Autoregressive Reasoning over Chains of Facts with Transformers
By Ruben Cartuyvels, Graham Spinks, Marie-Francine Moens. 2020
15 citations recorded by the source corpus.
This paper proposes an iterative inference algorithm for multi-hop explanation regeneration, that retrieves relevant factual evidence in the form of text snippets, given a natural language question and its answer. Combining multiple sources of evidence or facts for multi-hop reasoning becomes increasingly hard when the number of sources needed to make an inference grows. Our algorithm copes with this by decomposing the selection of facts from a corpus autoregressively, conditioning the next iteration on previously selected facts. This allows us to use a pairwise learning-to-rank loss. We validate our method on datasets of the TextGraphs 2019 and 2020 Shared Tasks for explanation regeneration. Existing work on this task either evaluates facts in isolation or artificially limits the possible chains of facts, thus limiting multi-hop inference. We demonstrate that our algorithm, when used with a pre-trained transformer model, outperforms the previous state-of-the-art in terms of precision, training time and inference efficiency.
Subfields: Artificial Intelligence, Molecular Biology, Natural Language Processing, Prompting & Reasoning
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Swarat Chaudhuri
- Kaj Bostrom
- Mokanarangan Thayaparan
- André Freitas
- Greg Durrett
- Xinyu Zhao
- Marco Valentino
- Alexander M. Rush
In Molecular Biology
In Natural Language Processing
- Swarat Chaudhuri
- Kaj Bostrom
- Mokanarangan Thayaparan
- André Freitas
- Greg Durrett
- Xinyu Zhao
- Marco Valentino
- Daniel Klein
In Prompting & Reasoning
- Mokanarangan Thayaparan
- André Freitas
- Kaj Bostrom
- Greg Durrett
- Xinyu Zhao
- Marco Valentino
- Swarat Chaudhuri
Citations in the corpus
- Flexible Generation of Natural Language Deductions (2021, background): “, 2020) and new types of models take inspiration from formal and informal reasoning (Clark et al., 2020; Saha et al., 2020; Cartuyvels et al., 2020; Betz et al., 2021).”
- Case-Based Abductive Natural Language Inference (2020, background): “These resources have been used in explanation regeneration tasks (Jansen and Ustalov, 2019; Cartuyvels et al., 2020; Valentino et al., 2021), but their applicability on downstream predictions is yet to be explored.”