Learning to Reason over Scene Graphs: A Case Study of Finetuning GPT-2 into a Robot Language Model for Grounded Task Planning
Long-horizon task planning is essential for the development of intelligent assistive and service robots. In this work, we investigate the applicability of a smaller class of large language models (LLMs), specifically GPT-2, in robotic task planning by learning to decompose tasks into subgoal specifi...
Gespeichert in:
Hauptverfasser: | , , , , , |
---|---|
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
container_end_page | |
---|---|
container_issue | |
container_start_page | |
container_title | |
container_volume | |
creator | Chalvatzaki, Georgia Younes, Ali Nandha, Daljeet Le, An Ribeiro, Leonardo F. R Gurevych, Iryna |
description | Long-horizon task planning is essential for the development of intelligent
assistive and service robots. In this work, we investigate the applicability of
a smaller class of large language models (LLMs), specifically GPT-2, in robotic
task planning by learning to decompose tasks into subgoal specifications for a
planner to execute sequentially. Our method grounds the input of the LLM on the
domain that is represented as a scene graph, enabling it to translate human
requests into executable robot plans, thereby learning to reason over
long-horizon tasks, as encountered in the ALFRED benchmark. We compare our
approach with classical planning and baseline methods to examine the
applicability and generalizability of LLM-based planners. Our findings suggest
that the knowledge stored in an LLM can be effectively grounded to perform
long-horizon task planning, demonstrating the promising potential for the
future application of neuro-symbolic planning methods in robotics. |
doi_str_mv | 10.48550/arxiv.2305.07716 |
format | Article |
fullrecord | <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2305_07716</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2305_07716</sourcerecordid><originalsourceid>FETCH-LOGICAL-a676-bc29074400455774f56a1270b6d339de399e6697e8a1296ebd6bc2fdbcd0f0273</originalsourceid><addsrcrecordid>eNotkM1Og0AUhdm4MNUHcOV9AXD4m-m4a4hFExqblj25MHeQiDPNAI1d-O5SdHWSk3xfco7nPYQsSNZpyp7QfXfnIIpZGjAhQn7r_RSEznSmhdHCgXCwBuyZHBwbMgS5w9PH8AwbyHAgOI6TuoDVsO0MjdPC5fvSj6AzM49wsLUdoUDTTtgS7KyiHrR1s8hORpGCEodP2PdorvCdd6OxH-j-P1deuX0ps1e_eM_fsk3hIxfcr5tIMpEkjCVpKkSiU45hJFjNVRxLRbGUxLkUtJ5ryalWfEa0qhvFNItEvPIe_7TL_urkui90l-r6Q7X8EP8C-5JXgA</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Learning to Reason over Scene Graphs: A Case Study of Finetuning GPT-2 into a Robot Language Model for Grounded Task Planning</title><source>arXiv.org</source><creator>Chalvatzaki, Georgia ; Younes, Ali ; Nandha, Daljeet ; Le, An ; Ribeiro, Leonardo F. R ; Gurevych, Iryna</creator><creatorcontrib>Chalvatzaki, Georgia ; Younes, Ali ; Nandha, Daljeet ; Le, An ; Ribeiro, Leonardo F. R ; Gurevych, Iryna</creatorcontrib><description>Long-horizon task planning is essential for the development of intelligent
assistive and service robots. In this work, we investigate the applicability of
a smaller class of large language models (LLMs), specifically GPT-2, in robotic
task planning by learning to decompose tasks into subgoal specifications for a
planner to execute sequentially. Our method grounds the input of the LLM on the
domain that is represented as a scene graph, enabling it to translate human
requests into executable robot plans, thereby learning to reason over
long-horizon tasks, as encountered in the ALFRED benchmark. We compare our
approach with classical planning and baseline methods to examine the
applicability and generalizability of LLM-based planners. Our findings suggest
that the knowledge stored in an LLM can be effectively grounded to perform
long-horizon task planning, demonstrating the promising potential for the
future application of neuro-symbolic planning methods in robotics.</description><identifier>DOI: 10.48550/arxiv.2305.07716</identifier><language>eng</language><subject>Computer Science - Artificial Intelligence ; Computer Science - Robotics</subject><creationdate>2023-05</creationdate><rights>http://creativecommons.org/licenses/by/4.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,881</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2305.07716$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2305.07716$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Chalvatzaki, Georgia</creatorcontrib><creatorcontrib>Younes, Ali</creatorcontrib><creatorcontrib>Nandha, Daljeet</creatorcontrib><creatorcontrib>Le, An</creatorcontrib><creatorcontrib>Ribeiro, Leonardo F. R</creatorcontrib><creatorcontrib>Gurevych, Iryna</creatorcontrib><title>Learning to Reason over Scene Graphs: A Case Study of Finetuning GPT-2 into a Robot Language Model for Grounded Task Planning</title><description>Long-horizon task planning is essential for the development of intelligent
assistive and service robots. In this work, we investigate the applicability of
a smaller class of large language models (LLMs), specifically GPT-2, in robotic
task planning by learning to decompose tasks into subgoal specifications for a
planner to execute sequentially. Our method grounds the input of the LLM on the
domain that is represented as a scene graph, enabling it to translate human
requests into executable robot plans, thereby learning to reason over
long-horizon tasks, as encountered in the ALFRED benchmark. We compare our
approach with classical planning and baseline methods to examine the
applicability and generalizability of LLM-based planners. Our findings suggest
that the knowledge stored in an LLM can be effectively grounded to perform
long-horizon task planning, demonstrating the promising potential for the
future application of neuro-symbolic planning methods in robotics.</description><subject>Computer Science - Artificial Intelligence</subject><subject>Computer Science - Robotics</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotkM1Og0AUhdm4MNUHcOV9AXD4m-m4a4hFExqblj25MHeQiDPNAI1d-O5SdHWSk3xfco7nPYQsSNZpyp7QfXfnIIpZGjAhQn7r_RSEznSmhdHCgXCwBuyZHBwbMgS5w9PH8AwbyHAgOI6TuoDVsO0MjdPC5fvSj6AzM49wsLUdoUDTTtgS7KyiHrR1s8hORpGCEodP2PdorvCdd6OxH-j-P1deuX0ps1e_eM_fsk3hIxfcr5tIMpEkjCVpKkSiU45hJFjNVRxLRbGUxLkUtJ5ryalWfEa0qhvFNItEvPIe_7TL_urkui90l-r6Q7X8EP8C-5JXgA</recordid><startdate>20230512</startdate><enddate>20230512</enddate><creator>Chalvatzaki, Georgia</creator><creator>Younes, Ali</creator><creator>Nandha, Daljeet</creator><creator>Le, An</creator><creator>Ribeiro, Leonardo F. R</creator><creator>Gurevych, Iryna</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20230512</creationdate><title>Learning to Reason over Scene Graphs: A Case Study of Finetuning GPT-2 into a Robot Language Model for Grounded Task Planning</title><author>Chalvatzaki, Georgia ; Younes, Ali ; Nandha, Daljeet ; Le, An ; Ribeiro, Leonardo F. R ; Gurevych, Iryna</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a676-bc29074400455774f56a1270b6d339de399e6697e8a1296ebd6bc2fdbcd0f0273</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Computer Science - Artificial Intelligence</topic><topic>Computer Science - Robotics</topic><toplevel>online_resources</toplevel><creatorcontrib>Chalvatzaki, Georgia</creatorcontrib><creatorcontrib>Younes, Ali</creatorcontrib><creatorcontrib>Nandha, Daljeet</creatorcontrib><creatorcontrib>Le, An</creatorcontrib><creatorcontrib>Ribeiro, Leonardo F. R</creatorcontrib><creatorcontrib>Gurevych, Iryna</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Chalvatzaki, Georgia</au><au>Younes, Ali</au><au>Nandha, Daljeet</au><au>Le, An</au><au>Ribeiro, Leonardo F. R</au><au>Gurevych, Iryna</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Learning to Reason over Scene Graphs: A Case Study of Finetuning GPT-2 into a Robot Language Model for Grounded Task Planning</atitle><date>2023-05-12</date><risdate>2023</risdate><abstract>Long-horizon task planning is essential for the development of intelligent
assistive and service robots. In this work, we investigate the applicability of
a smaller class of large language models (LLMs), specifically GPT-2, in robotic
task planning by learning to decompose tasks into subgoal specifications for a
planner to execute sequentially. Our method grounds the input of the LLM on the
domain that is represented as a scene graph, enabling it to translate human
requests into executable robot plans, thereby learning to reason over
long-horizon tasks, as encountered in the ALFRED benchmark. We compare our
approach with classical planning and baseline methods to examine the
applicability and generalizability of LLM-based planners. Our findings suggest
that the knowledge stored in an LLM can be effectively grounded to perform
long-horizon task planning, demonstrating the promising potential for the
future application of neuro-symbolic planning methods in robotics.</abstract><doi>10.48550/arxiv.2305.07716</doi><oa>free_for_read</oa></addata></record> |
fulltext | fulltext_linktorsrc |
identifier | DOI: 10.48550/arxiv.2305.07716 |
ispartof | |
issn | |
language | eng |
recordid | cdi_arxiv_primary_2305_07716 |
source | arXiv.org |
subjects | Computer Science - Artificial Intelligence Computer Science - Robotics |
title | Learning to Reason over Scene Graphs: A Case Study of Finetuning GPT-2 into a Robot Language Model for Grounded Task Planning |
url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-15T10%3A25%3A23IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Learning%20to%20Reason%20over%20Scene%20Graphs:%20A%20Case%20Study%20of%20Finetuning%20GPT-2%20into%20a%20Robot%20Language%20Model%20for%20Grounded%20Task%20Planning&rft.au=Chalvatzaki,%20Georgia&rft.date=2023-05-12&rft_id=info:doi/10.48550/arxiv.2305.07716&rft_dat=%3Carxiv_GOX%3E2305_07716%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true |