arXiv subjects

Eric Malmi

Publications and source records attributed to Eric Malmi.

At least 19 recordsLinked to original sources

arXiv2024

Mastering Board Games by External and Internal Planning with Language Models

Advancing planning and reasoning capabilities of Large Language Models (LLMs) is one of the key prerequisites towards unlocking their potential for performing reliably in complex and impactful domains. In this paper, we aim to demonstrate this across board games (Chess, Fischer Random / Chess960, Connect Four, and Hex), and we show that search-based planning can yield significant improvements in LLM game-playing strength. We introduce, compare and contrast two major approaches: In external search, the model guides Monte Carlo Tree Search (MCTS) rollouts and evaluations without calls to an external game engine, and in internal search, the model is trained to generate in-context a linearized tree of search and a resulting final choice. Both build on a language model pre-trained on relevant domain knowledge, reliably capturing the transition and value functions in the respective environments, with minimal hallucinations. We evaluate our LLM search implementations against game-specific state-of-the-art engines, showcasing substantial improvements in strength over the base model, and reaching Grandmaster-level performance in chess while operating closer to the human search budget. Our proposed approach, combining search with domain knowledge, is not specific to board games, hinting at more general future applications.

cs.AI↗

arXiv2024

West-of-N: Synthetic Preferences for Self-Improving Reward Models

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.

cs.CL↗

arXiv2024

Effects of Research Paper Promotion via ArXiv and X

In the evolving landscape of scientific publishing, it is important to understand the drivers of high-impact research, to equip scientists with actionable strategies to enhance the reach of their work, and to understand trends in the use of modern scientific publishing tools to inform their further development. Here, we study trends in the use of early preprint publications and revisions on ArXiv and the use of X (formerly Twitter) for promotion of such papers in computer science and physics. We find that early submissions to ArXiv and promotion on X have soared in recent years. Estimating the effect that the use of each of these modern affordances has on the number of citations of scientific publications, we find that peer-reviewed conference papers in computer science that are submitted early to ArXiv gain on average $21.1 \pm 17.4$ more citations, revised on ArXiv gain $18.4 \pm 17.6$ more citations, and promoted on X gain $44.4 \pm 8$ more citations in the first 5 years from an initial publication. In contrast, journal articles in physics experience comparatively lower boosts in citation counts, with increases of $3.9 \pm 1.1$, $4.3 \pm 0.9$, and $6.9 \pm 3.5$ citations respectively for the same interventions. Our results show that promoting one's work on ArXiv or X has a large impact on the number of citations, as well as the number of influential citations computed by Semantic Scholar, and thereby on the career of researchers. These effects are present also for publications in physics, but they are relatively smaller. The larger relative effect sizes, effects of promotion accumulating over time, and elevated unpredictability of the number of citations in computer science than in physics suggest a greater role of world-of-mouth spreading in computer science than in physics.

cs.DL↗

arXiv2023

Small Language Models Improve Giants by Rewriting Their Outputs

Despite the impressive performance of large language models (LLMs), they often lag behind specialized models in various tasks. LLMs only use a fraction of the existing training data for in-context learning, while task-specific models harness the full dataset for fine-tuning. In this work, we tackle the problem of leveraging training data to improve the performance of LLMs without fine-tuning. Our approach directly targets LLM predictions without requiring access to their weights. We create a pool of candidates from the LLM through few-shot prompting and we employ a compact model, the LM-corrector (LMCor), specifically trained to merge these candidates to produce an enhanced output. Our experiments on four natural language generation tasks demonstrate that even a small LMCor model (250M) substantially improves the few-shot performance of LLMs (62B), matching and even outperforming standard fine-tuning. Furthermore, we illustrate the robustness of LMCor against different prompts, thereby minimizing the need for extensive prompt engineering. Finally, we show that LMCor can be seamlessly integrated with different LLMs at inference, serving as a plug-and-play module to improve their performance.

cs.CL↗

arXiv2022

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.

cs.CL↗

arXiv2022

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.

cs.CL↗

arXiv2022

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

We present EdiT5 - a novel semi-autoregressive text-editing model designed to combine the strengths of non-autoregressive text-editing and autoregressive decoding. EdiT5 is faster during inference than conventional sequence-to-sequence (seq2seq) models, while being capable of modelling 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 step uses an autoregressive decoder. Depending on the task, EdiT5 on average requires significantly fewer autoregressive steps, demonstrating speedups of up to 25x when compared to seq2seq models. Quality-wise, EdiT5 is initialized with a pre-trained T5 checkpoint yielding comparable performance to T5 in high-resource settings when evaluated on three NLG tasks: Sentence Fusion, Grammatical Error Correction, and Decontextualization while clearly outperforming T5 in low-resource settings.

cs.CL↗

arXiv2021

Controlled Text Generation as Continuous Optimization with Multiple Constraints

As large-scale language model pretraining pushes the state-of-the-art in text generation, recent work has turned to controlling attributes of the text such models generate. While modifying the pretrained models via fine-tuning remains the popular approach, it incurs a significant computational cost and can be infeasible due to lack of appropriate data. As an alternative, we propose MuCoCO -- a flexible and modular algorithm for controllable inference from pretrained models. We formulate the decoding process as an optimization problem which allows for multiple attributes we aim to control to be easily incorporated as differentiable constraints to the optimization. By relaxing this discrete optimization to a continuous one, we make use of Lagrangian multipliers and gradient-descent based techniques to generate the desired text. We evaluate our approach on controllable machine translation and style transfer with multiple sentence-level attributes and observe significant improvements over baselines.

cs.CL↗

arXiv2021

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.

cs.CL↗

arXiv2020

Semantically Driven Sentence Fusion: Modeling and Evaluation

Sentence fusion is the task of joining related sentences into coherent text. Current training and evaluation schemes for this task are based on single reference ground-truths and do not account for valid fusion variants. We show that this hinders models from robustly capturing the semantic relationship between input sentences. To alleviate this, we present an approach in which ground-truth solutions are automatically expanded into multiple references via curated equivalence classes of connective phrases. We apply this method to a large-scale dataset and use the augmented dataset for both model training and evaluation. To improve the learning of semantic representation using multiple references, we enrich the model with auxiliary discourse classification tasks under a multi-tasking framework. Our experiments highlight the improvements of our approach over state-of-the-art models.

cs.CL↗

arXiv2020

Unsupervised Text Style Transfer with Padded Masked Language Models

We propose Masker, an unsupervised text-editing method for style transfer. To tackle cases when no parallel source-target pairs are available, we train masked language models (MLMs) for both the source and the target domain. Then we find the text spans where the two models disagree the most in terms of likelihood. This allows us to identify the source tokens to delete to transform the source text to match the style of the target domain. The deleted tokens are replaced with the target MLM, and by using a padded MLM variant, we avoid having to predetermine the number of inserted tokens. Our experiments on sentence fusion and sentiment transfer demonstrate that Masker performs competitively in a fully unsupervised setting. Moreover, in low-resource settings, it improves supervised methods' accuracy by over 10 percentage points when pre-training them on silver training data generated by Masker.

cs.CL↗

arXiv2020

Rapformer: Conditional Rap Lyrics Generation with Denoising Autoencoders

The ability to combine symbols to generate language is a defining characteristic of human intelligence, particularly in the context of artistic story-telling through lyrics. We develop a method for synthesizing a rap verse based on the content of any text (e.g., a news article), or for augmenting pre-existing rap lyrics. Our method, called Rapformer, is based on training a Transformer-based denoising autoencoder to reconstruct rap lyrics from content words extracted from the lyrics, trying to preserve the essential meaning, while matching the target style. Rapformer features a novel BERT-based paraphrasing scheme for rhyme enhancement which increases the average rhyme density of output lyrics by 10%. Experimental results on three diverse input domains show that Rapformer is capable of generating technically fluent verses that offer a good trade-off between content preservation and style transfer. Furthermore, a Turing-test-like experiment reveals that Rapformer fools human lyrics experts 25% of the time.

cs.CL↗

arXiv2020

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.

cs.CL↗

arXiv2019

Encode, Tag, Realize: High-Precision Text Editing

We propose LaserTagger - a sequence tagging approach that casts text generation as a text editing task. Target texts are reconstructed from the inputs using three main edit operations: keeping a token, deleting it, and adding a phrase before the token. To predict the edit operations, we propose a novel model, which combines a BERT encoder with an autoregressive Transformer decoder. This approach is evaluated on English text on four tasks: sentence fusion, sentence splitting, abstractive summarization, and grammar correction. LaserTagger achieves new state-of-the-art results on three of these tasks, performs comparably to a set of strong seq2seq baselines with a large number of training examples, and outperforms them when the number of examples is limited. Furthermore, we show that at inference time tagging can be more than two orders of magnitude faster than comparable seq2seq models, making it more attractive for running in a live environment.

cs.CL↗

arXiv2019

DiscoFuse: A Large-Scale Dataset for Discourse-Based Sentence Fusion

Sentence fusion is the task of joining several independent sentences into a single coherent text. Current datasets for sentence fusion are small and insufficient for training modern neural models. In this paper, we propose a method for automatically-generating fusion examples from raw text and present DiscoFuse, a large scale dataset for discourse-based sentence fusion. We author a set of rules for identifying a diverse set of discourse phenomena in raw text, and decomposing the text into two independent sentences. We apply our approach on two document collections: Wikipedia and Sports articles, yielding 60 million fusion examples annotated with discourse information required to reconstruct the fused text. We develop a sequence-to-sequence model on DiscoFuse and thoroughly analyze its strengths and weaknesses with respect to the various discourse phenomena, using both automatic as well as human evaluation. Finally, we conduct transfer learning experiments with WebSplit, a recent dataset for text simplification. We show that pretraining on DiscoFuse substantially improves performance on WebSplit when viewed as a sentence fusion task.

cs.CL↗

arXiv2018

Responsible team players wanted: an analysis of soft skill requirements in job advertisements

During the past decades the importance of soft skills for labour market outcomes has grown substantially. This carries implications for labour market inequality, since previous research shows that soft skills are not valued equally across race and gender. This work explores the role of soft skills in job advertisements by drawing on methods from computational science as well as on theoretical and empirical insights from economics, sociology and psychology. We present a semi-automatic approach based on crowdsourcing and text mining for extracting a list of soft skills. We find that soft skills are a crucial component of job ads, especially of low-paid jobs and jobs in female-dominated professions. Our work shows that soft skills can serve as partial predictors of the gender composition in job categories and that not all soft skills receive equal wage returns at the labour market. Especially "female" skills are frequently associated with wage penalties. Our results expand the growing literature on the association of soft skills on wage inequality and highlight their importance for occupational gender segregation at labour markets.

cs.CY↗

arXiv2018

Learning Representations for Soft Skill Matching

Employers actively look for talents having not only specific hard skills but also various soft skills. To analyze the soft skill demands on the job market, it is important to be able to detect soft skill phrases from job advertisements automatically. However, a naive matching of soft skill phrases can lead to false positive matches when a soft skill phrase, such as friendly, is used to describe a company, a team, or another entity, rather than a desired candidate. In this paper, we propose a phrase-matching-based approach which differentiates between soft skill phrases referring to a candidate vs. something else. The disambiguation is formulated as a binary text classification problem where the prediction is made for the potential soft skill based on the context where it occurs. To inform the model about the soft skill for which the prediction is made, we develop several approaches, including soft skill masking and soft skill tagging. We compare several neural network based approaches, including CNN, LSTM and Hierarchical Attention Model. The proposed tagging-based input representation using LSTM achieved the highest recall of 83.92% on the job dataset when fixing a precision to 95%.

cs.CL↗

arXiv2018

Computationally Inferred Genealogical Networks Uncover Long-Term Trends in Assortative Mating

Genealogical networks, also known as family trees or population pedigrees, are commonly studied by genealogists wanting to know about their ancestry, but they also provide a valuable resource for disciplines such as digital demography, genetics, and computational social science. These networks are typically constructed by hand through a very time-consuming process, which requires comparing large numbers of historical records manually. We develop computational methods for automatically inferring large-scale genealogical networks. A comparison with human-constructed networks attests to the accuracy of the proposed methods. To demonstrate the applicability of the inferred large-scale genealogical networks, we present a longitudinal analysis on the mating patterns observed in a network. This analysis shows a consistent tendency of people choosing a spouse with a similar socioeconomic status, a phenomenon known as assortative mating. Interestingly, we do not observe this tendency to consistently decrease (nor increase) over our study period of 150 years.

cs.SI↗