Toward Robust RALMs: Revealing the Impact of Imperfect Retrieval on Retrieval-Augmented Language Models

Retrieval Augmented Language Models (RALMs) have gained significant attention for their ability to generate accurate answer and improve efficiency. However, RALMs are inherently vulnerable to imperfect information due to their reliance on the imperfect retriever or knowledge source. We identify three common scenarios-unanswerable, adversarial, conflicting-where retrieved document sets can confuse RALM with plausible real-world examples. We present the first comprehensive investigation to assess how well RALMs detect and handle such problematic scenarios. Among these scenarios, to systematically examine adversarial robustness we propose a new adversarial attack method, Generative model-based ADVersarial attack (GenADV) and a novel metric Robustness under Additional Document (RAD). Our findings reveal that RALMs often fail to identify the unanswerability or contradiction of a document set, which frequently leads to hallucinations. Moreover, we show the addition of an adversary significantly degrades RALM's performance, with the model becoming even more vulnerable when the two scenarios overlap (adversarial+unanswerable). Our research identifies critical areas for assessing and enhancing the robustness of RALMs, laying the foundation for the development of more robust models.

IntGrad MT: Eliciting LLMs' Machine Translation Capabilities with Sentence Interpolation and Gradual MT

Recent Large Language Models (LLMs) have demonstrated strong performance in translation without needing to be finetuned on additional parallel corpora. However, they still underperform for low-resource language pairs. Previous works have focused on mitigating this issue by leveraging relevant few-shot examples or external resources such as dictionaries or grammar books, making models heavily reliant on these nonparametric sources of information. In this paper, we propose a novel method named IntGrad MT that focuses on fully exploiting an LLM's inherent translation capability. IntGrad MT achieves this by constructing a chain of few-shot examples, each consisting of a source sentence and the model's own translation, that rise incrementally in difficulty. IntGrad MT employs two techniques: Sentence Interpolation, which generates a sequence of sentences that gradually change from an easy sentence to translate to a difficult one, and Gradual MT, which sequentially translates this chain using translations of earlier sentences as few-shot examples for the translation of subsequent ones. With this approach, we observe a substantial enhancement in the xCOMET scores of various LLMs for multiple languages, especially in low-resource languages such as Hindi(8.26), Swahili(7.10), Bengali(6.97) and Marathi(13.03). Our approach presents a practical way of enhancing LLMs' performance without extra training.

Towards Efficient Visual-Language Alignment of the Q-Former for Visual Reasoning Tasks

Recent advancements in large language models have demonstrated enhanced capabilities in visual reasoning tasks by employing additional encoders for aligning different modalities. While the Q-Former has been widely used as a general encoder for aligning several modalities including image, video, audio, and 3D with large language models, previous works on its efficient training and the analysis of its individual components have been limited. In this work, we investigate the effectiveness of parameter efficient fine-tuning (PEFT) the Q-Former using InstructBLIP with visual reasoning benchmarks ScienceQA and IconQA. We observe that applying PEFT to the Q-Former achieves comparable performance to full fine-tuning using under 2% of the trainable parameters. Additionally, we employ AdaLoRA for dynamic parameter budget reallocation to examine the relative importance of the Q-Former's sublayers with 4 different benchmarks. Our findings reveal that the self-attention layers are noticeably more important in perceptual visual-language reasoning tasks, and relative importance of FFN layers depends on the complexity of visual-language patterns involved in tasks. The code is available at https://github.com/AttentionX/InstructBLIP_PEFT.

Enhancing Robustness of Retrieval-Augmented Language Models with In-Context Learning

Retrieval-Augmented Language Models (RALMs) have significantly improved performance in open-domain question answering (QA) by leveraging external knowledge. However, RALMs still struggle with unanswerable queries, where the retrieved contexts do not contain the correct answer, and with conflicting information, where different sources provide contradictory answers due to imperfect retrieval. This study introduces an in-context learning-based approach to enhance the reasoning capabilities of RALMs, making them more robust in imperfect retrieval scenarios. Our method incorporates Machine Reading Comprehension (MRC) demonstrations, referred to as cases, to boost the model's capabilities to identify unanswerabilities and conflicts among the retrieved contexts. Experiments on two open-domain QA datasets show that our approach increases accuracy in identifying unanswerable and conflicting scenarios without requiring additional fine-tuning. This work demonstrates that in-context learning can effectively enhance the robustness of RALMs in open-domain QA tasks.

Locate&Edit: Energy-based Text Editing for Efficient, Flexible, and Faithful Controlled Text Generation

Recent approaches to controlled text generation (CTG) often involve manipulating the weights or logits of base language models (LMs) at decoding time. However, these methods are inapplicable to latest black-box LMs and ineffective at preserving the core semantics of the base LM's original generations. In this work, we propose Locate&Edit(L&E), an efficient and flexible energy-based approach to CTG, which edits text outputs from a base LM using off-the-shelf energy models. Given text outputs from the base LM, L&E first locates spans that are most relevant to constraints (e.g., toxicity) utilizing energy models, and then edits these spans by replacing them with more suitable alternatives. Importantly, our method is compatible with black-box LMs, as it requires only the text outputs. Also, since L&E doesn't mandate specific architecture for its component models, it can work with a diverse combination of available off-the-shelf models. Moreover, L&E preserves the base LM's original generations, by selectively modifying constraint-related aspects of the texts and leaving others unchanged. These targeted edits also ensure that L&E operates efficiently. Our experiments confirm that L&E achieves superior semantic preservation of the base LM generations and speed, while simultaneously obtaining competitive or improved constraint satisfaction. Furthermore, we analyze how the granularity of energy distribution impacts CTG performance and find that fine-grained, regression-based energy models improve constraint satisfaction, compared to conventional binary classifier energy models.

RE-RAG: Improving Open-Domain QA Performance and Interpretability with Relevance Estimator in Retrieval-Augmented Generation

Retrieval-augmented generation (RAG) frame work is showing state-of-the-art performance on open-domain question answering tasks by referencing external knowledge. However, the RAG system faces challenges with performance degradation when it is fed contexts of low relevance or when the relative relevance among the input contexts is inaccurately assessed. In this work, we propose a RE-RAG framework that injects an explicit context relevance estimator (RE) into the RAG system. RE-RAG re-evaluates the retrieved contexts with the proposed context RE and passes the more relevant contexts along with their measure importance to the generator. To train context RE, we propose an unsupervised learning method, which does not utilize any labeled document ranking data to train the context RE. To examine the efficacy of RE-RAG, we examine its performance on Natural Questions and TriviaQA datasets. RE-RAG achieves on-par performance compared to the FiD variants while utilizing fewer contexts (0.25x). We show that the proposed context RE, which was trained with the T5 model, is also applicable to RAG with LLMs(ChatGPT) by improving the performance on NQ (+6.4EM) and TQA (+2.8EM), respecitvely. Lastly, we display that RE can add interpretability to RAG framework as RE score highly correlates with the RE-RAG accuracy. Consequently, RE can be utilized to filter out unanswerable scenarios where context does not contain answers with 38.9%-51.3% accuracy just by examining a set of retrieved contexts.

An Analysis under a Unified Fomulation of Learning Algorithms with Output Constraints

Neural networks (NN) perform well in diverse tasks, but sometimes produce nonsensical results to humans. Most NN models "solely" learn from (input, output) pairs, occasionally conflicting with human knowledge. Many studies indicate injecting human knowledge by reducing output constraints during training can improve model performance and reduce constraint violations. While there have been several attempts to compare different existing algorithms under the same programming framework, nonetheless, there has been no previous work that categorizes learning algorithms with output constraints in a unified manner. Our contributions are as follows: (1) We categorize the previous studies based on three axes: type of constraint loss used (e.g. probabilistic soft logic, REINFORCE), exploration strategy of constraint-violating examples, and integration mechanism of learning signals from main task and constraint. (2) We propose new algorithms to integrate the information of main task and constraint injection, inspired by continual-learning algorithms. (3) Furthermore, we propose the $H\beta$-score as a metric for considering the main task metric and constraint violation simultaneously. To provide a thorough analysis, we examine all the algorithms on three NLP tasks: natural language inference (NLI), synthetic transduction examples (STE), and semantic role labeling (SRL). We explore and reveal the key factors of various algorithms associated with achieving high $H\beta$-scores.

Comparing Neighbors Together Makes it Easy: Jointly Comparing Multiple Candidates for Efficient and Effective Retrieval

A common retrieve-and-rerank paradigm involves retrieving a broad set of relevant candidates using a scalable bi-encoder, followed by expensive but more accurate cross-encoders to a limited candidate set. However, this small subset often leads to error propagation from the bi-encoders, thereby restricting the performance of the overall pipeline. To address these issues, we propose the Comparing Multiple Candidates (CMC) framework, which compares a query and multiple candidate embeddings jointly through shallow self-attention layers. While providing contextualized representations, CMC is scalable enough to handle multiple comparisons simultaneously, where comparing 2K candidates takes only twice as long as comparing 100. Practitioners can use CMC as a lightweight and effective reranker to improve top-1 accuracy. Moreover, when integrated with another retriever, CMC reranking can function as a virtually enhanced retriever. This configuration adds only negligible latency compared to using a single retriever (virtual), while significantly improving recall at K (enhanced).} Through experiments, we demonstrate that CMC, as a virtually enhanced retriever, significantly improves Recall@k (+6.7, +3.5%-p for R@16, R@64) compared to the initial retrieval stage on the ZeSHEL dataset. Meanwhile, we conduct experiments for direct reranking on entity, passage, and dialogue ranking. The results indicate that CMC is not only faster (11x) than cross-encoders but also often more effective, with improved prediction performance in Wikipedia entity linking (+0.7%-p) and DSTC7 dialogue ranking (+3.3%-p). The code and link to datasets are available at https://github.com/yc-song/cmc

Case-Based Reasoning Approach for Solving Financial Question Answering

Measuring a machine's understanding of human language often involves assessing its reasoning skills, i.e. logical process of deriving answers to questions. While recent language models have shown remarkable proficiency in text based tasks, their efficacy in complex reasoning problems involving heterogeneous information such as text, tables, and numbers remain uncertain. Addressing this gap, FinQA introduced a numerical reasoning dataset for financial documents and simultaneously proposed a program generation approach . Our investigation reveals that half of the errors (48%) stem from incorrect operations being generated. To address this issue, we propose a novel approach to tackle numerical reasoning problems using case based reasoning (CBR), an artificial intelligence paradigm that provides problem solving guidance by offering similar cases (i.e. similar questions and corresponding logical programs). Our model retrieves relevant cases to address a given question, and then generates an answer based on the retrieved cases and contextual information. Through experiments on the FinQA dataset, we demonstrate competitive performance of our approach and additionally show that by expanding case repository, we can help solving complex multi step programs which FinQA showed weakness of.

Multistage Collaborative Knowledge Distillation from Large Language Models

We study semi-supervised sequence prediction tasks where labeled data are too scarce to effectively finetune a model and at the same time few-shot prompting of a large language model (LLM) has suboptimal performance. This happens when a task, such as parsing, is expensive to annotate and also unfamiliar to a pretrained LLM. In this paper, we present a discovery that student models distilled from a prompted LLM can often generalize better than their teacher on such tasks. Leveraging this finding, we propose a new distillation method, multistage collaborative knowledge distillation from an LLM (MCKD), for such tasks. MCKD first prompts an LLM using few-shot in-context learning to produce pseudolabels for unlabeled data. Then, at each stage of distillation, a pair of students are trained on disjoint partitions of the pseudolabeled data. Each student subsequently produces new and improved pseudolabels for the unseen partition to supervise the next round of student(s) with. We show the benefit of multistage cross-partition labeling on two constituency parsing tasks. On CRAFT biomedical parsing, 3-stage MCKD with 50 labeled examples matches the performance of supervised finetuning with 500 examples and outperforms the prompted LLM and vanilla KD by 7.5% and 3.7% parsing F1, respectively.