no code implementations • 13 Mar 2024 • Yao Fu, Dong-Ki Kim, Jaekyeom Kim, Sungryull Sohn, Lajanugen Logeswaran, Kyunghoon Bae, Honglak Lee
The primary limitation of large language models (LLMs) is their restricted understanding of the world.
1 code implementation • 7 Dec 2023 • Sungryull Sohn, Yiwei Lyu, Anthony Liu, Lajanugen Logeswaran, Dong-Ki Kim, Dongsub Shim, Honglak Lee
Our TOD-Flow graph learns what a model can, should, and should not predict, effectively reducing the search space and providing a rationale for the model's prediction.
no code implementations • 16 Nov 2023 • Lajanugen Logeswaran, Sungryull Sohn, Yiwei Lyu, Anthony Zhe Liu, Dong-Ki Kim, Dongsub Shim, Moontae Lee, Honglak Lee
One of the fundamental skills required for an agent acting in an environment to complete tasks is the ability to understand what actions are plausible at any given point.
no code implementations • 25 Oct 2023 • Dong-Ki Kim, Sungryull Sohn, Lajanugen Logeswaran, Dongsub Shim, Honglak Lee
Recently, there has been an increasing interest in automated prompt optimization based on reinforcement learning (RL).
Multi-agent Reinforcement Learning reinforcement-learning +1
1 code implementation • 24 Oct 2023 • Zheyuan Zhang, Shane Storks, Fengyuan Hu, Sungryull Sohn, Moontae Lee, Honglak Lee, Joyce Chai
We incorporate these interlinked dual processes in fine-tuning and in-context learning with PLMs, applying them to two language understanding tasks that require coherent physical commonsense reasoning.
no code implementations • 16 Mar 2023 • Anthony Z. Liu, Lajanugen Logeswaran, Sungryull Sohn, Honglak Lee
Planning is an important capability of artificial agents that perform long-horizon tasks in real-world environments.
no code implementations • 17 Feb 2023 • Lajanugen Logeswaran, Sungryull Sohn, Yunseok Jang, Moontae Lee, Honglak Lee
This work explores the problem of generating task graphs of real-world activities.
no code implementations • 17 Feb 2023 • Yunseok Jang, Sungryull Sohn, Lajanugen Logeswaran, Tiange Luo, Moontae Lee, Honglak Lee
Real-world tasks consist of multiple inter-dependent subtasks (e. g., a dirty pan needs to be washed before it can be used for cooking).
no code implementations • 25 May 2022 • Sungryull Sohn, Hyunjae Woo, Jongwook Choi, lyubing qiang, Izzeddin Gur, Aleksandra Faust, Honglak Lee
Different from the previous meta-rl methods trying to directly infer the unstructured task embedding, our multi-task subtask graph inferencer (MTSGI) first infers the common high-level task structure in terms of the subtask graph from the training tasks, and use it as a prior to improve the task inference in testing.
Hierarchical Reinforcement Learning Meta Reinforcement Learning +2
1 code implementation • 28 Mar 2022 • Anthony Z. Liu, Sungryull Sohn, Mahdi Qazwini, Honglak Lee
These subtasks are defined in terms of entities (e. g., "apple", "pear") that can be recombined to form new subtasks (e. g., "pickup apple", and "pickup pear").
1 code implementation • NeurIPS 2021 • Christopher Hoang, Sungryull Sohn, Jongwook Choi, Wilka Carvalho, Honglak Lee
SFL leverages the ability of successor features (SF) to capture transition dynamics, using it to drive exploration by estimating state-novelty and to enable high-level planning by abstracting the state-space as a non-parametric landmark-based graph.
1 code implementation • 13 Jul 2021 • Sungryull Sohn, Sungtae Lee, Jongwook Choi, Harm van Seijen, Mehdi Fatemi, Honglak Lee
We propose the k-Shortest-Path (k-SP) constraint: a novel constraint on the agent's trajectory that improves the sample efficiency in sparse-reward MDPs.
no code implementations • 28 Oct 2020 • Wilka Carvalho, Anthony Liang, Kimin Lee, Sungryull Sohn, Honglak Lee, Richard L. Lewis, Satinder Singh
In this work, we show that one can learn object-interaction tasks from scratch without supervision by learning an attentive object-model as an auxiliary task during task learning with an object-centric relational RL agent.
no code implementations • 8 Feb 2020 • Sungryull Sohn, Yin-Lam Chow, Jayden Ooi, Ofir Nachum, Honglak Lee, Ed Chi, Craig Boutilier
In batch reinforcement learning (RL), one often constrains a learned policy to be close to the behavior (data-generating) policy, e. g., by constraining the learned action distribution to differ from the behavior policy by some maximum degree that is the same at each state.
1 code implementation • ICLR 2020 • Sungryull Sohn, Hyunjae Woo, Jongwook Choi, Honglak Lee
We propose and address a novel few-shot RL problem, where a task is characterized by a subtask graph which describes a set of subtasks and their dependencies that are unknown to the agent.
1 code implementation • NeurIPS 2018 • Sungryull Sohn, Junhyuk Oh, Honglak Lee
We introduce a new RL problem where the agent is required to generalize to a previously-unseen environment characterized by a subtask graph which describes a set of subtasks and their dependencies.
no code implementations • ICLR 2018 • Sungryull Sohn, Junhyuk Oh, Honglak Lee
Unlike existing approaches which explicitly describe what the agent should do, our problem only describes properties of subtasks and relationships between them, which requires the agent to perform a complex reasoning to find the optimal subtask to execute.
2 code implementations • ICML 2017 • Ruben Villegas, Jimei Yang, Yuliang Zou, Sungryull Sohn, Xunyu Lin, Honglak Lee
To avoid inherent compounding errors in recursive pixel-level prediction, we propose to first estimate high-level structure in the input frames, then predict how that structure evolves in the future, and finally by observing a single frame from the past and the predicted high-level structure, we construct the future frames without having to observe any of the pixel-level predictions.