{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:UOXHR7CDY5CDUNSNZV6OL7IOHP","short_pith_number":"pith:UOXHR7CD","schema_version":"1.0","canonical_sha256":"a3ae78fc43c7443a364dcd7ce5fd0e3bef90646adf40fd07c18bfcc2ef3cf01e","source":{"kind":"arxiv","id":"2607.07050","version":1},"attestation_state":"computed","paper":{"title":"Behavior Leverage Imbalance in Multi-Teacher On-Policy Distillation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Chengjun Mao, Guang Chen, Jiabin Shen","submitted_at":"2026-07-08T06:26:13Z","abstract_excerpt":"Agentic language models must learn when to call tools, when to consume tool responses, and when to answer directly. This makes multi-teacher on-policy distillation a natural training strategy: one teacher can specialize in tool calls, another in direct responses, and the student can learn from both on\n  its own generated distribution. We show that this strategy can induce a behavior shift that is invisible from aggregate losses alone. In a two-teacher tool-use setting, vanilla generalized knowledge distillation improves tool-call recall but also moves the model toward over-calling, where it ca"},"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":"2607.07050","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-07-08T06:26:13Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"b63d5e1d3f87835ed76d7dfc9a19967050a7911466a8ee622863b1d187e744fc","abstract_canon_sha256":"dd2eb8c61f4e798e8bbdbdd68d0b62868d5e91ae1207b918d99931447f3da060"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-09T01:20:13.287974Z","signature_b64":"Njz7mh+1rFU81ZDyVhFv8X6CCAwPp9Sy5+Tuae+mdkSGt1vn5rLH+9LW2L9ftZtPWt5nygKBbwbGnCz9JrMuDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a3ae78fc43c7443a364dcd7ce5fd0e3bef90646adf40fd07c18bfcc2ef3cf01e","last_reissued_at":"2026-07-09T01:20:13.287621Z","signature_status":"signed_v1","first_computed_at":"2026-07-09T01:20:13.287621Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Behavior Leverage Imbalance in Multi-Teacher On-Policy Distillation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Chengjun Mao, Guang Chen, Jiabin Shen","submitted_at":"2026-07-08T06:26:13Z","abstract_excerpt":"Agentic language models must learn when to call tools, when to consume tool responses, and when to answer directly. This makes multi-teacher on-policy distillation a natural training strategy: one teacher can specialize in tool calls, another in direct responses, and the student can learn from both on\n  its own generated distribution. We show that this strategy can induce a behavior shift that is invisible from aggregate losses alone. In a two-teacher tool-use setting, vanilla generalized knowledge distillation improves tool-call recall but also moves the model toward over-calling, where it ca"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.07050","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/2607.07050/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":"2607.07050","created_at":"2026-07-09T01:20:13.287677+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.07050v1","created_at":"2026-07-09T01:20:13.287677+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.07050","created_at":"2026-07-09T01:20:13.287677+00:00"},{"alias_kind":"pith_short_12","alias_value":"UOXHR7CDY5CD","created_at":"2026-07-09T01:20:13.287677+00:00"},{"alias_kind":"pith_short_16","alias_value":"UOXHR7CDY5CDUNSN","created_at":"2026-07-09T01:20:13.287677+00:00"},{"alias_kind":"pith_short_8","alias_value":"UOXHR7CD","created_at":"2026-07-09T01:20:13.287677+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/UOXHR7CDY5CDUNSNZV6OL7IOHP","json":"https://pith.science/pith/UOXHR7CDY5CDUNSNZV6OL7IOHP.json","graph_json":"https://pith.science/api/pith-number/UOXHR7CDY5CDUNSNZV6OL7IOHP/graph.json","events_json":"https://pith.science/api/pith-number/UOXHR7CDY5CDUNSNZV6OL7IOHP/events.json","paper":"https://pith.science/paper/UOXHR7CD"},"agent_actions":{"view_html":"https://pith.science/pith/UOXHR7CDY5CDUNSNZV6OL7IOHP","download_json":"https://pith.science/pith/UOXHR7CDY5CDUNSNZV6OL7IOHP.json","view_paper":"https://pith.science/paper/UOXHR7CD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.07050&json=true","fetch_graph":"https://pith.science/api/pith-number/UOXHR7CDY5CDUNSNZV6OL7IOHP/graph.json","fetch_events":"https://pith.science/api/pith-number/UOXHR7CDY5CDUNSNZV6OL7IOHP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UOXHR7CDY5CDUNSNZV6OL7IOHP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UOXHR7CDY5CDUNSNZV6OL7IOHP/action/storage_attestation","attest_author":"https://pith.science/pith/UOXHR7CDY5CDUNSNZV6OL7IOHP/action/author_attestation","sign_citation":"https://pith.science/pith/UOXHR7CDY5CDUNSNZV6OL7IOHP/action/citation_signature","submit_replication":"https://pith.science/pith/UOXHR7CDY5CDUNSNZV6OL7IOHP/action/replication_record"}},"created_at":"2026-07-09T01:20:13.287677+00:00","updated_at":"2026-07-09T01:20:13.287677+00:00"}