# Leveraging Explicit Procedural Instructions for Data-Efficient Action Prediction Source: [Leveraging Explicit Procedural Instructions for Data-Efficient Action Prediction]() Collection index: [llms.txt]() Data updated: 2026-01-04T04:32:38.389Z Authors: Julia White, Arushi Raghuvanshi, Yada Pruksachatkun Published: 2023-06-06T18:42:08.000Z ## Abstract Task-oriented dialogues often require agents to enact complex, multi-step procedures in order to meet user requests. While large language models have found success automating these dialogues in constrained environments, their widespread deployment is limited by the substantial quantities of task-specific data required for training. The following paper presents a data-efficient solution to constructing dialogue systems, leveraging explicit instructions derived from agent guidelines, such as company policies or customer service manuals. Our proposed Knowledge-Augmented Dialogue System (KADS) combines a large language model with a knowledge retrieval module that pulls documents outlining relevant procedures from a predefined set of policies, given a user-agent interaction. To train this system, we introduce a semi-supervised pre-training scheme that employs dialogue-document matching and action-oriented masked language modeling with partial parameter freezing. We evaluate the effectiveness of our approach on prominent task-oriented dialogue datasets, Action-Based Conversations Dataset and Schema-Guided Dialogue, for two dialogue tasks: action state tracking and workflow discovery. Our results demonstrate that procedural knowledge augmentation improves accuracy predicting in- and out-of-distribution actions while preserving high performance in settings with low or sparse data. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Action State Tracking on ABCD in-distribution (test)](): Action State Tracking; ABCD in-distribution (test); B-Slot Acc: 85.2. - [Action State Tracking on SGD (in-distribution, test)](): Action State Tracking; SGD in-distribution (test); B-Slot Accuracy: 63.2. - [Action Selection Task (AST) on ABCD (out-of-distribution)](): Action Selection Task (AST); ABCD (out-of-distribution); B-Slot Accuracy: 11.6. - [Action Selection Task (AST) on SGD (out-of-distribution)](): Action Selection Task (AST); SGD (out-of-distribution); B-Slot Acc: 49.8. - [Workflow Discovery on SGD (in-distribution, test)](): Workflow Discovery; SGD in-distribution (test); B-Slot Accuracy: 53.1. ## Original sources and implementation [Original paper]() [PDF]()