• Corpus ID: 253097695

Towards Teachable Reasoning Systems: Using a Dynamic Memory of User Feedback for Continual System Improvement

@inproceedings{Dalvi2022TowardsTR,
  title={Towards Teachable Reasoning Systems: Using a Dynamic Memory of User Feedback for Continual System Improvement},
  author={Bhavana Dalvi and Oyvind Tafjord and Peter Clark},
  year={2022}
}
Our goal is a teachable reasoning system for question-answering (QA), where a user can interact with faithful answer explanations, and correct its errors so that the system improves over time. Our approach is to augment a QA model with a dynamic memory of user feedback, containing user-supplied corrections to erroneous model beliefs that users identify during interaction. Retrievals from memory are used as additional context for QA, to help avoid previous mistakes in similar new situa-tions - a… 

Figures from this paper

Entailer: Answering Questions with Faithful and Truthful Chains of Reasoning

The approach is to recursively combine a trained backward-chaining model, capable of generating a set of premises entailing an answer hypothesis, with a verifier that checks that the model itself believes those premises (and the entailment itself) through self-querying.

References

SHOWING 1-10 OF 31 REFERENCES

Learning to repair: Repairing model output errors after deployment using a dynamic memory of feedback

It is shown that the memory-enhancedFBN ET system, FBN ET, learns to apply user feedback effectively to repair such errors, while making a start at avoiding similar past mistakes on new, unseen examples.

MemPrompt: Memory-assisted Prompt Editing with User Feedback

It is shown how a (simulated) user can interac-tively teach a deployed GPT -3, substantially increasing its accuracy over the queries with different kinds of misunderstandings by the GPT 3, a step towards the low-cost utility enhancement for very large pre-trained LMs.

BeliefBank: Adding Memory to a Pre-Trained Language Model for a Systematic Notion of Belief

This work describes two mechanisms to improve belief consistency in the overall system, enabling PTLM-based architectures with a systematic notion of belief to construct a more coherent picture of the world, and improve over time without model retraining.

Entailer: Answering Questions with Faithful and Truthful Chains of Reasoning

The approach is to recursively combine a trained backward-chaining model, capable of generating a set of premises entailing an answer hypothesis, with a verifier that checks that the model itself believes those premises (and the entailment itself) through self-querying.

Learning To Retrieve Prompts for In-Context Learning

This work proposes an efficient method for retrieving prompts for in-context learning using annotated data and an LM, and trains an efficient dense retriever from this data, which is used to retrieve training examples as prompts at test time.

Explaining Answers with Entailment Trees

ENTAILMENTBANK is created, the first dataset to contain multistep entailment trees, providing a new type of dataset (multistep entails) and baselines, offering a new avenue for the community to generate richer, more systematic explanations.

Editing Factual Knowledge in Language Models

This work presents KnowledgeEditor, a method which can be used to edit factual knowledge and, thus, fix ‘bugs’ or unexpected predictions without the need for expensive re-training or fine-tuning.

Interactive Task Learning from GUI-Grounded Natural Language Instructions and Demonstrations

We show SUGILITE, an intelligent task automation agent that can learn new tasks and relevant associated concepts interactively from the user’s natural language instructions and demonstrations, using

BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage

The goal of this research program is to enable the community to study ever-improving responsible agents that learn through interaction in BlenderBot 3, a 175B parameter dialogue model capable of open-domain conversation with access to the internet and a long-term memory.

Learning to Attend On Essential Terms: An Enhanced Retriever-Reader Model for Open-domain Question Answering

This paper proposes a retriever-reader model that learns to attend on essential terms during the question answering process, and builds an essential term selector which first identifies the most important words in a question, then reformulates the query and searches for related evidence.