{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:4PPQOK5YJVO3TPWVGDUGBWAMXR","short_pith_number":"pith:4PPQOK5Y","schema_version":"1.0","canonical_sha256":"e3df072bb84d5db9bed530e860d80cbc7c8e13565f3bfcd583dfdaba5f8050a7","source":{"kind":"arxiv","id":"2507.11538","version":1},"attestation_state":"computed","paper":{"title":"How Many Instructions Can LLMs Follow at Once?","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Brendan Whiting, Daniel Jaroslawicz, Karime Maamari, Parth Shah","submitted_at":"2025-07-15T17:59:42Z","abstract_excerpt":"Production-grade LLM systems require robust adherence to dozens or even hundreds of instructions simultaneously. However, the instruction-following capabilities of LLMs at high instruction densities have not yet been characterized, as existing benchmarks only evaluate models on tasks with a single or few instructions. We introduce IFScale, a simple benchmark of 500 keyword-inclusion instructions for a business report writing task to measure how instruction-following performance degrades as instruction density increases. We evaluate 20 state-of-the-art models across seven major providers and fi"},"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":"2507.11538","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.AI","submitted_at":"2025-07-15T17:59:42Z","cross_cats_sorted":[],"title_canon_sha256":"506547be22cc3c55f4f6ed7a9dce81b624a1663daef535fd0aabe54ed4e21aea","abstract_canon_sha256":"123f530d48b648a3e993925b5b1c416e803a0e44be24a4f8d4da8a9a51eb7f8f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:37:40.597736Z","signature_b64":"WxUEiORDJFua9EtrYGB7uTM0C+1sXPKLLvi454Xb5fb5vuz9qOdTYXwT3JZYwzECL+I24prre65ihmCW5BdXCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e3df072bb84d5db9bed530e860d80cbc7c8e13565f3bfcd583dfdaba5f8050a7","last_reissued_at":"2026-07-05T11:37:40.597155Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:37:40.597155Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Many Instructions Can LLMs Follow at Once?","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Brendan Whiting, Daniel Jaroslawicz, Karime Maamari, Parth Shah","submitted_at":"2025-07-15T17:59:42Z","abstract_excerpt":"Production-grade LLM systems require robust adherence to dozens or even hundreds of instructions simultaneously. However, the instruction-following capabilities of LLMs at high instruction densities have not yet been characterized, as existing benchmarks only evaluate models on tasks with a single or few instructions. We introduce IFScale, a simple benchmark of 500 keyword-inclusion instructions for a business report writing task to measure how instruction-following performance degrades as instruction density increases. We evaluate 20 state-of-the-art models across seven major providers and fi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.11538","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/2507.11538/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":"2507.11538","created_at":"2026-07-05T11:37:40.597223+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.11538v1","created_at":"2026-07-05T11:37:40.597223+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.11538","created_at":"2026-07-05T11:37:40.597223+00:00"},{"alias_kind":"pith_short_12","alias_value":"4PPQOK5YJVO3","created_at":"2026-07-05T11:37:40.597223+00:00"},{"alias_kind":"pith_short_16","alias_value":"4PPQOK5YJVO3TPWV","created_at":"2026-07-05T11:37:40.597223+00:00"},{"alias_kind":"pith_short_8","alias_value":"4PPQOK5Y","created_at":"2026-07-05T11:37:40.597223+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.07539","citing_title":"Prompt Governance? On Governing Technologies Governed by Natural Language","ref_index":148,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05716","citing_title":"More Is Not Always Better: Cross-Component Interference in LLM Agent Scaffolding","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4PPQOK5YJVO3TPWVGDUGBWAMXR","json":"https://pith.science/pith/4PPQOK5YJVO3TPWVGDUGBWAMXR.json","graph_json":"https://pith.science/api/pith-number/4PPQOK5YJVO3TPWVGDUGBWAMXR/graph.json","events_json":"https://pith.science/api/pith-number/4PPQOK5YJVO3TPWVGDUGBWAMXR/events.json","paper":"https://pith.science/paper/4PPQOK5Y"},"agent_actions":{"view_html":"https://pith.science/pith/4PPQOK5YJVO3TPWVGDUGBWAMXR","download_json":"https://pith.science/pith/4PPQOK5YJVO3TPWVGDUGBWAMXR.json","view_paper":"https://pith.science/paper/4PPQOK5Y","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.11538&json=true","fetch_graph":"https://pith.science/api/pith-number/4PPQOK5YJVO3TPWVGDUGBWAMXR/graph.json","fetch_events":"https://pith.science/api/pith-number/4PPQOK5YJVO3TPWVGDUGBWAMXR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4PPQOK5YJVO3TPWVGDUGBWAMXR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4PPQOK5YJVO3TPWVGDUGBWAMXR/action/storage_attestation","attest_author":"https://pith.science/pith/4PPQOK5YJVO3TPWVGDUGBWAMXR/action/author_attestation","sign_citation":"https://pith.science/pith/4PPQOK5YJVO3TPWVGDUGBWAMXR/action/citation_signature","submit_replication":"https://pith.science/pith/4PPQOK5YJVO3TPWVGDUGBWAMXR/action/replication_record"}},"created_at":"2026-07-05T11:37:40.597223+00:00","updated_at":"2026-07-05T11:37:40.597223+00:00"}