{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:3JMZEFV2NOUKAK6VBHRJZ6ES3L","short_pith_number":"pith:3JMZEFV2","schema_version":"1.0","canonical_sha256":"da599216ba6ba8a02bd509e29cf892daf685521d996d33df5521a0cce9e0b6af","source":{"kind":"arxiv","id":"2507.15788","version":1},"attestation_state":"computed","paper":{"title":"Small LLMs Do Not Learn a Generalizable Theory of Mind via Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Hanan Salam, Sneheel Sarangi","submitted_at":"2025-07-21T16:47:59Z","abstract_excerpt":"Recent advancements in large language models (LLMs) have demonstrated emergent capabilities in complex reasoning, largely spurred by rule-based Reinforcement Learning (RL) techniques applied during the post-training. This has raised the question of whether similar methods can instill more nuanced, human-like social intelligence, such as a Theory of Mind (ToM), in LLMs. This paper investigates whether small-scale LLMs can acquire a robust and generalizable ToM capability through RL with verifiable rewards (RLVR). We conduct a systematic evaluation by training models on various combinations of p"},"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":"2507.15788","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-07-21T16:47:59Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"52898d728d827f9a36b6ceb185fc80b247dcb69f87be3de6a3a661ee6eec8c77","abstract_canon_sha256":"bd25b121fd31b68623bfeb234644fe75b6579c3809eba0bdae9254622398776f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:40:44.280263Z","signature_b64":"9yr3a2//F/SZJ0KFJBXMKWKE4P2Bz1MX0oj/AEZMBSwjnj0Pinc7vqcTDIvTWlL96KlETZH9WWXNS3S2RSx0Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"da599216ba6ba8a02bd509e29cf892daf685521d996d33df5521a0cce9e0b6af","last_reissued_at":"2026-07-05T11:40:44.279793Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:40:44.279793Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Small LLMs Do Not Learn a Generalizable Theory of Mind via Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Hanan Salam, Sneheel Sarangi","submitted_at":"2025-07-21T16:47:59Z","abstract_excerpt":"Recent advancements in large language models (LLMs) have demonstrated emergent capabilities in complex reasoning, largely spurred by rule-based Reinforcement Learning (RL) techniques applied during the post-training. This has raised the question of whether similar methods can instill more nuanced, human-like social intelligence, such as a Theory of Mind (ToM), in LLMs. This paper investigates whether small-scale LLMs can acquire a robust and generalizable ToM capability through RL with verifiable rewards (RLVR). We conduct a systematic evaluation by training models on various combinations of p"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.15788","kind":"arxiv","version":1},"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/2507.15788/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":"2507.15788","created_at":"2026-07-05T11:40:44.279845+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.15788v1","created_at":"2026-07-05T11:40:44.279845+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.15788","created_at":"2026-07-05T11:40:44.279845+00:00"},{"alias_kind":"pith_short_12","alias_value":"3JMZEFV2NOUK","created_at":"2026-07-05T11:40:44.279845+00:00"},{"alias_kind":"pith_short_16","alias_value":"3JMZEFV2NOUKAK6V","created_at":"2026-07-05T11:40:44.279845+00:00"},{"alias_kind":"pith_short_8","alias_value":"3JMZEFV2","created_at":"2026-07-05T11:40:44.279845+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.09092","citing_title":"From Shortcuts to Reasoning: Robust Post-Training of Theory of Mind with Reinforcement Learning","ref_index":67,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3JMZEFV2NOUKAK6VBHRJZ6ES3L","json":"https://pith.science/pith/3JMZEFV2NOUKAK6VBHRJZ6ES3L.json","graph_json":"https://pith.science/api/pith-number/3JMZEFV2NOUKAK6VBHRJZ6ES3L/graph.json","events_json":"https://pith.science/api/pith-number/3JMZEFV2NOUKAK6VBHRJZ6ES3L/events.json","paper":"https://pith.science/paper/3JMZEFV2"},"agent_actions":{"view_html":"https://pith.science/pith/3JMZEFV2NOUKAK6VBHRJZ6ES3L","download_json":"https://pith.science/pith/3JMZEFV2NOUKAK6VBHRJZ6ES3L.json","view_paper":"https://pith.science/paper/3JMZEFV2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.15788&json=true","fetch_graph":"https://pith.science/api/pith-number/3JMZEFV2NOUKAK6VBHRJZ6ES3L/graph.json","fetch_events":"https://pith.science/api/pith-number/3JMZEFV2NOUKAK6VBHRJZ6ES3L/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3JMZEFV2NOUKAK6VBHRJZ6ES3L/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3JMZEFV2NOUKAK6VBHRJZ6ES3L/action/storage_attestation","attest_author":"https://pith.science/pith/3JMZEFV2NOUKAK6VBHRJZ6ES3L/action/author_attestation","sign_citation":"https://pith.science/pith/3JMZEFV2NOUKAK6VBHRJZ6ES3L/action/citation_signature","submit_replication":"https://pith.science/pith/3JMZEFV2NOUKAK6VBHRJZ6ES3L/action/replication_record"}},"created_at":"2026-07-05T11:40:44.279845+00:00","updated_at":"2026-07-05T11:40:44.279845+00:00"}