# Coherence-Driven Multimodal Safety Dialogue with Active Learning for Embodied Agents

arXiv: https://arxiv.org/abs/2410.14141 · HTML: https://mlsanity.com/paper/2410.14141/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Placed in Agent threats, guardrails and governance, one of 16 directions in LLM agents and tool use.
- Builds on 17 of its 44 references that are in the map.
- First posted on arXiv 18 Oct 2024.

## Placed in

- LLM agents and tool use / [Agent threats, guardrails and governance](https://mlsanity.com/agents/agent-threats-guardrails-and-governance/)

## Builds on (most cited first, 17 of 17)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Do As I Can, Not As I Say: Grounding Language in Robotic Affordances](https://mlsanity.com/paper/2204.01691/) (2022)
- [MiniLM: Deep Self-Attention Distillation for Task-Agnostic Compression of Pre-Trained Transformers](https://mlsanity.com/paper/2002.10957/) (2020)
- [Language Models as Zero-Shot Planners: Extracting Actionable Knowledge for Embodied Agents](https://mlsanity.com/paper/2201.07207/) (2022)
- [A Discourse-Aware Attention Model for Abstractive Summarization of Long Documents](https://mlsanity.com/paper/1804.05685/) (2018)
- [LLaVA-Plus: Learning to Use Tools for Creating Multimodal Agents](https://mlsanity.com/paper/2311.05437/) (2023)
- [Cold-start Active Learning through Self-supervised Language Modeling](https://mlsanity.com/paper/2010.09535/) (2020)
- [On the Safety of Conversational Models: Taxonomy, Dataset, and Benchmark](https://mlsanity.com/paper/2110.08466/) (2021)
- [Discourse-Aware Neural Rewards for Coherent Text Generation](https://mlsanity.com/paper/1805.03766/) (2018)
- [Clue: Cross-modal Coherence Modeling for Caption Generation](https://mlsanity.com/paper/2005.00908/) (2020)
- [DisCGen: A Framework for Discourse-Informed Counterspeech Generation](https://mlsanity.com/paper/2311.18147/) (2023)
- [D-CALM: A Dynamic Clustering-based Active Learning Approach for Mitigating Bias](https://mlsanity.com/paper/2305.17013/) (2023)
- [Extending Implicit Discourse Relation Recognition to the PDTB-3](https://mlsanity.com/paper/2010.06294/) (2020)
- [Active Learning for Robust and Representative LLM Generation in Safety-Critical Scenarios](https://mlsanity.com/paper/2410.11114/) (2024)
- [Pre-trained Language Model Based Active Learning for Sentence Matching](https://mlsanity.com/paper/2010.05522/) (2020)
- [An Active Learning Framework for Inclusive Generation by Large Language Models](https://mlsanity.com/paper/2410.13641/) (2024)

## Built on by (most cited first, 8 of 8)

- [HomeSafeBench: Benchmarking Embodied Vision-Language Models in Free-Exploration Home Safety Inspection](https://mlsanity.com/paper/2509.23690/) (2025)
- [Adaptive Querying for Reward Learning from Human Feedback](https://mlsanity.com/paper/2412.07990/) (2024)
- [Moonworks Lunara Aesthetic Dataset](https://mlsanity.com/paper/2601.07941/) (2026)
- [Moonworks Lunara Aesthetic II: An Image Variation Dataset](https://mlsanity.com/paper/2602.01666/) (2026)
- [Mind the Gap: A Framework for Assessing Pitfalls in Multimodal Active Learning](https://mlsanity.com/paper/2603.29677/) (2026)
- [TSHA: A Benchmark for Visual Language Models in Trustworthy Safety Hazard Assessment Scenarios](https://mlsanity.com/paper/2603.29759/) (2026)
- [Moonworks Lunara: Modeling Artistic Intelligence](https://mlsanity.com/paper/2609.22272/) (2026)
- [Security and Privacy in Large-Model-Driven Embodied Agents: Attacks, Defenses, and Future Directions](https://mlsanity.com/paper/2609.27847/) (2026)

## Abstract (arXiv)

When assisting people in daily tasks, robots need to accurately interpret visual cues and respond effectively in diverse safety-critical situations, such as sharp objects on the floor. In this context, we present M-CoDAL, a multimodal-dialogue system specifically designed for embodied agents to better understand and communicate in safety-critical situations. The system leverages discourse coherence relations to enhance its contextual understanding and communication abilities. To train this system, we introduce a novel clustering-based active learning mechanism that utilizes an external Large Language Model (LLM) to identify informative instances. Our approach is evaluated using a newly created multimodal dataset comprising 1K safety violations extracted from 2K Reddit images. These violations are annotated using a Large Multimodal Model (LMM) and verified by human annotators. Results with this dataset demonstrate that our approach improves resolution of safety situations, user sentiment, as well as safety of the conversation. Next, we deploy our dialogue system on a Hello Robot Stretch robot and conduct a within-subject user study with real-world participants. In the study, participants role-play two safety scenarios with different levels of severity with the robot and receive interventions from our model and a baseline system powered by OpenAI's ChatGPT. The study results corroborate and extend the findings from the automated evaluation, showing that our proposed system is more persuasive in a real-world embodied agent setting.