{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:TQPC6CUXS7XXSQGSS6HPUNKU2U","short_pith_number":"pith:TQPC6CUX","schema_version":"1.0","canonical_sha256":"9c1e2f0a9797ef7940d2978efa3554d51e159d20b0e6d4e47398ffbc3d50bad2","source":{"kind":"arxiv","id":"2607.07646","version":1},"attestation_state":"computed","paper":{"title":"RL Post-Training Builds Compositional Reasoning Strategies","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Andrew Saxe, Azwar Abdulsalam, Nishil Patel","submitted_at":"2026-07-08T17:04:42Z","abstract_excerpt":"Does RL post-training merely amplify primitive skills already latent in a base model, or can it compose primitive skills into new higher-level strategies? We study this question in a fully observable rewrite-grammar environment where the pretraining distribution is known and every generated rewrite can be audited. A Transformer is pretrained on primitive symbol-rewrite chains and post-trained on a Trace-based reasoning task with only a binary final-answer reward. RL solves held-out problems that remain rarely solved by the pretrained model even under much larger sampling budgets, while rejecti"},"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":"2607.07646","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-07-08T17:04:42Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"fbf40a49852233c9d84a14cfab55e5243e713f6e059981899352eccc9fcd1692","abstract_canon_sha256":"2c81278d8a27c7f627afed6dfd04543992dc6f390a7d1834f521ee61d07986f6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-09T01:20:37.900037Z","signature_b64":"RfDchhjg0ZSjoBcXq9l//Ojbq/tYgESMkOXFfNM391suQ93LeDbnWJBvcwGi9DwMspc86HiuPW/tMZ6Qk4n0BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9c1e2f0a9797ef7940d2978efa3554d51e159d20b0e6d4e47398ffbc3d50bad2","last_reissued_at":"2026-07-09T01:20:37.899510Z","signature_status":"signed_v1","first_computed_at":"2026-07-09T01:20:37.899510Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"RL Post-Training Builds Compositional Reasoning Strategies","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Andrew Saxe, Azwar Abdulsalam, Nishil Patel","submitted_at":"2026-07-08T17:04:42Z","abstract_excerpt":"Does RL post-training merely amplify primitive skills already latent in a base model, or can it compose primitive skills into new higher-level strategies? We study this question in a fully observable rewrite-grammar environment where the pretraining distribution is known and every generated rewrite can be audited. A Transformer is pretrained on primitive symbol-rewrite chains and post-trained on a Trace-based reasoning task with only a binary final-answer reward. RL solves held-out problems that remain rarely solved by the pretrained model even under much larger sampling budgets, while rejecti"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.07646","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/2607.07646/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":"2607.07646","created_at":"2026-07-09T01:20:37.899574+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.07646v1","created_at":"2026-07-09T01:20:37.899574+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.07646","created_at":"2026-07-09T01:20:37.899574+00:00"},{"alias_kind":"pith_short_12","alias_value":"TQPC6CUXS7XX","created_at":"2026-07-09T01:20:37.899574+00:00"},{"alias_kind":"pith_short_16","alias_value":"TQPC6CUXS7XXSQGS","created_at":"2026-07-09T01:20:37.899574+00:00"},{"alias_kind":"pith_short_8","alias_value":"TQPC6CUX","created_at":"2026-07-09T01:20:37.899574+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TQPC6CUXS7XXSQGSS6HPUNKU2U","json":"https://pith.science/pith/TQPC6CUXS7XXSQGSS6HPUNKU2U.json","graph_json":"https://pith.science/api/pith-number/TQPC6CUXS7XXSQGSS6HPUNKU2U/graph.json","events_json":"https://pith.science/api/pith-number/TQPC6CUXS7XXSQGSS6HPUNKU2U/events.json","paper":"https://pith.science/paper/TQPC6CUX"},"agent_actions":{"view_html":"https://pith.science/pith/TQPC6CUXS7XXSQGSS6HPUNKU2U","download_json":"https://pith.science/pith/TQPC6CUXS7XXSQGSS6HPUNKU2U.json","view_paper":"https://pith.science/paper/TQPC6CUX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.07646&json=true","fetch_graph":"https://pith.science/api/pith-number/TQPC6CUXS7XXSQGSS6HPUNKU2U/graph.json","fetch_events":"https://pith.science/api/pith-number/TQPC6CUXS7XXSQGSS6HPUNKU2U/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TQPC6CUXS7XXSQGSS6HPUNKU2U/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TQPC6CUXS7XXSQGSS6HPUNKU2U/action/storage_attestation","attest_author":"https://pith.science/pith/TQPC6CUXS7XXSQGSS6HPUNKU2U/action/author_attestation","sign_citation":"https://pith.science/pith/TQPC6CUXS7XXSQGSS6HPUNKU2U/action/citation_signature","submit_replication":"https://pith.science/pith/TQPC6CUXS7XXSQGSS6HPUNKU2U/action/replication_record"}},"created_at":"2026-07-09T01:20:37.899574+00:00","updated_at":"2026-07-09T01:20:37.899574+00:00"}