SequenceMatch: Imitation Learning for Autoregressive Sequence Modelling with Backtracking

• URL: http://arxiv.org/abs/2306.05426v3
• Date: Mon, 6 May 2024 16:02:30 GMT
• Title: SequenceMatch: Imitation Learning for Autoregressive Sequence Modelling with Backtracking
• Authors: Chris Cundy, Stefano Ermon,
• Abstract summary: A maximum-likelihood (MLE) objective does not match a downstream use-case of autoregressively generating high-quality sequences. We formulate sequence generation as an imitation learning (IL) problem. This allows us to minimize a variety of divergences between the distribution of sequences generated by an autoregressive model and sequences from a dataset. Our resulting method, SequenceMatch, can be implemented without adversarial training or architectural changes.
• Score: 60.109453252858806
• License: http://creativecommons.org/licenses/by-sa/4.0/
• Abstract: In many domains, autoregressive models can attain high likelihood on the task of predicting the next observation. However, this maximum-likelihood (MLE) objective does not necessarily match a downstream use-case of autoregressively generating high-quality sequences. The MLE objective weights sequences proportionally to their frequency under the data distribution, with no guidance for the model's behaviour out of distribution (OOD): leading to compounding error during autoregressive generation. In order to address this compounding error problem, we formulate sequence generation as an imitation learning (IL) problem. This allows us to minimize a variety of divergences between the distribution of sequences generated by an autoregressive model and sequences from a dataset, including divergences with weight on OOD generated sequences. The IL framework also allows us to incorporate backtracking by introducing a backspace action into the generation process. This further mitigates the compounding error problem by allowing the model to revert a sampled token if it takes the sequence OOD. Our resulting method, SequenceMatch, can be implemented without adversarial training or architectural changes. We identify the SequenceMatch-$\chi^2$ divergence as a more suitable training objective for autoregressive models which are used for generation. We show that empirically, SequenceMatch training leads to improvements over MLE on text generation with language models and arithmetic.

Related papers

• Promises and Pitfalls of Generative Masked Language Modeling: Theoretical Framework and Practical Guidelines [74.42485647685272]
  We focus on Generative Masked Language Models (GMLMs) We train a model to fit conditional probabilities of the data distribution via masking, which are subsequently used as inputs to a Markov Chain to draw samples from the model. We adapt the T5 model for iteratively-refined parallel decoding, achieving 2-3x speedup in machine translation with minimal sacrifice in quality.
  arXiv  Detail & Related papers  (2024-07-22T18:00:00Z)
• Non-autoregressive Sequence-to-Sequence Vision-Language Models [63.77614880533488]
  We propose a parallel decoding sequence-to-sequence vision-language model that marginalizes over multiple inference paths in the decoder. The model achieves performance on-par with its state-of-the-art autoregressive counterpart, but is faster at inference time.
  arXiv  Detail & Related papers  (2024-03-04T17:34:59Z)
• Extrapolative Controlled Sequence Generation via Iterative Refinement [22.42501277690634]
  We study the problem of extrapolative controlled generation, i.e., generating sequences with attribute values beyond the range seen in training. In this work, we propose Iterative Controlled Extrapolation (ICE) which iteratively makes local edits to a sequence to enable extrapolation. Results on one natural language task (sentiment analysis) and two protein engineering tasks (ACE2 stability and AAV fitness) show that ICE considerably outperforms state-of-the-art approaches despite its simplicity.
  arXiv  Detail & Related papers  (2023-03-08T13:21:27Z)
• Mutual Exclusivity Training and Primitive Augmentation to Induce Compositionality [84.94877848357896]
  Recent datasets expose the lack of the systematic generalization ability in standard sequence-to-sequence models. We analyze this behavior of seq2seq models and identify two contributing factors: a lack of mutual exclusivity bias and the tendency to memorize whole examples. We show substantial empirical improvements using standard sequence-to-sequence models on two widely-used compositionality datasets.
  arXiv  Detail & Related papers  (2022-11-28T17:36:41Z)
• Calibrating Sequence likelihood Improves Conditional Language Generation [39.35161650538767]
  Conditional language models are predominantly trained with maximum likelihood estimation (MLE) While MLE trained models assign high probability to plausible sequences given the context, the model probabilities often do not accurately rank-order generated sequences by quality. We introduce sequence likelihood calibration (SLiC) where the likelihood of model generated sequences are calibrated to better align with reference sequences in the model's latent space.
  arXiv  Detail & Related papers  (2022-09-30T19:16:16Z)
• Robust Attack Graph Generation [11.419463747286716]
  We present a method to learn automaton models that are more robust to input modifications. It iteratively aligns sequences to a learned model, modifies the sequences to their aligned versions, and re-learns the model.
  arXiv  Detail & Related papers  (2022-06-15T19:26:39Z)
• An EM Approach to Non-autoregressive Conditional Sequence Generation [49.11858479436565]
  Autoregressive (AR) models have been the dominating approach to conditional sequence generation. Non-autoregressive (NAR) models have been recently proposed to reduce the latency by generating all output tokens in parallel. This paper proposes a new approach that jointly optimize both AR and NAR models in a unified Expectation-Maximization framework.
  arXiv  Detail & Related papers  (2020-06-29T20:58:57Z)
• On the Discrepancy between Density Estimation and Sequence Generation [92.70116082182076]
  log-likelihood is highly correlated with BLEU when we consider models within the same family. We observe no correlation between rankings of models across different families.
  arXiv  Detail & Related papers  (2020-02-17T20:13:35Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.