{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:D4GTU253PYNHL2BF3FNK7WA5QQ","short_pith_number":"pith:D4GTU253","schema_version":"1.0","canonical_sha256":"1f0d3a6bbb7e1a75e825d95aafd81d842b6f823f78e3f5ebb05a3101c2864d45","source":{"kind":"arxiv","id":"2608.06638","version":1},"attestation_state":"computed","paper":{"title":"MI-MIDI: Mechanistic Interpretability of Text-to-MIDI Generation Models via Probing, Lenses and Steering","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.SD","authors_text":"Jakub Po\\'cwiardowski, Mateusz Modrzejewski","submitted_at":"2026-08-06T23:09:37Z","abstract_excerpt":"Mechanistic interpretability of music generation has concentrated on audio models, leaving symbolic models largely unexplored. We analyze two public text-to-MIDI systems of contrasting design: the purpose-built encoder--decoder text2midi and MIDI-LLM, a Llama~3.2~1B model extended with MIDI tokens using linear probing, the logit and tuned lenses, activation patching and difference-in-means steering. Across these methods, we recover musically meaningful structure and show how architecture shapes its formation and control. Pitch, instrumentation, harmony and texture are linearly decodable in bot"},"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":"2608.06638","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.SD","submitted_at":"2026-08-06T23:09:37Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"91a7e75bc8deb35b92de6c5cb01bddbd713301eabdecbf043df44192bf46f888","abstract_canon_sha256":"92a340bbba9cfa89daaa2561605768cca6797ae3414a9df21cf1e797c9f17b43"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-10T01:11:33.385776Z","signature_b64":"9l04y9zyap6shuFkdnMcfXqcqT+tzNjxCPLbvNgwuhArYI4OdZ/uIxOSpZH0oXvDyX813wtfvDDMzKISfdzVBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1f0d3a6bbb7e1a75e825d95aafd81d842b6f823f78e3f5ebb05a3101c2864d45","last_reissued_at":"2026-08-10T01:11:33.383405Z","signature_status":"signed_v1","first_computed_at":"2026-08-10T01:11:33.383405Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MI-MIDI: Mechanistic Interpretability of Text-to-MIDI Generation Models via Probing, Lenses and Steering","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.SD","authors_text":"Jakub Po\\'cwiardowski, Mateusz Modrzejewski","submitted_at":"2026-08-06T23:09:37Z","abstract_excerpt":"Mechanistic interpretability of music generation has concentrated on audio models, leaving symbolic models largely unexplored. We analyze two public text-to-MIDI systems of contrasting design: the purpose-built encoder--decoder text2midi and MIDI-LLM, a Llama~3.2~1B model extended with MIDI tokens using linear probing, the logit and tuned lenses, activation patching and difference-in-means steering. Across these methods, we recover musically meaningful structure and show how architecture shapes its formation and control. Pitch, instrumentation, harmony and texture are linearly decodable in bot"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.06638","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/2608.06638/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":"2608.06638","created_at":"2026-08-10T01:11:33.384389+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.06638v1","created_at":"2026-08-10T01:11:33.384389+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.06638","created_at":"2026-08-10T01:11:33.384389+00:00"},{"alias_kind":"pith_short_12","alias_value":"D4GTU253PYNH","created_at":"2026-08-10T01:11:33.384389+00:00"},{"alias_kind":"pith_short_16","alias_value":"D4GTU253PYNHL2BF","created_at":"2026-08-10T01:11:33.384389+00:00"},{"alias_kind":"pith_short_8","alias_value":"D4GTU253","created_at":"2026-08-10T01:11:33.384389+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/D4GTU253PYNHL2BF3FNK7WA5QQ","json":"https://pith.science/pith/D4GTU253PYNHL2BF3FNK7WA5QQ.json","graph_json":"https://pith.science/api/pith-number/D4GTU253PYNHL2BF3FNK7WA5QQ/graph.json","events_json":"https://pith.science/api/pith-number/D4GTU253PYNHL2BF3FNK7WA5QQ/events.json","paper":"https://pith.science/paper/D4GTU253"},"agent_actions":{"view_html":"https://pith.science/pith/D4GTU253PYNHL2BF3FNK7WA5QQ","download_json":"https://pith.science/pith/D4GTU253PYNHL2BF3FNK7WA5QQ.json","view_paper":"https://pith.science/paper/D4GTU253","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.06638&json=true","fetch_graph":"https://pith.science/api/pith-number/D4GTU253PYNHL2BF3FNK7WA5QQ/graph.json","fetch_events":"https://pith.science/api/pith-number/D4GTU253PYNHL2BF3FNK7WA5QQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/D4GTU253PYNHL2BF3FNK7WA5QQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/D4GTU253PYNHL2BF3FNK7WA5QQ/action/storage_attestation","attest_author":"https://pith.science/pith/D4GTU253PYNHL2BF3FNK7WA5QQ/action/author_attestation","sign_citation":"https://pith.science/pith/D4GTU253PYNHL2BF3FNK7WA5QQ/action/citation_signature","submit_replication":"https://pith.science/pith/D4GTU253PYNHL2BF3FNK7WA5QQ/action/replication_record"}},"created_at":"2026-08-10T01:11:33.384389+00:00","updated_at":"2026-08-10T01:11:33.384389+00:00"}