{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:IUJGPJPPDOP22FNF2F2H4WXZTV","short_pith_number":"pith:IUJGPJPP","schema_version":"1.0","canonical_sha256":"451267a5ef1b9fad15a5d1747e5af99d4236eb74790a4484a3642c5063c03330","source":{"kind":"arxiv","id":"2406.07904","version":2},"attestation_state":"computed","paper":{"title":"Grounding Multimodal Large Language Models in Actions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Alexander Toshev, Andrew Szot, Bogdan Mazoure, Devon Hjelm, Harsh Agrawal, Zsolt Kira","submitted_at":"2024-06-12T06:12:04Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) have demonstrated a wide range of capabilities across many domains, including Embodied AI. In this work, we study how to best ground a MLLM into different embodiments and their associated action spaces, with the goal of leveraging the multimodal world knowledge of the MLLM. We first generalize a number of methods through a unified architecture and the lens of action space adaptors. For continuous actions, we show that a learned tokenization allows for sufficient modeling precision, yielding the best performance on downstream tasks. For discrete actions,"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2406.07904","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-12T06:12:04Z","cross_cats_sorted":[],"title_canon_sha256":"6ecda4831360f777399fcf86423bcd0a1ac1a54f25fdecf176219afbb99bd9be","abstract_canon_sha256":"f9ee097402b4884975a289f8878fe6c557ee85d10214749cbdff209e81ed3e45"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:45:50.186591Z","signature_b64":"zSOrzUolXXnjpikpESb5Bfbb3ZVo3KJuGbJkvokBDS73AXprGnX/I7aJVumuG8hem2UVglXgh/Izt3BwrE9XCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"451267a5ef1b9fad15a5d1747e5af99d4236eb74790a4484a3642c5063c03330","last_reissued_at":"2026-07-05T09:45:50.186117Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:45:50.186117Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Grounding Multimodal Large Language Models in Actions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Alexander Toshev, Andrew Szot, Bogdan Mazoure, Devon Hjelm, Harsh Agrawal, Zsolt Kira","submitted_at":"2024-06-12T06:12:04Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) have demonstrated a wide range of capabilities across many domains, including Embodied AI. In this work, we study how to best ground a MLLM into different embodiments and their associated action spaces, with the goal of leveraging the multimodal world knowledge of the MLLM. We first generalize a number of methods through a unified architecture and the lens of action space adaptors. For continuous actions, we show that a learned tokenization allows for sufficient modeling precision, yielding the best performance on downstream tasks. For discrete actions,"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.07904","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2406.07904/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2406.07904","created_at":"2026-07-05T09:45:50.186172+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.07904v2","created_at":"2026-07-05T09:45:50.186172+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.07904","created_at":"2026-07-05T09:45:50.186172+00:00"},{"alias_kind":"pith_short_12","alias_value":"IUJGPJPPDOP2","created_at":"2026-07-05T09:45:50.186172+00:00"},{"alias_kind":"pith_short_16","alias_value":"IUJGPJPPDOP22FNF","created_at":"2026-07-05T09:45:50.186172+00:00"},{"alias_kind":"pith_short_8","alias_value":"IUJGPJPP","created_at":"2026-07-05T09:45:50.186172+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2501.00398","citing_title":"TSPE: Task-Specific Prompt Ensemble for Improved Zero-Shot Audio Classification","ref_index":30,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IUJGPJPPDOP22FNF2F2H4WXZTV","json":"https://pith.science/pith/IUJGPJPPDOP22FNF2F2H4WXZTV.json","graph_json":"https://pith.science/api/pith-number/IUJGPJPPDOP22FNF2F2H4WXZTV/graph.json","events_json":"https://pith.science/api/pith-number/IUJGPJPPDOP22FNF2F2H4WXZTV/events.json","paper":"https://pith.science/paper/IUJGPJPP"},"agent_actions":{"view_html":"https://pith.science/pith/IUJGPJPPDOP22FNF2F2H4WXZTV","download_json":"https://pith.science/pith/IUJGPJPPDOP22FNF2F2H4WXZTV.json","view_paper":"https://pith.science/paper/IUJGPJPP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.07904&json=true","fetch_graph":"https://pith.science/api/pith-number/IUJGPJPPDOP22FNF2F2H4WXZTV/graph.json","fetch_events":"https://pith.science/api/pith-number/IUJGPJPPDOP22FNF2F2H4WXZTV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IUJGPJPPDOP22FNF2F2H4WXZTV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IUJGPJPPDOP22FNF2F2H4WXZTV/action/storage_attestation","attest_author":"https://pith.science/pith/IUJGPJPPDOP22FNF2F2H4WXZTV/action/author_attestation","sign_citation":"https://pith.science/pith/IUJGPJPPDOP22FNF2F2H4WXZTV/action/citation_signature","submit_replication":"https://pith.science/pith/IUJGPJPPDOP22FNF2F2H4WXZTV/action/replication_record"}},"created_at":"2026-07-05T09:45:50.186172+00:00","updated_at":"2026-07-05T09:45:50.186172+00:00"}