An official website of the United States government
Large Language Models (LLMs) have achieved remarkable success across a wide array of tasks. Due to the impressive planning and reasoning abilities of LLMs, they have been used as autonomous agents to do many tasks automatically. Recently, based on the development of using one LLM as a single planning or decision-making agent, LLM-based multi-agent systems have achieved considerable progress in complex problem-solving and world simulation. To provide the community with an overview of this dynamic field, we present this survey to offer an in-depth discussion on the essential aspects of multi-agent systems based on LLMs, as well as the challenges. Our goal is for readers to gain substantial insights on the following questions: What domains and environments do LLM-based multi-agents simulate? How are these agents profiled and how do they communicate? What mechanisms contribute to the growth of agents' capacities? For those interested in delving into this field of study, we also summarize the commonly used datasets or benchmarks for them to have convenient access. To keep researchers updated on the latest studies, we maintain an open-source GitHub repository, dedicated to outlining the research on LLM-based multi-agent systems.
more »
« less
Augmenting large language models with chemistry tools
Abstract Large language models (LLMs) have shown strong performance in tasks across domains but struggle with chemistry-related problems. These models also lack access to external knowledge sources, limiting their usefulness in scientific applications. We introduce ChemCrow, an LLM chemistry agent designed to accomplish tasks across organic synthesis, drug discovery and materials design. By integrating 18 expert-designed tools and using GPT-4 as the LLM, ChemCrow augments the LLM performance in chemistry, and new capabilities emerge. Our agent autonomously planned and executed the syntheses of an insect repellent and three organocatalysts and guided the discovery of a novel chromophore. Our evaluation, including both LLM and expert assessments, demonstrates ChemCrow’s effectiveness in automating a diverse set of chemical tasks. Our work not only aids expert chemists and lowers barriers for non-experts but also fosters scientific advancement by bridging the gap between experimental and computational chemistry.
more »
« less
- Award ID(s):
- 1751471
- PAR ID:
- 10505760
- Publisher / Repository:
- Nature Publishing Group
- Date Published:
- Journal Name:
- Nature Machine Intelligence
- Volume:
- 6
- Issue:
- 5
- ISSN:
- 2522-5839
- Format(s):
- Medium: X Size: p. 525-535
- Size(s):
- p. 525-535
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Large Language Models (LLMs) have a natural role in answering complex queries about data streams, but the high computational cost of LLM inference makes them infeasible in many such tasks. We propose online cascade learning as an approach to address this challenge. The objective here is to learn a “cascade” of models, starting with lower-capacity models (such as logistic regression) and ending with a powerful LLM, along with a deferral policy that determines the model to be used on a given input. We formulate the task of learning cascades online as an imitation-learning problem, where smaller models are updated over time imitating the LLM expert demonstrations, and give a no-regret algorithm for the problem. Experimental results across four benchmarks show that our method parallels LLMs in accuracy while cutting down inference costs by as much as 90% with strong robustness against input distribution shifts, underscoring its efficacy and adaptability in stream processing.more » « less
-
We present an algorithm for skill discovery from expert demonstrations. The algorithm first utilizes Large Language Models (LLMs) to propose an initial segmentation of the trajectories. Following that, a hierarchical variational inference framework incorporates the LLM-generated segmentation information to discover reusable skills by merging trajectory segments. To further control the trade-off between compression and reusability, we introduce a novel auxiliary objective based on the Minimum Description Length principle that helps guide this skill discovery process. Our results demonstrate that agents equipped with our method are able to discover skills that help accelerate learning and outperform baseline skill learning approaches on new long-horizon tasks in BabyAI, a grid world navigation environment, as well as ALFRED, a household simulation environment.more » « less
-
Language models are aligned to emulate the collective voice of many, resulting in outputs that align with no one in particular. Steering LLMs away from generic output is possible through supervised finetuning or RLHF, but requires prohibitively large datasets for new ad-hoc tasks. We argue that it is instead possible to align an LLM to a specific setting by leveraging a very small number (< 10) of demonstrations as feedback. Our method, Demonstration ITerated Task Optimization (DITTO), directly aligns language model outputs to a user's demonstrated behaviors. Derived using ideas from online imitation learning, DITTO cheaply generates online comparison data by treating users' demonstrations as preferred over output from the LLM and its intermediate checkpoints. Concretely, DITTO operates by having an LLM generate examples that are presumed to be inferior to expert demonstrations. The method iteratively constructs pairwise preference relationships between these LLM-generated samples and expert demonstrations, potentially including comparisons between different training checkpoints. These constructed preference pairs are then used to train the model using a preference optimization algorithm (e.g. DPO). We evaluate DITTO's ability to learn fine-grained style and task alignment across domains such as news articles, emails, and blog posts. Additionally, we conduct a user study soliciting a range of demonstrations from participants (N = 16). Across our benchmarks and user study, we find that win-rates for DITTO outperform few-shot prompting, supervised fine-tuning, and other self-play methods by an avg. of 19% points. By using demonstrations as feedback directly, DITTO offers a novel method for effective customization of LLMs.more » « less
-
While Reinforcement learning (RL), especially Deep RL (DRL), has shown outstanding performance in video games, little evidence has shown that DRL can be successfully applied to human-centric tasks where the ultimate RL goal is to make the \textit{human-agent interactions} productive and fruitful. In real-life, complex, human-centric tasks, such as education and healthcare, data can be noisy and limited. Batch RL is designed for handling such situations where data is \textit{limited yet noisy}, and where \textit{building simulations is challenging}. In two consecutive empirical studies, we investigated Batch DRL for pedagogical policy induction, to choose student learning activities in an Intelligent Tutoring System. In Fall 2018 (F18), we compared the Batch DRL policy to an Expert policy, but found no significant difference between the DRL and Expert policies. In Spring 2019 (S19), we augmented the Batch DRL-induced policy with \textit{a simple act of explanation} by showing a message such as \textit{"The AI agent thinks you should view this problem as a Worked Example to learn how some new rules work."}. We compared this policy against two conditions, the Expert policy, and a student decision making policy. Our results show that 1) the Batch DRL policy with explanations significantly improved student learning performance more than the Expert policy; and 2) no significant differences were found between the Expert policy and student decision making. Overall, our results suggest that \textit{pairing simple explanations with the Batch DRL policy} can be an important and effective technique for applying RL to real-life, human-centric tasks.more » « less
