{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:DLQYTAPNEMZIV22MXTYL6QTIMX","short_pith_number":"pith:DLQYTAPN","schema_version":"1.0","canonical_sha256":"1ae18981ed23328aeb4cbcf0bf426865fc7dee8020911409e4951cb8463b159f","source":{"kind":"arxiv","id":"2607.22639","version":1},"attestation_state":"computed","paper":{"title":"TRACE: Business Rule-Grounded Reasoning Curriculum for Knowledge-Preserving Parametric Tool Retrieval in Enterprise LLMs","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Ashutosh Hathidara, Christopher Toukmaji, Karthikeyan Asokkumar, Mayank Shrivastava, Sai Shruthi Sistla","submitted_at":"2026-06-22T23:15:35Z","abstract_excerpt":"Parametric retrieval enables LLMs to retrieve tools implicitly by assigning each API a unique virtual token and training the model to generate it via constrained beam search. Toolsense shows that this regime has two critical drawbacks: it destroys parametric tool knowledge during training, and its beam-search decoding is too slow for real-time deployment. We introduce TRACE (Tool Retrieval via Augmented Chain-of-thought and Enterprise rules), a two-stage curriculum that resolves this dissociation. Stage 1 reuses the multi-format memorization SFT from ToolSense to seed tool knowledge with LoRA."},"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":"2607.22639","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.AI","submitted_at":"2026-06-22T23:15:35Z","cross_cats_sorted":[],"title_canon_sha256":"6525166967303c5b6e4e5504c3924685245878ecd3a9b49556ffcee2c4aa1caa","abstract_canon_sha256":"670a1b55453c395bcb95e869f9e037c7d7cb5557cc0054996985e44c7e597cdd"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-28T00:21:46.535709Z","signature_b64":"UwFZMZ0m3F1Ma0pK0gEqQ8/U74q2S3bYVmx5ryCc58XGWJAWiBm5+Gm85IOAfgVAUFTRcgI9DRgAAc2SVxppCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1ae18981ed23328aeb4cbcf0bf426865fc7dee8020911409e4951cb8463b159f","last_reissued_at":"2026-07-28T00:21:46.534673Z","signature_status":"signed_v1","first_computed_at":"2026-07-28T00:21:46.534673Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TRACE: Business Rule-Grounded Reasoning Curriculum for Knowledge-Preserving Parametric Tool Retrieval in Enterprise LLMs","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Ashutosh Hathidara, Christopher Toukmaji, Karthikeyan Asokkumar, Mayank Shrivastava, Sai Shruthi Sistla","submitted_at":"2026-06-22T23:15:35Z","abstract_excerpt":"Parametric retrieval enables LLMs to retrieve tools implicitly by assigning each API a unique virtual token and training the model to generate it via constrained beam search. Toolsense shows that this regime has two critical drawbacks: it destroys parametric tool knowledge during training, and its beam-search decoding is too slow for real-time deployment. We introduce TRACE (Tool Retrieval via Augmented Chain-of-thought and Enterprise rules), a two-stage curriculum that resolves this dissociation. Stage 1 reuses the multi-format memorization SFT from ToolSense to seed tool knowledge with LoRA."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.22639","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/2607.22639/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":"2607.22639","created_at":"2026-07-28T00:21:46.535139+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.22639v1","created_at":"2026-07-28T00:21:46.535139+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.22639","created_at":"2026-07-28T00:21:46.535139+00:00"},{"alias_kind":"pith_short_12","alias_value":"DLQYTAPNEMZI","created_at":"2026-07-28T00:21:46.535139+00:00"},{"alias_kind":"pith_short_16","alias_value":"DLQYTAPNEMZIV22M","created_at":"2026-07-28T00:21:46.535139+00:00"},{"alias_kind":"pith_short_8","alias_value":"DLQYTAPN","created_at":"2026-07-28T00:21:46.535139+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/DLQYTAPNEMZIV22MXTYL6QTIMX","json":"https://pith.science/pith/DLQYTAPNEMZIV22MXTYL6QTIMX.json","graph_json":"https://pith.science/api/pith-number/DLQYTAPNEMZIV22MXTYL6QTIMX/graph.json","events_json":"https://pith.science/api/pith-number/DLQYTAPNEMZIV22MXTYL6QTIMX/events.json","paper":"https://pith.science/paper/DLQYTAPN"},"agent_actions":{"view_html":"https://pith.science/pith/DLQYTAPNEMZIV22MXTYL6QTIMX","download_json":"https://pith.science/pith/DLQYTAPNEMZIV22MXTYL6QTIMX.json","view_paper":"https://pith.science/paper/DLQYTAPN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.22639&json=true","fetch_graph":"https://pith.science/api/pith-number/DLQYTAPNEMZIV22MXTYL6QTIMX/graph.json","fetch_events":"https://pith.science/api/pith-number/DLQYTAPNEMZIV22MXTYL6QTIMX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DLQYTAPNEMZIV22MXTYL6QTIMX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DLQYTAPNEMZIV22MXTYL6QTIMX/action/storage_attestation","attest_author":"https://pith.science/pith/DLQYTAPNEMZIV22MXTYL6QTIMX/action/author_attestation","sign_citation":"https://pith.science/pith/DLQYTAPNEMZIV22MXTYL6QTIMX/action/citation_signature","submit_replication":"https://pith.science/pith/DLQYTAPNEMZIV22MXTYL6QTIMX/action/replication_record"}},"created_at":"2026-07-28T00:21:46.535139+00:00","updated_at":"2026-07-28T00:21:46.535139+00:00"}