Can LLMs Reliably Simulate Human Learner Actions? A Simulation Authoring Framework for Open-Ended Learning Environments

Simulating learner actions helps stress-test open-ended interactive learning environments and prototype new adaptations before deployment. While recent studies show the promise of using large language models (LLMs) for simulating human behavior, such approaches have not gone beyond rudimentary proof-of-concept stages due to key limitations. First, LLMs are highly sensitive to minor prompt variations, raising doubts about their ability to generalize to new scenarios without extensive prompt engineering. Moreover, apparently successful outcomes can often be unreliable, either because domain experts unintentionally guide LLMs to produce expected results, leading to self-fulfilling prophecies; or because the LLM has encountered highly similar scenarios in its training data, meaning that models may not be simulating behavior so much as regurgitating memorized content. To address these challenges, we propose Hyp-Mix, a simulation authoring framework that allows experts to develop and evaluate simulations by combining testable hypotheses about learner behavior. Testing this framework in a physics learning environment, we found that GPT-4 Turbo maintains calibrated behavior even as the underlying learner model changes, providing the first evidence that LLMs can be used to simulate realistic behaviors in open-ended interactive learning environments, a necessary prerequisite for useful LLM behavioral simulation.

Making Task-Oriented Dialogue Datasets More Natural by Synthetically Generating Indirect User Requests

Indirect User Requests (IURs), such as "It's cold in here" instead of "Could you please increase the temperature?" are common in human-human task-oriented dialogue and require world knowledge and pragmatic reasoning from the listener. While large language models (LLMs) can handle these requests effectively, smaller models deployed on virtual assistants often struggle due to resource constraints. Moreover, existing task-oriented dialogue benchmarks lack sufficient examples of complex discourse phenomena such as indirectness. To address this, we propose a set of linguistic criteria along with an LLM-based pipeline for generating realistic IURs to test natural language understanding (NLU) and dialogue state tracking (DST) models before deployment in a new domain. We also release IndirectRequests, a dataset of IURs based on the Schema Guided Dialog (SGD) corpus, as a comparative testbed for evaluating the performance of smaller models in handling indirect requests.

IndirectRequests: Making Task-Oriented Dialogue Datasets More Natural by Synthetically Generating Indirect User Requests

Existing benchmark corpora of task-oriented dialogue are collected either using a "machines talking to machines" approach or by giving template-based goal descriptions to crowdworkers. These methods, however, often produce utterances that are markedly different from natural human conversations in which people often convey their preferences in indirect ways, such as through small talk. We term such utterances as Indirect User Requests (IURs). Understanding such utterances demands considerable world knowledge and reasoning capabilities on the listener's part. Our study introduces an LLM-based pipeline to automatically generate realistic, high-quality IURs for a given domain, with the ultimate goal of supporting research in natural language understanding (NLU) and dialogue state tracking (DST) for task-oriented dialogue systems. Our findings show that while large LLMs such as GPT-3.5 and GPT-4 generate high-quality IURs, achieving similar quality with smaller models is more challenging. We release IndirectRequests, a dataset of IURs that advances beyond the initial Schema-Guided Dialog (SGD) dataset in that it provides a challenging testbed for testing the "in the wild" performance of NLU and DST models.

Towards Compositionally Generalizable Semantic Parsing in Large Language Models: A Survey

Compositional generalization is the ability of a model to generalize to complex, previously unseen types of combinations of entities from just having seen the primitives. This type of generalization is particularly relevant to the semantic parsing community for applications such as task-oriented dialogue, text-to-SQL parsing, and information retrieval, as they can harbor infinite complexity. Despite the success of large language models (LLMs) in a wide range of NLP tasks, unlocking perfect compositional generalization still remains one of the few last unsolved frontiers. The past few years has seen a surge of interest in works that explore the limitations of, methods to improve, and evaluation metrics for compositional generalization capabilities of LLMs for semantic parsing tasks. In this work, we present a literature survey geared at synthesizing recent advances in analysis, methods, and evaluation schemes to offer a starting point for both practitioners and researchers in this area.

Can Similarity-Based Domain-Ordering Reduce Catastrophic Forgetting for Intent Recognition?

Task-oriented dialogue systems are expected to handle a constantly expanding set of intents and domains even after they have been deployed to support more and more functionalities. To live up to this expectation, it becomes critical to mitigate the catastrophic forgetting problem (CF) that occurs in continual learning (CL) settings for a task such as intent recognition. While existing dialogue systems research has explored replay-based and regularization-based methods to this end, the effect of domain ordering on the CL performance of intent recognition models remains unexplored. If understood well, domain ordering has the potential to be an orthogonal technique that can be leveraged alongside existing techniques such as experience replay. Our work fills this gap by comparing the impact of three domain-ordering strategies (min-sum path, max-sum path, random) on the CL performance of a generative intent recognition model. Our findings reveal that the min-sum path strategy outperforms the others in reducing catastrophic forgetting when training on the 220M T5-Base model. However, this advantage diminishes with the larger 770M T5-Large model. These results underscores the potential of domain ordering as a complementary strategy for mitigating catastrophic forgetting in continually learning intent recognition models, particularly in resource-constrained scenarios.

Towards a Neural Era in Dialogue Management for Collaboration: A Literature Survey

Dialogue-based human-AI collaboration can revolutionize collaborative problem-solving, creative exploration, and social support. To realize this goal, the development of automated agents proficient in skills such as negotiating, following instructions, establishing common ground, and progressing shared tasks is essential. This survey begins by reviewing the evolution of dialogue management paradigms in collaborative dialogue systems, from traditional handcrafted and information-state based methods to AI planning-inspired approaches. It then shifts focus to contemporary data-driven dialogue management techniques, which seek to transfer deep learning successes from form-filling and open-domain settings to collaborative contexts. The paper proceeds to analyze a selected set of recent works that apply neural approaches to collaborative dialogue management, spotlighting prevailing trends in the field. This survey hopes to provide foundational background for future advancements in collaborative dialogue management, particularly as the dialogue systems community continues to embrace the potential of large language models.

Agreement Tracking for Multi-Issue Negotiation Dialogues

Automated negotiation support systems aim to help human negotiators reach more favorable outcomes in multi-issue negotiations (e.g., an employer and a candidate negotiating over issues such as salary, hours, and promotions before a job offer). To be successful, these systems must accurately track agreements reached by participants in real-time. Existing approaches either focus on task-oriented dialogues or produce unstructured outputs, rendering them unsuitable for this objective. Our work introduces the novel task of agreement tracking for two-party multi-issue negotiations, which requires continuous monitoring of agreements within a structured state space. To address the scarcity of annotated corpora with realistic multi-issue negotiation dialogues, we use GPT-3 to build GPT-Negochat, a synthesized dataset that we make publicly available. We present a strong initial baseline for our task by transfer-learning a T5 model trained on the MultiWOZ 2.4 corpus. Pre-training T5-small and T5-base on MultiWOZ 2.4's DST task enhances results by 21% and 9% respectively over training solely on GPT-Negochat. We validate our method's sample-efficiency via smaller training subset experiments. By releasing GPT-Negochat and our baseline models, we aim to encourage further research in multi-issue negotiation dialogue agreement tracking.