{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:DPRS2Y47ID7PPQJBOGQGGOMOKD","short_pith_number":"pith:DPRS2Y47","canonical_record":{"source":{"id":"2509.01684","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-09-01T18:04:10Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"30c378a0fec91f9bd42c83fde413edc7beff9f387a9232b9329682369310aad6","abstract_canon_sha256":"53f4c596c795b97f8414f877001af9d6e78c46c582dda7918bb385adbf93d75c"},"schema_version":"1.0"},"canonical_sha256":"1be32d639f40fef7c12171a063398e50fd315531c4a5823aed036798601e9036","source":{"kind":"arxiv","id":"2509.01684","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2509.01684","created_at":"2026-07-05T12:04:18Z"},{"alias_kind":"arxiv_version","alias_value":"2509.01684v1","created_at":"2026-07-05T12:04:18Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.01684","created_at":"2026-07-05T12:04:18Z"},{"alias_kind":"pith_short_12","alias_value":"DPRS2Y47ID7P","created_at":"2026-07-05T12:04:18Z"},{"alias_kind":"pith_short_16","alias_value":"DPRS2Y47ID7PPQJB","created_at":"2026-07-05T12:04:18Z"},{"alias_kind":"pith_short_8","alias_value":"DPRS2Y47","created_at":"2026-07-05T12:04:18Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:DPRS2Y47ID7PPQJBOGQGGOMOKD","target":"record","payload":{"canonical_record":{"source":{"id":"2509.01684","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-09-01T18:04:10Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"30c378a0fec91f9bd42c83fde413edc7beff9f387a9232b9329682369310aad6","abstract_canon_sha256":"53f4c596c795b97f8414f877001af9d6e78c46c582dda7918bb385adbf93d75c"},"schema_version":"1.0"},"canonical_sha256":"1be32d639f40fef7c12171a063398e50fd315531c4a5823aed036798601e9036","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:04:18.849162Z","signature_b64":"tJ60Gv8aSW2AszF1uA7lI23o4vwCiEfwO3D9FE341LPC2AEK/E18KVwh993BI3wCwpE6zWmgQYoZ/S/jzCRBBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1be32d639f40fef7c12171a063398e50fd315531c4a5823aed036798601e9036","last_reissued_at":"2026-07-05T12:04:18.848598Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:04:18.848598Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2509.01684","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T12:04:18Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"78/t9+OsWCU0RTtWc+rGJ/dHIshRpOJcLld12s2ljWifNZjuSc8OmMcWPBzpeu83HxdkELEigo4/1IV9QH6vCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T23:10:16.169838Z"},"content_sha256":"5a9bae491c2b2c11503fbb7f0ff9b4d6df522c8caa221c9a2a2ecadfc4946045","schema_version":"1.0","event_id":"sha256:5a9bae491c2b2c11503fbb7f0ff9b4d6df522c8caa221c9a2a2ecadfc4946045"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:DPRS2Y47ID7PPQJBOGQGGOMOKD","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Reinforcement Learning for Machine Learning Engineering Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Joy He-Yueya, Percy Liang, Sherry Yang","submitted_at":"2025-09-01T18:04:10Z","abstract_excerpt":"Existing agents for solving tasks such as ML engineering rely on prompting powerful language models. As a result, these agents do not improve with more experience. In this paper, we show that agents backed by weaker models that improve via reinforcement learning (RL) can outperform agents backed by much larger, but static models. We identify two major challenges with RL in this setting. First, actions can take a variable amount of time (e.g., executing code for different solutions), which leads to asynchronous policy gradient updates that favor faster but suboptimal solutions. To tackle variab"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.01684","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/2509.01684/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T12:04:18Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"/oCLvlLlY1lNoHQJBmCIp93LOHSkXDw1YlKs4/q1tRTD9gGNSWkzKKGp+Mo52F3oaEXlSg0AJ9fJ+95PDga+Bw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T23:10:16.170739Z"},"content_sha256":"f9e8b3ee4affeb998ad28e1e4ac97f54704baeb7c60e191e4a8f3e73a362c4d3","schema_version":"1.0","event_id":"sha256:f9e8b3ee4affeb998ad28e1e4ac97f54704baeb7c60e191e4a8f3e73a362c4d3"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/DPRS2Y47ID7PPQJBOGQGGOMOKD/bundle.json","state_url":"https://pith.science/pith/DPRS2Y47ID7PPQJBOGQGGOMOKD/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/DPRS2Y47ID7PPQJBOGQGGOMOKD/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-05T23:10:16Z","links":{"resolver":"https://pith.science/pith/DPRS2Y47ID7PPQJBOGQGGOMOKD","bundle":"https://pith.science/pith/DPRS2Y47ID7PPQJBOGQGGOMOKD/bundle.json","state":"https://pith.science/pith/DPRS2Y47ID7PPQJBOGQGGOMOKD/state.json","well_known_bundle":"https://pith.science/.well-known/pith/DPRS2Y47ID7PPQJBOGQGGOMOKD/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:DPRS2Y47ID7PPQJBOGQGGOMOKD","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"53f4c596c795b97f8414f877001af9d6e78c46c582dda7918bb385adbf93d75c","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-09-01T18:04:10Z","title_canon_sha256":"30c378a0fec91f9bd42c83fde413edc7beff9f387a9232b9329682369310aad6"},"schema_version":"1.0","source":{"id":"2509.01684","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2509.01684","created_at":"2026-07-05T12:04:18Z"},{"alias_kind":"arxiv_version","alias_value":"2509.01684v1","created_at":"2026-07-05T12:04:18Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.01684","created_at":"2026-07-05T12:04:18Z"},{"alias_kind":"pith_short_12","alias_value":"DPRS2Y47ID7P","created_at":"2026-07-05T12:04:18Z"},{"alias_kind":"pith_short_16","alias_value":"DPRS2Y47ID7PPQJB","created_at":"2026-07-05T12:04:18Z"},{"alias_kind":"pith_short_8","alias_value":"DPRS2Y47","created_at":"2026-07-05T12:04:18Z"}],"graph_snapshots":[{"event_id":"sha256:f9e8b3ee4affeb998ad28e1e4ac97f54704baeb7c60e191e4a8f3e73a362c4d3","target":"graph","created_at":"2026-07-05T12:04:18Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2509.01684/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Existing agents for solving tasks such as ML engineering rely on prompting powerful language models. As a result, these agents do not improve with more experience. In this paper, we show that agents backed by weaker models that improve via reinforcement learning (RL) can outperform agents backed by much larger, but static models. We identify two major challenges with RL in this setting. First, actions can take a variable amount of time (e.g., executing code for different solutions), which leads to asynchronous policy gradient updates that favor faster but suboptimal solutions. To tackle variab","authors_text":"Joy He-Yueya, Percy Liang, Sherry Yang","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-09-01T18:04:10Z","title":"Reinforcement Learning for Machine Learning Engineering Agents"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.01684","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:5a9bae491c2b2c11503fbb7f0ff9b4d6df522c8caa221c9a2a2ecadfc4946045","target":"record","created_at":"2026-07-05T12:04:18Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"53f4c596c795b97f8414f877001af9d6e78c46c582dda7918bb385adbf93d75c","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-09-01T18:04:10Z","title_canon_sha256":"30c378a0fec91f9bd42c83fde413edc7beff9f387a9232b9329682369310aad6"},"schema_version":"1.0","source":{"id":"2509.01684","kind":"arxiv","version":1}},"canonical_sha256":"1be32d639f40fef7c12171a063398e50fd315531c4a5823aed036798601e9036","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"1be32d639f40fef7c12171a063398e50fd315531c4a5823aed036798601e9036","first_computed_at":"2026-07-05T12:04:18.848598Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T12:04:18.848598Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"tJ60Gv8aSW2AszF1uA7lI23o4vwCiEfwO3D9FE341LPC2AEK/E18KVwh993BI3wCwpE6zWmgQYoZ/S/jzCRBBw==","signature_status":"signed_v1","signed_at":"2026-07-05T12:04:18.849162Z","signed_message":"canonical_sha256_bytes"},"source_id":"2509.01684","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:5a9bae491c2b2c11503fbb7f0ff9b4d6df522c8caa221c9a2a2ecadfc4946045","sha256:f9e8b3ee4affeb998ad28e1e4ac97f54704baeb7c60e191e4a8f3e73a362c4d3"],"state_sha256":"d3b1c692cfd00eb919779ce56acbd5a80222907bc9a9197f237195c80dd97160"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"IYO/cFDs8fw5v0sfpmK2993+dyfs5A3lGBd0W23cl8BZ0b1HTHEAjbXLjp3cNHY6cIj6Bp7XQrP41aczXf+2CQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-05T23:10:16.176788Z","bundle_sha256":"52922bf3ae24fff0fa5aa00e5fbb17b24f5b09e456ca78e8ed07c74baa22f471"}}