Multilingual pre-trained Large Language Models (LLMs) are incredibly effective at Question Answering (QA), a core task in Natural Language Understanding, achieving high accuracies on several multilingual benchmarks. However, little is known about how well their confidences are calibrated. In this paper, we comprehensively benchmark the calibration of several multilingual LLMs (MLLMs) on a variety of QA tasks. We perform extensive experiments, spanning encoder-only, encoder-decoder, and decoder-only QA models (size varying from 110M to 7B parameters) and diverse languages, including both high- and low-resource ones. We study different dimensions of calibration in in-distribution, out-of-distribution, and cross-lingual transfer settings, and investigate strategies to improve it, including post-hoc methods and regularized fine-tuning. For decoder-only LLMs such as LlaMa2, we additionally find that in-context learning improves confidence calibration on multilingual data. We also conduct several ablation experiments to study the effect of language distances, language corpus size, and model size on calibration, and how multilingual models compare with their monolingual counterparts for diverse tasks and languages. Our experiments suggest that the multilingual QA models are poorly calibrated for languages other than English and incorporating a small set of cheaply translated multilingual samples during fine-tuning/calibration effectively enhances the calibration performance.
Generative Large Language Models (LLMs) have achieved remarkable advancements in various NLP tasks. In this work, our aim is to explore the multilingual capabilities of large language models by using machine translation as a task involving English and 22 Indian languages. We first investigate the translation capabilities of raw large language models, followed by exploring the in-context learning capabilities of the same raw models. We fine-tune these large language models using parameter efficient fine-tuning methods such as LoRA and additionally with full fine-tuning. Through our study, we have identified the best performing large language model for the translation task involving LLMs, which is based on LLaMA. Our results demonstrate significant progress, with average BLEU scores of 13.42, 15.93, 12.13, 12.30, and 12.07, as well as CHRF scores of 43.98, 46.99, 42.55, 42.42, and 45.39, respectively, using 2-stage fine-tuned LLaMA-13b for English to Indian languages on IN22 (conversational), IN22 (general), flores200-dev, flores200-devtest, and newstest2019 testsets. Similarly, for Indian languages to English, we achieved average BLEU scores of 14.03, 16.65, 16.17, 15.35 and 12.55 along with chrF scores of 36.71, 40.44, 40.26, 39.51, and 36.20, respectively, using fine-tuned LLaMA-13b on IN22 (conversational), IN22 (general), flores200-dev, flores200-devtest, and newstest2019 testsets. Overall, our findings highlight the potential and strength of large language models for machine translation capabilities, including for languages that are currently underrepresented in LLMs.
In-context learning (ICL) has become the default method for using large language models (LLMs), making the exploration of its limitations and understanding the underlying causes crucial. In this paper, we find that ICL falls short of handling specification-heavy tasks, which are tasks with complicated and extensive task specifications, requiring several hours for ordinary humans to master, such as traditional information extraction tasks. The performance of ICL on these tasks mostly cannot reach half of the state-of-the-art results. To explore the reasons behind this failure, we conduct comprehensive experiments on 18 specification-heavy tasks with various LLMs and identify three primary reasons: inability to specifically understand context, misalignment in task schema comprehension with humans, and inadequate long-text understanding ability. Furthermore, we demonstrate that through fine-tuning, LLMs can achieve decent performance on these tasks, indicating that the failure of ICL is not an inherent flaw of LLMs, but rather a drawback of existing alignment methods that renders LLMs incapable of handling complicated specification-heavy tasks via ICL. To substantiate this, we perform dedicated instruction tuning on LLMs for these tasks and observe a notable improvement. We hope the analyses in this paper could facilitate advancements in alignment methods enabling LLMs to meet more sophisticated human demands.
Existing Knowledge Base Question Answering (KBQA) architectures are hungry for annotated data, which make them costly and time-consuming to deploy. We introduce the problem of few-shot transfer learning for KBQA, where the target domain offers only a few labeled examples, but a large labeled training dataset is available in a source domain. We propose a novel KBQA architecture called FuSIC-KBQA that performs KB-retrieval using multiple source-trained retrievers, re-ranks using an LLM and uses this as input for LLM few-shot in-context learning to generate logical forms. These are further refined using execution-guided feedback. Experiments over multiple source-target KBQA pairs of varying complexity show that FuSIC-KBQA significantly outperforms adaptations of SoTA KBQA models for this setting. Additional experiments show that FuSIC-KBQA also outperforms SoTA KBQA models in the in-domain setting when training data is limited.
Language models (LMs) have achieved notable success in numerous NLP tasks, employing both fine-tuning and in-context learning (ICL) methods. While language models demonstrate exceptional performance, they face robustness challenges due to spurious correlations arising from imbalanced label distributions in training data or ICL exemplars. Previous research has primarily concentrated on word, phrase, and syntax features, neglecting the concept level, often due to the absence of concept labels and difficulty in identifying conceptual content in input texts. This paper introduces two main contributions. First, we employ ChatGPT to assign concept labels to texts, assessing concept bias in models during fine-tuning or ICL on test data. We find that LMs, when encountering spurious correlations between a concept and a label in training or prompts, resort to shortcuts for predictions. Second, we introduce a data rebalancing technique that incorporates ChatGPT-generated counterfactual data, thereby balancing label distribution and mitigating spurious correlations. Our method's efficacy, surpassing traditional token removal approaches, is validated through extensive testing.
Large Language Models (LLMs) have achieved remarkable success in natural language tasks, yet understanding their reasoning processes remains a significant challenge. We address this by introducing XplainLLM, a dataset accompanying an explanation framework designed to enhance LLM transparency and reliability. Our dataset comprises 24,204 instances where each instance interprets the LLM's reasoning behavior using knowledge graphs (KGs) and graph attention networks (GAT), and includes explanations of LLMs such as the decoder-only Llama-3 and the encoder-only RoBERTa. XplainLLM also features a framework for generating grounded explanations and the debugger-scores for multidimensional quality analysis. Our explanations include why-choose and why-not-choose components, reason-elements, and debugger-scores that collectively illuminate the LLM's reasoning behavior. Our evaluations demonstrate XplainLLM's potential to reduce hallucinations and improve grounded explanation generation in LLMs. XplainLLM is a resource for researchers and practitioners to build trust and verify the reliability of LLM outputs.
We propose the on-the-fly ensembling of a machine translation model with an LLM, prompted on the same task and input. We perform experiments on 4 language pairs (both directions) with varying data amounts. We find that a slightly weaker-at-translation LLM can improve translations of a NMT model, and ensembling with an LLM can produce better translations than ensembling two stronger MT models. We combine our method with various techniques from LLM prompting, such as in context learning and translation context.
Multilingual generative models obtain remarkable cross-lingual in-context learning capabilities through pre-training on large-scale corpora. However, they still exhibit a performance bias toward high-resource languages and learn isolated distributions of multilingual sentence representations, which may hinder knowledge transfer across languages. To bridge this gap, we propose a simple yet effective cross-lingual alignment framework exploiting pairs of translation sentences. It aligns the internal sentence representations across different languages via multilingual contrastive learning and aligns outputs by following cross-lingual instructions in the target language. Experimental results show that even with less than 0.1 {\textperthousand} of pre-training tokens, our alignment framework significantly boosts the cross-lingual abilities of generative language models and mitigates the performance gap. Further analyses reveal that it results in a better internal multilingual representation distribution of multilingual models.
In-context learning (ICL) is now a common method for teaching large language models (LLMs) new tasks: given labeled examples in the input context, the LLM learns to perform the task without weight updates. Do models guided via ICL infer the underlying structure of the task defined by the context, or do they rely on superficial heuristics that only generalize to identically distributed examples? We address this question using transformations tasks and an NLI task that assess sensitivity to syntax - a requirement for robust language understanding. We further investigate whether out-of-distribution generalization can be improved via chain-of-thought prompting, where the model is provided with a sequence of intermediate computation steps that illustrate how the task ought to be performed. In experiments with models from the GPT, PaLM, and Llama 2 families, we find large variance across LMs. The variance is explained more by the composition of the pre-training corpus and supervision methods than by model size; in particular, models pre-trained on code generalize better, and benefit more from chain-of-thought prompting.
Large language models (LLMs) such as GPT-3.5 and CodeLlama are powerful models for code generation and understanding. Fine-tuning these models comes with a high computational cost and requires a large labeled dataset. Alternatively, in-context learning techniques allow models to learn downstream tasks with only a few examples. Recently, researchers have shown how in-context learning performs well in bug detection and repair. In this paper, we propose code-pair classification task in which both the buggy and non-buggy versions are given to the model, and the model identifies the buggy ones. We evaluate our task in real-world dataset of bug detection and two most powerful LLMs. Our experiments indicate that an LLM can often pick the buggy from the non-buggy version of the code, and the code-pair classification task is much easier compared to be given a snippet and deciding if and where a bug exists.
In-context learning (ICL) has shown impressive results in few-shot learning tasks, yet its underlying mechanism is still not fully understood. A recent line of work suggests that ICL performs gradient descent (GD)-based optimization implicitly. While appealing, much of the research focuses on simplified settings, where the parameters of a shallow model are optimized. In this work, we revisit evidence for ICL-GD correspondence on realistic NLP tasks and models. We find gaps in evaluation, both in terms of problematic metrics and insufficient baselines. We show that surprisingly, even untrained models achieve comparable ICL-GD similarity scores despite not exhibiting ICL. Next, we explore a major discrepancy in the flow of information throughout the model between ICL and GD, which we term Layer Causality. We propose a simple GD-based optimization procedure that respects layer causality, and show it improves similarity scores significantly.
Recent advancements in natural language processing have led to the proliferation of large language models (LLMs). These models have been shown to yield good performance, using in-context learning, even on tasks and languages they are not trained on. However, their performance on African languages is largely understudied relative to high-resource languages. We present an analysis of four popular large language models (mT0, Aya, LLaMa 2, and GPT-4) on six tasks (topic classification, sentiment classification, machine translation, summarization, question answering, and named entity recognition) across 60 African languages, spanning different language families and geographical regions. Our results suggest that all LLMs produce lower performance for African languages, and there is a large gap in performance compared to high-resource languages (such as English) for most tasks. We find that GPT-4 has an average to good performance on classification tasks, yet its performance on generative tasks such as machine translation and summarization is significantly lacking. Surprisingly, we find that mT0 had the best overall performance for cross-lingual QA, better than the state-of-the-art supervised model (i.e. fine-tuned mT5) and GPT-4 on African languages. Similarly, we find the recent Aya model to have comparable result to mT0 in almost all tasks except for topic classification where it outperform mT0. Overall, LLaMa 2 showed the worst performance, which we believe is due to its English and code-centric~(around 98%) pre-training corpus. Our findings confirms that performance on African languages continues to remain a hurdle for the current LLMs, underscoring the need for additional efforts to close this gap.
We present BYOKG, a universal question-answering (QA) system that can operate on any knowledge graph (KG), requires no human-annotated training data, and can be ready to use within a day -- attributes that are out-of-scope for current KGQA systems. BYOKG draws inspiration from the remarkable ability of humans to comprehend information present in an unseen KG through exploration -- starting at random nodes, inspecting the labels of adjacent nodes and edges, and combining them with their prior world knowledge. In BYOKG, exploration leverages an LLM-backed symbolic agent that generates a diverse set of query-program exemplars, which are then used to ground a retrieval-augmented reasoning procedure to predict programs for arbitrary questions. BYOKG is effective over both small- and large-scale graphs, showing dramatic gains in QA accuracy over a zero-shot baseline of 27.89 and 58.02 F1 on GrailQA and MetaQA, respectively. On GrailQA, we further show that our unsupervised BYOKG outperforms a supervised in-context learning method, demonstrating the effectiveness of exploration. Lastly, we find that performance of BYOKG reliably improves with continued exploration as well as improvements in the base LLM, notably outperforming a state-of-the-art fine-tuned model by 7.08 F1 on a sub-sampled zero-shot split of GrailQA.
Are Large language models (LLMs) temporally grounded? Since LLMs cannot perceive and interact with the environment, it is impossible to answer this question directly. Instead, we provide LLMs with textual narratives and probe them with respect to their common-sense knowledge of the structure and duration of events, their ability to order events along a timeline, and self-consistency within their temporal model (e.g., temporal relations such as after and before are mutually exclusive for any pair of events). We evaluate state-of-the-art LLMs (such as LLaMA 2 and GPT-4) on three tasks reflecting these abilities. Generally, we find that LLMs lag significantly behind both human performance as well as small-scale, specialised LMs. In-context learning, instruction tuning, and chain-of-thought prompting reduce this gap only to a limited degree. Crucially, LLMs struggle the most with self-consistency, displaying incoherent behaviour in at least 27.23% of their predictions. Contrary to expectations, we also find that scaling the model size does not guarantee positive gains in performance. To explain these results, we study the sources from which LLMs may gather temporal information: we find that sentence ordering in unlabelled texts, available during pre-training, is only weakly correlated with event ordering. Moreover, public instruction tuning mixtures contain few temporal tasks. Hence, we conclude that current LLMs lack a consistent temporal model of textual narratives. Code, datasets, and LLM outputs are available at https://github.com/yfqiu-nlp/temporal-llms.
Transformer neural networks can exhibit a surprising capacity for in-context learning (ICL) despite not being explicitly trained for it. Prior work has provided a deeper understanding of how ICL emerges in transformers, e.g. through the lens of mechanistic interpretability, Bayesian inference, or by examining the distributional properties of training data. However, in each of these cases, ICL is treated largely as a persistent phenomenon; namely, once ICL emerges, it is assumed to persist asymptotically. Here, we show that the emergence of ICL during transformer training is, in fact, often transient. We train transformers on synthetic data designed so that both ICL and in-weights learning (IWL) strategies can lead to correct predictions. We find that ICL first emerges, then disappears and gives way to IWL, all while the training loss decreases, indicating an asymptotic preference for IWL. The transient nature of ICL is observed in transformers across a range of model sizes and datasets, raising the question of how much to "overtrain" transformers when seeking compact, cheaper-to-run models. We find that L2 regularization may offer a path to more persistent ICL that removes the need for early stopping based on ICL-style validation tasks. Finally, we present initial evidence that ICL transience may be caused by competition between ICL and IWL circuits.
Zero-shot In-context learning is the phenomenon where models can perform the task simply given the instructions. However, pre-trained large language models are known to be poorly calibrated for this task. One of the most effective approaches to handling this bias is to adopt a contrastive decoding objective, which accounts for the prior probability of generating the next token by conditioning on some context. This work introduces an Anti-Language Model objective with a decay factor designed to address the weaknesses of In-context Machine Translation. We conduct our experiments across 3 model types and sizes, 3 language directions, and for both greedy decoding and beam search ($B=5$). The proposed method outperforms other state-of-art decoding objectives, with up to $20$ BLEU point improvement from the default objective observed in some settings.
The remarkable ability of Large Language Models (LLMs) to understand and follow instructions has sometimes been limited by their in-context learning (ICL) performance in low-resource languages. To address this, we introduce a novel approach that leverages cross-lingual retrieval-augmented in-context learning (CREA-ICL). By extracting semantically similar prompts from high-resource languages, we aim to improve the zero-shot performance of multilingual pre-trained language models (MPLMs) across diverse tasks. Though our approach yields steady improvements in classification tasks, it faces challenges in generation tasks. Our evaluation offers insights into the performance dynamics of retrieval-augmented in-context learning across both classification and generation domains.
Recent studies demonstrated that large language models (LLMs) can excel in many tasks via in-context learning (ICL). However, recent works show that ICL-prompted models tend to produce inaccurate results when presented with adversarial inputs. In this work, we investigate whether augmenting ICL with natural language explanations (NLEs) improves the robustness of LLMs on adversarial datasets covering natural language inference and paraphrasing identification. We prompt LLMs with a small set of human-generated NLEs to produce further NLEs, yielding more accurate results than both a zero-shot-ICL setting and using only human-generated NLEs. Our results on five popular LLMs (GPT3.5-turbo, Llama2, Vicuna, Zephyr, and Mistral) show that our approach yields over 6% improvement over baseline approaches for eight adversarial datasets: HANS, ISCS, NaN, ST, PICD, PISP, ANLI, and PAWS. Furthermore, previous studies have demonstrated that prompt selection strategies significantly enhance ICL on in-distribution test sets. However, our findings reveal that these strategies do not match the efficacy of our approach for robustness evaluations, resulting in an accuracy drop of 8% compared to the proposed approach.
Recently, advancements in large language models (LLMs) have shown an unprecedented ability across various language tasks. This paper investigates the potential application of LLMs to slot filling with noisy ASR transcriptions, via both in-context learning and task-specific fine-tuning. Dedicated prompt designs and fine-tuning approaches are proposed to improve the robustness of LLMs for slot filling with noisy ASR transcriptions. Moreover, a linearised knowledge injection (LKI) scheme is also proposed to integrate dynamic external knowledge into LLMs. Experiments were performed on SLURP to quantify the performance of LLMs, including GPT-3.5-turbo, GPT-4, LLaMA-13B and Vicuna-13B (v1.1 and v1.5) with different ASR error rates. The use of the proposed fine-tuning together with the LKI scheme for LLaMA-13B achieved an 8.3% absolute SLU-F1 improvement compared to the strong Flan-T5-base baseline system on a limited data setup.
In-context learning (ICL) has become one of the most popular learning paradigms. While there is a growing body of literature focusing on prompt engineering, there is a lack of systematic analysis comparing the effects of prompts across different models and tasks. To address this gap, we present a comprehensive prompt analysis based on the sensitivity of a function. Our analysis reveals that sensitivity is an unsupervised proxy for model performance, as it exhibits a strong negative correlation with accuracy. We use gradient-based saliency scores to empirically demonstrate how different prompts affect the relevance of input tokens to the output, resulting in different levels of sensitivity. Furthermore, we introduce sensitivity-aware decoding which incorporates sensitivity estimation as a penalty term in the standard greedy decoding. We show that this approach is particularly helpful when information in the input is scarce. Our work provides a fresh perspective on the analysis of prompts, and contributes to a better understanding of the mechanism of ICL.