Offline Regularised Reinforcement Learning for Large Language Models Alignment

The dominant framework for alignment of large language models (LLM), whether through reinforcement learning from human feedback or direct preference optimisation, is to learn from preference data. This involves building datasets where each element is a quadruplet composed of a prompt, two independent responses (completions of the prompt) and a human preference between the two independent responses, yielding a preferred and a dis-preferred response. Such data is typically scarce and expensive to collect. On the other hand, \emph{single-trajectory} datasets where each element is a triplet composed of a prompt, a response and a human feedback is naturally more abundant. The canonical element of such datasets is for instance an LLM's response to a user's prompt followed by a user's feedback such as a thumbs-up/down. Consequently, in this work, we propose DRO, or \emph{Direct Reward Optimisation}, as a framework and associated algorithms that do not require pairwise preferences. DRO uses a simple mean-squared objective that can be implemented in various ways. We validate our findings empirically, using T5 encoder-decoder language models, and show DRO's performance over selected baselines such as Kahneman-Tversky Optimization (KTO). Thus, we confirm that DRO is a simple and empirically compelling method for single-trajectory policy optimisation.

West-of-N: Synthetic Preference Generation for Improved Reward Modeling

The success of reinforcement learning from human feedback (RLHF) in language model alignment is strongly dependent on the quality of the underlying reward model. In this paper, we present a novel approach to improve reward model quality by generating synthetic preference data, thereby augmenting the training dataset with on-policy, high-quality preference pairs. Motivated by the promising results of Best-of-N sampling strategies in language model training, we extend their application to reward model training. This results in a self-training strategy to generate preference pairs by selecting the best and worst candidates in a pool of responses to a given query. Empirically, we find that this approach improves the performance of any reward model, with an effect comparable to the addition of a similar quantity of human preference data. This work opens up new avenues of research for improving RLHF for language model alignment, by offering synthetic preference generation as a solution to reward modeling challenges.

Small Language Models Improve Giants by Rewriting Their Outputs

Large language models (LLMs) have demonstrated impressive few-shot learning capabilities, but they often underperform compared to fine-tuned models on challenging tasks. Furthermore, their large size and restricted access only through APIs make task-specific fine-tuning impractical. Moreover, LLMs are sensitive to different aspects of prompts (e.g., the selection and order of demonstrations) and can thus require time-consuming prompt engineering. In this light, we propose a method to correct LLM outputs without relying on their weights. First, we generate a pool of candidates by few-shot prompting an LLM. Second, we refine the LLM-generated outputs using a smaller model, the LM-corrector (LMCor), which is trained to rank, combine and rewrite the candidates to produce the final target output. Our experiments demonstrate that even a small LMCor model (250M) substantially improves the few-shot performance of LLMs (62B) across diverse tasks. Moreover, we illustrate that the LMCor exhibits robustness against different prompts, thereby minimizing the need for extensive prompt engineering. Finally, we showcase that the LMCor can be seamlessly integrated with different LLMs at inference time, serving as a plug-and-play module to improve their performance.

Teaching Small Language Models to Reason

Chain of thought prompting successfully improves the reasoning capabilities of large language models, achieving state of the art results on a range of datasets. However, these reasoning capabilities only appear to emerge in models with a size of over 100 billion parameters. In this paper, we explore the transfer of such reasoning capabilities to models with less than 100 billion parameters via knowledge distillation. Specifically, we finetune a student model on the chain of thought outputs generated by a larger teacher model. Our experiments show that the proposed method improves task performance across arithmetic, commonsense and symbolic reasoning datasets. For example, the accuracy of T5 XXL on GSM8K improves from 8.11% to 21.99% when finetuned on PaLM-540B generated chains of thought.

Text Generation with Text-Editing Models

Text-editing models have recently become a prominent alternative to seq2seq models for monolingual text-generation tasks such as grammatical error correction, simplification, and style transfer. These tasks share a common trait - they exhibit a large amount of textual overlap between the source and target texts. Text-editing models take advantage of this observation and learn to generate the output by predicting edit operations applied to the source sequence. In contrast, seq2seq models generate outputs word-by-word from scratch thus making them slow at inference time. Text-editing models provide several benefits over seq2seq models including faster inference speed, higher sample efficiency, and better control and interpretability of the outputs. This tutorial provides a comprehensive overview of text-editing models and current state-of-the-art approaches, and analyzes their pros and cons. We discuss challenges related to productionization and how these models can be used to mitigate hallucination and bias, both pressing challenges in the field of text generation.

EdiT5: Semi-Autoregressive Text-Editing with T5 Warm-Start

We present EdiT5 - a novel semi-autoregressive text-editing approach designed to combine the strengths of non-autoregressive text-editing and autoregressive decoding. EdiT5 is faster at inference times than conventional sequence-to-sequence (seq2seq) models, while being capable of modeling flexible input-output transformations. This is achieved by decomposing the generation process into three sub-tasks: (1) tagging to decide on the subset of input tokens to be preserved in the output, (2) re-ordering to define their order in the output text, and (3) insertion to infill the missing tokens that are not present in the input. The tagging and re-ordering steps, which are responsible for generating the largest portion of the output, are non-autoregressive, while the insertion uses an autoregressive decoder. Depending on the task, EdiT5 requires significantly fewer autoregressive steps demonstrating speedups of up to 25x when compared to classic seq2seq models. Quality-wise, EdiT5 is initialized with a pre-trained T5 checkpoint yielding comparable performance to T5 in high-resource settings and clearly outperforms it on low-resource settings when evaluated on three NLG tasks: Sentence Fusion, Grammatical Error Correction, and Decontextualization.

RED-ACE: Robust Error Detection for ASR using Confidence Embeddings

ASR Error Detection (AED) models aim to post-process the output of Automatic Speech Recognition (ASR) systems, in order to detect transcription errors. Modern approaches usually use text-based input, comprised solely of the ASR transcription hypothesis, disregarding additional signals from the ASR model. Instead, we propose to utilize the ASR system's word-level confidence scores for improving AED performance. Specifically, we add an ASR Confidence Embedding (ACE) layer to the AED model's encoder, allowing us to jointly encode the confidence scores and the transcribed text into a contextualized representation. Our experiments show the benefits of ASR confidence scores for AED, their complementary effect over the textual signal, as well as the effectiveness and robustness of ACE for combining these signals. To foster further research, we publish a novel AED dataset consisting of ASR outputs on the LibriSpeech corpus with annotated transcription errors.

A Simple Recipe for Multilingual Grammatical Error Correction

This paper presents a simple recipe to train state-of-the-art multilingual Grammatical Error Correction (GEC) models. We achieve this by first proposing a language-agnostic method to generate a large number of synthetic examples. The second ingredient is to use large-scale multilingual language models (up to 11B parameters). Once fine-tuned on language-specific supervised sets we surpass the previous state-of-the-art results on GEC benchmarks in four languages: English, Czech, German and Russian. Having established a new set of baselines for GEC, we make our results easily reproducible and accessible by releasing a cLang-8 dataset. It is produced by using our best model, which we call gT5, to clean the targets of a widely used yet noisy lang-8 dataset. cLang-8 greatly simplifies typical GEC training pipelines composed of multiple fine-tuning stages -- we demonstrate that performing a single fine-tuning step on cLang-8 with the off-the-shelf language models yields further accuracy improvements over an already top-performing gT5 model for English.

Felix: Flexible Text Editing Through Tagging and Insertion

We present Felix --- a flexible text-editing approach for generation, designed to derive the maximum benefit from the ideas of decoding with bi-directional contexts and self-supervised pre-training. In contrast to conventional sequence-to-sequence (seq2seq) models, Felix is efficient in low-resource settings and fast at inference time, while being capable of modeling flexible input-output transformations. We achieve this by decomposing the text-editing task into two sub-tasks: tagging to decide on the subset of input tokens and their order in the output text and insertion to in-fill the missing tokens in the output not present in the input. The tagging model employs a novel Pointer mechanism, while the insertion model is based on a Masked Language Model. Both of these models are chosen to be non-autoregressive to guarantee faster inference. Felix performs favourably when compared to recent text-editing methods and strong seq2seq baselines when evaluated on four NLG tasks: Sentence Fusion, Machine Translation Automatic Post-Editing, Summarization, and Text Simplification.

Controllable Sentence Simplification: Employing Syntactic and Lexical Constraints

Sentence simplification aims to make sentences easier to read and understand. Recent approaches have shown promising results with sequence-to-sequence models which have been developed assuming homogeneous target audiences. In this paper we argue that different users have different simplification needs (e.g. dyslexics vs. non-native speakers), and propose CROSS, ContROllable Sentence Simplification model, which allows to control both the level of simplicity and the type of the simplification. We achieve this by enriching a Transformer-based architecture with syntactic and lexical constraints (which can be set or learned from data). Empirical results on two benchmark datasets show that constraints are key to successful simplification, offering flexible generation output.