSE), a simple-yet-effective method to directly boost the safety of existing instruction-tuned LLMs without any additional training. The principle of ROSE is to improve the probability of desired safe output via suppressing the undesired output induced by the carefully-designed reverse prompts. Experiments on 6 safety and 2 general-purpose tasks show that, our ROSE not only brings consistent and significant safety improvements (up to +13.8% safety score) upon 5 types of instruction-tuned LLMs, but also benefits the general-purpose ability of LLMs. In-depth analyses explore the underlying mechanism of ROSE, and reveal when and where to use it.
Since most large language models (LLMs) are trained once and never updated, they struggle to dynamically adapt to our ever-changing world. In this work, we present FreshQA, a dynamic QA benchmark that tests a model’s ability to answer questions that may require reasoning over up-to-date world knowledge. We develop a two-mode human evaluation procedure to measure both correctness and hallucination, which we use to benchmark both closed and open-source LLMs by collecting >50K human judgments. We observe that all LLMs struggle to answer questions that require fast-changing world knowledge as well as questions with false premises that need to be debunked. In response, we develop FreshPrompt, a few-shot prompting method that curates and organizes relevant information from a search engine into an LLM’s prompt. Our experiments show that FreshPrompt outperforms both competing search engine-augmented prompting methods such as Self-Ask (Press et al., 2022) as well as commercial systems such as Perplexity.AI. To facilitate future work, we additionally develop FreshEval, a reliable autorater for quick evaluation and comparison on FreshQA. Our latest results with FreshEval suggest that open-source LLMs such as Mixtral (Jiang et al., 2024), when combined with FreshPrompt, are competitive with closed-source and commercial systems on search-augmented QA.
A Unified Joint Approach with Topological Context Learning and Rule Augmentation for Knowledge Graph Completion
PDF ↗Knowledge graph completion (KGC) task is to infer the missing knowledge in the knowledge graph based on known factual triples. However, present KGC approaches still face the following two challenges. Those methods perform simple linear update on relation representation, and only local neighborhood information is aggregated, which makes it difficult to capture logic semantic between relations and global topological context information. To tackle the above challenges, we propose a unified joint approach with Topological Context learning and Rule Augmentation (TCRA) for KGC. The TCRA framework consists of an entity topological context learning mechanism based on dual-branch hierarchical graph attention network, and a relation rule context learning mechanism based on Rule-Transformer and rule-to-relation aggregator. The former mechanism encodes the topological structure features of entities, aggregates the local neighborhood topological context information of entities on the three levels (entity, relation and triple), and build clusters of global head or tail entities related to the same relation. It can capture the local and global topological context information of entities related to the same relation. The latter mechanism introduces chain-like Horn rules as the context information of relations, and encodes the logical semantic of relations to enrich the relation representation. Experimental performances on three benchmark datasets FB15k-237, WN18RR and Kinship indicate the effectiveness and superiority of our proposed approach. The codes are publicly available.
Large language models (LLMs) have presented remarkable capabilities in the wide range of natural language understanding and reasoning tasks. Despite their success, a few works indicate that LLMs suffer from the “reversal curse”, in which LLMs can’t employ the inverted structure “B is A” when they are trained based on “A is B”. To explore the effect of the “reversal curse” for LLMs on complex mathematical reasoning tasks, we present two reversal datasets upon GSM8K and MathQA and verify that LLMs also struggle to solve reversal mathematical problems. We analyze the potential reason and attribute it to the insufficient modeling of the relationship between reasoning steps caused by the left-to-right objective. Consequently, based on the characteristics of multi-step reasoning, we design a novel training method to improve the general and reversal reasoning abilities. Finally, we conduct experiments on four mathematical datasets, and the results demonstrate that our method significantly improves the general reasoning capacities and alleviates the reversal problem. Our datasets and codes are available at https: //github.com/AllForward/ReversalMath.
Domain adaptation remains a challenge in the realm of Neural Machine Translation (NMT), even in the era of large language models (LLMs). Existing non-parametric approaches like nearest neighbor machine translation have made small Autoregressive Translation (AT) models achieve efficient domain generalization and adaptation without updating parameters, but leaving the Non-Autoregressive Translation (NAT) counterparts under-explored. To fill this blank, we introduce Bi-kNN, an innovative and efficient domain adaptation approach for NAT models that tailors a k-nearest-neighbor algorithm for NAT. Specifically, we introduce an effective datastore construction and correlated updating strategies to conform the parallel nature of NAT. Additionally, we train a meta-network that seamlessly integrates the NN distribution with the NMT distribution robustly during the iterative decoding process of NAT. Our experimental results across four benchmark datasets demonstrate that our Bi-kNN not only achieves significant improvements over the Base-NAT model (7.8 BLEU on average) but also exhibits enhanced efficiency.
The programming skill is one crucial ability for Large Language Models (LLMs), necessitating a deep understanding of programming languages (PLs) and their correlation with natural languages (NLs). We examine the impact of pre-training data on code-focused LLMs’ performance by assessing the comment density as a measure of PL-NL alignment. Given the scarcity of code-comment aligned data in pre-training corpora, we introduce a novel data augmentation method that generates comments for existing code, coupled with a data filtering strategy that filters out code data poorly correlated with natural language. We conducted experiments on three code-focused LLMs and observed consistent improvements in performance on two widely-used programming skill benchmarks. Notably, the model trained on the augmented data outperformed both the model used for generating comments and the model further trained on the data without augmentation.
Advancing automated programming necessitates robust and comprehensive code generation benchmarks, yet current evaluation frameworks largely neglect object-oriented programming (OOP) in favour of functional programming (FP), e.g., HumanEval and MBPP. To address this, our study introduces a pioneering OOP-focused benchmark, featuring 431 Python programs that encompass essential OOP concepts and features like classes and encapsulation methods. We propose a novel evaluation metric, pass@o, tailored for OOP, enhancing traditional pass@k metric. Our evaluation of 23 leading large language models (LLMs), including both general and code-specialized models, reveals three key insights: 1) pass@o offers a more relevant and comprehensive assessment for OOP code generation; 2) Despite excelling in FP, code-specialized LLMs like WizardCoder lag in OOP compared to models like ChatGPT; 3) The poor performance of all advanced LLMs on our OOP benchmark highlights a critical need for improvements in this field. Our benchmark and scripts will be publicly released at GitHub.
Connecting text and visual modalities plays an essential role in generative intelligence. For this reason, inspired by the success of large language models, significant research efforts are being devoted to the development of Multimodal Large Language Models (MLLMs). These models can seamlessly integrate visual and textual modalities, while providing a dialogue-based interface and instruction-following capabilities. In this paper, we provide a comprehensive review of recent visual-based MLLMs, analyzing their architectural choices, multimodal alignment strategies, and training techniques. We also conduct a detailed analysis of these models across a wide range of tasks, including visual grounding, image generation and editing, visual understanding, and domain-specific applications. Additionally, we compile and describe training datasets and evaluation benchmarks, conducting comparisons among existing models in terms of performance and computational requirements. Overall, this survey offers a comprehensive overview of the current state of the art, laying the groundwork for future MLLMs.
Recent advancements in Chinese Spelling Correction (CSC) predominantly leverage pre-trained language models (PLMs). However, a notable challenge with fine-tuned PLM-based CSC models is their tendency to over-correct, leading to poor generalization for error patterns outside the standard distribution. To address this, we developed a teacher network guided by prior knowledge for distillation learning of CSC models. Unlike traditional teacher networks, which depend on task-related pre-training, our method infuses task-related prior information into the teacher network, offering guidance beyond mere labels to the student network. This strategy significantly enhances the CSC model’s language modeling capabilities, crucial for minimizing over-correction. Importantly, our approach is model-independent and the teacher network does not require task-related pre-training, making it broadly applicable for enhancing various PLM-based CSC models with minimal additional computational resources. Extensive experiments on widely used benchmarks demonstrate that our method achieves new state-of-the-art results. Additionally, we explored the potential of generalizing our method to other non-autoregressive text-generation tasks.
Recent studies have shown that fusing text labels and context sentences is an effective method for learning prototype representations in few-shot relation extraction. However, the **inconsistency of prototype representations** across different few-shot tasks persists due to different context sentences for the same relation, even with the integration of text labels into prototype representations. Conversely, the text label for each relation is unique and consistent, 1)which prompts us to propose a **dual prototype learning method**. Unlike previous methods that only construct support-based prototypes, we additionally construct label-based prototypes. Furthermore, we introduce a graph-based prototype adjustment module to construct topological information between support-based and label-based prototypes, thereby generating a more effective similarity measure through a simple linear combination. In addition, relations of different granularities have different distribution widths in the same semantic space, the **imbalanced distribution in the semantic space** leads to a lack of comparability among relations. To create a more discriminative semantic space, 2)we propose a **granularity-aware prototype learning method** that unifies the distribution width of relations, making relations of different granularities have similar distribution widths. Experimental results on two public benchmark datasets show that our proposed methods achieve state-of-the-art performance in few-shot relation classification.
Hypothetical induction is recognized as the main reasoning type when scientists make observations about the world and try to propose hypotheses to explain those observations. Past research on hypothetical induction is under a constrained setting: (1) the observation annotations in the dataset are carefully manually handpicked sentences (resulting in a close-domain setting); and (2) the ground truth hypotheses are mostly commonsense knowledge, making the task less challenging. In this work, we tackle these problems by proposing the first dataset for social science academic hypotheses discovery, with the final goal to create systems that automatically generate valid, novel, and helpful scientific hypotheses, given only a pile of raw web corpus. Unlike previous settings, the new dataset requires (1) using open-domain data (raw web corpus) as observations; and (2) proposing hypotheses even new to humanity. A multi-module framework is developed for the task, including three different feedback mechanisms to boost performance, which exhibits superior performance in terms of both GPT-4 based and expert-based evaluation.To the best of our knowledge, this is the first work showing that LLMs are able to generate novel (”not existing in literature”) and valid (”reflecting reality”) scientific hypotheses.
Large language models (LLMs) have demonstrated impressive reasoning capabilities, yet there is ongoing debate about these abilities and the potential data contamination problem recently. This paper aims to evaluate the reasoning capacities of LLMs, specifically in solving recent competition-level programming problems in Codeforces, which are expert-crafted and unique, requiring deep understanding and robust reasoning skills. We first provide a comprehensive evaluation of GPT-4’s perceived zero-shot performance on this task, considering various aspects such as problems’ release time, difficulties, and types of errors encountered. Surprisingly, the perceived performance of GPT-4 has experienced a cliff like decline in problems after September 2021 consistently across all the difficulties and types of problems, which shows the potential data contamination, as well as the challenges for any existing LLM to solve unseen complex reasoning problems. We further explore various approaches such as fine-tuning, Chain-of-Thought prompting and problem description simplification. Unfortunately, none of them is able to consistently mitigate the challenges. Through our work, we emphasize the importance of this excellent data source for assessing the genuine reasoning capabilities of LLMs, and foster the development of LLMs with stronger reasoning abilities and better generalization in the future.
Agents powered by large language models (LLMs) inherit important limitations, such as the restricted context length, dependency on human-engineered exemplars (e.g., for task decomposition), and insufficient generalization. To address these challenges, we propose RaDA, a novel planning method for Web agents that does not require manual exemplars, efficiently leverages the LLMs’ context, and enhances generalization. RaDA disentangles planning into two stages: for a new given task, during Retrieval-augmented Task Decomposition (RaD), it decomposes tasks into high-level subtasks; next, during Retrieval-augmented Action Generation (RaA), it traverses the trajectory obtained with RaD to iteratively synthesize actions based on dynamically retrieved exemplars. We compare RaDA with strong baselines covering a broad space of design choices, using both GPT-3.5 and GPT-4 as backbones; and we find consistent improvements over previous SOTA in two challenging benchmarks, CompWoB and Mind2Web, covering settings with different complexities. We show the contributions of RaDA via ablation studies and qualitative analysis; and we discuss the structural benefits of our more compositional design.
InstructGraph: Boosting Large Language Models via Graph-centric Instruction Tuning and Preference Alignment
PDF ↗Do current large language models (LLMs) better solve graph reasoning and generation tasks with parameter updates? In this paper, we propose InstructGraph, a framework that empowers LLMs with the abilities of graph reasoning and generation by instruction tuning and preference alignment. Specifically, we first propose a structured format verbalizer to unify all graph data into a universal code-like format, which can simply represent the graph without any external graph-specific encoders. Furthermore, a graph instruction tuning stage is introduced to guide LLMs in solving graph reasoning and generation tasks. Finally, we identify potential hallucination problems in graph tasks and sample negative instances for preference alignment, the target of which is to enhance the output’s reliability of the model. Extensive experiments across multiple graph-centric tasks exhibit that InstructGraph can achieve the best performance and outperform GPT-4 and LLaMA2 by more than 13% and 38%, respectively.
A Curious Case of Searching for the Correlation between Training Data and Adversarial Robustness of Transformer Textual Models
PDF ↗Existing works have shown that fine-tuned textual transformer models achieve state-of-the-art prediction performances but are also vulnerable to adversarial text perturbations. Traditional adversarial evaluation is often done only after fine-tuning the models and ignoring the training data. In this paper, we want to prove that there is also a strong correlation between training data and model robustness. To this end, we extract 13 different features representing a wide range of input fine-tuning corpora properties and use them to predict the adversarial robustness of the fine-tuned models. Focusing mostly on encoder-only transformer models BERT and RoBERTa with additional results for BART, ELECTRA and GPT2, we provide diverse evidence to support our argument. First, empirical analyses show that (a) extracted features can be used with a lightweight classifier such as Random Forest to effectively predict the attack success rate and (b) features with the most influence on the model robustness have a clear correlation with the robustness. Second, our framework can be used as a fast and effective additional tool for robustness evaluation since it (a) saves 30x-193x runtime compared to the traditional technique, (b) is transferable across models, (c) can be used under adversarial training, and (d) robust to statistical randomness. Our code is publicly available at https://github.com/CaptainCuong/RobustText_ACL2024.
This paper explores to construct a general text evaluator based on open-source Large Language Models (LLMs), a domain predominantly occupied by commercial counterparts such as GPT-4. Recognizing the limitations of open-source models like Llama in evaluative tasks, we introduce InstructEval, a general multi-aspect text evaluator developed through instruction tuning of open-source LLMs. To overcome the shortage of annotated resources for multi-aspect evaluations, InstructEval combines extensive open Human Preference Modeling (HPM) datasets with a small set of multi-aspect annotated data.This approach not only enhances effectiveness in overall evaluation tasks but also exhibits improved performance in multi-aspect evaluation tasks.As demonstrated by our extensive experiments, InstructEval achieves comparable or superior performance to commercial LLMs like ChatGPT or GPT-4 in terms of both overall and multi-aspect evaluation.
We propose Referral-Augmented Retrieval (RAR), a simple technique that concatenates document indices with referrals: text from other documents that cite or link to the given document. We find that RAR provides significant performance gains for tasks across paper retrieval, entity retrieval, and open-domain question-answering in both zero-shot and in-domain (e.g., fine-tuned) settings. We examine how RAR provides especially strong improvements on more structured tasks, and can greatly outperform generative text expansion techniques such as DocT5Query and Query2Doc, with a 37% and 21% absolute improvement on ACL paper retrieval, respectively. We also compare three ways to aggregate referrals for RAR. Overall, we believe RAR can help revive and re-contextualize the classic information retrieval idea of using anchor texts to improve the representations of documents in a wide variety of corpuses in the age of neural retrieval.
Despite the success of data augmentation in improving CLIP model, existing methods that utilize LLM or SAM to enrich the information in captions still suffer from several limitations, including insufficient detail and excessive hallucinations, ultimately resulting in compromised alignment and masking the true potential of dense information. This can lead to erroneous conclusions about CLIP’s ability to handle rich data, impeding the development of more effective models. To address the limitations of existing methods, we introduce a novel pipeline that generates highly detailed, factually accurate captions for images, which facilitates in-depth analysis of the potential for dense information in multimodal alignment. Contrary to previous findings, our investigation revealed that lengthening captions boosts performance across diverse benchmarks, even surpassing the effectiveness of meticulously crafted hard negative samples. Building on these insights, DELIP is introduced, demonstrably enhancing both foundational multimodal alignment and compositional reasoning abilities. Finally, we explore strategies to expand the context window of the text encoder, unlocking the potential of richer data for CLIP and paving the way for advancements in leveraging dense information for multimodal alignment.
Reinforcement Tuning for Detecting Stances and Debunking Rumors Jointly with Large Language Models
PDF ↗Learning multi-task models for jointly detecting stance and verifying rumors poses challenges due to the need for training data of stance at post level and rumor veracity at claim level, which are difficult to obtain. To address this issue, we leverage large language models (LLMs) as the foundation annotators for the joint stance detection (SD) and rumor verification (RV) tasks, dubbed as JSDRV. We introduce a novel reinforcement tuning framework to enhance the joint predictive capabilities of LLM-based SD and RV components. Specifically, we devise a policy for selecting LLM-annotated data at the two levels, employing a hybrid reward mechanism to choose high-quality labels for effective LLM fine-tuning on both tasks. Results demonstrate that JSDRV improves the capabilities of LLMs in the joint tasks, not only outperforming state-of-the-art methods but also generalizing to non-LLMs accommodated as task models.