{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:WJA6YFXLX2N6HKH6R63RJH7IFL","short_pith_number":"pith:WJA6YFXL","canonical_record":{"source":{"id":"2605.08696","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-09T05:07:55Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"5cbab1c908ea664ea4ff0043f88adf9a69b600dd9686e2ee6a27b00510649139","abstract_canon_sha256":"7e92b75adf681edfe482f7f5acb99e9eb94aa2f8b206a21ff666d0a158e62d8e"},"schema_version":"1.0"},"canonical_sha256":"b241ec16ebbe9be3a8fe8fb7149fe82ae20d57b4d6c3f4fff95c03f6c22c370a","source":{"kind":"arxiv","id":"2605.08696","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.08696","created_at":"2026-05-20T01:05:16Z"},{"alias_kind":"arxiv_version","alias_value":"2605.08696v2","created_at":"2026-05-20T01:05:16Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.08696","created_at":"2026-05-20T01:05:16Z"},{"alias_kind":"pith_short_12","alias_value":"WJA6YFXLX2N6","created_at":"2026-05-20T01:05:16Z"},{"alias_kind":"pith_short_16","alias_value":"WJA6YFXLX2N6HKH6","created_at":"2026-05-20T01:05:16Z"},{"alias_kind":"pith_short_8","alias_value":"WJA6YFXL","created_at":"2026-05-20T01:05:16Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:WJA6YFXLX2N6HKH6R63RJH7IFL","target":"record","payload":{"canonical_record":{"source":{"id":"2605.08696","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-09T05:07:55Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"5cbab1c908ea664ea4ff0043f88adf9a69b600dd9686e2ee6a27b00510649139","abstract_canon_sha256":"7e92b75adf681edfe482f7f5acb99e9eb94aa2f8b206a21ff666d0a158e62d8e"},"schema_version":"1.0"},"canonical_sha256":"b241ec16ebbe9be3a8fe8fb7149fe82ae20d57b4d6c3f4fff95c03f6c22c370a","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T01:05:16.082626Z","signature_b64":"lglLOxDNLhTEx5dQNBANps5N5KLmjvZ9kPeCGJYsCI4hW6OTZ7juhxZL/oFhWprEnZX71O98hM69+2UaCwTtDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b241ec16ebbe9be3a8fe8fb7149fe82ae20d57b4d6c3f4fff95c03f6c22c370a","last_reissued_at":"2026-05-20T01:05:16.081779Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T01:05:16.081779Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.08696","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T01:05:16Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"RZ+amfM8QGltlJNIyDiZXLE6k5vf/dAlWbTvqwXAIvMRchrhVeTR+U6zk+GcNMwE4l+rClxpta1U34Y56yV0Ag==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-21T09:42:06.877542Z"},"content_sha256":"4540f1017a20aa6074cc4227191171a34e15c0dd7e258f5443b6c48a0682bf1a","schema_version":"1.0","event_id":"sha256:4540f1017a20aa6074cc4227191171a34e15c0dd7e258f5443b6c48a0682bf1a"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:WJA6YFXLX2N6HKH6R63RJH7IFL","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Structured Recurrent Mixers for Massively Parallelized Sequence Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"The Structured Recurrent Mixer enables algebraic conversion between parallel training and recurrent inference representations.","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Benjamin L. Badger","submitted_at":"2026-05-09T05:07:55Z","abstract_excerpt":"Over the last two decades, language modeling has experienced a shift from the use of predominantly recurrent architectures that process tokens sequentially during training and inference to non-recurrent models that process sequence elements in parallel during training, which results in greater training efficiency and stability at the expense of lower inference throughput. Here we introduce the Structured Recurrent Mixer, an architecture that allows for algebraic conversion between a sequence parallel representation at train time and a recurrent representation at inference, notably without the "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We show experimentally that this dual representation allows for greater training efficiency, higher input information capacity, and larger inference throughput and concurrency when compared to other linear complexity models. We provide Mojo/MAX inference implementations of SRMs exhibiting 12x the throughput and 170x the concurrency of similarly powerful Transformers inferenced on vLLM, increases characteristic of Pytorch implementations resulting in a 30% increase in compute-constant GSM8k Pass@k.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The algebraic conversion between parallel and recurrent representations preserves full model capacity and performance without information loss or the need for device-specific optimizations, and that experimental comparisons to other models are conducted under equivalent conditions.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Structured Recurrent Mixers enable algebraic switching between parallel training and recurrent inference representations, delivering higher efficiency, information capacity, and throughput than other linear-complexity models.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"The Structured Recurrent Mixer enables algebraic conversion between parallel training and recurrent inference representations.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"167de5f88db1ec1d62421f01e15f30104922e03c6417396d31b094ff25a0d5d2"},"source":{"id":"2605.08696","kind":"arxiv","version":2},"verdict":{"id":"f2b97cbf-0b4e-40c3-97a7-b38020b6aa58","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-12T01:24:07.994150Z","strongest_claim":"We show experimentally that this dual representation allows for greater training efficiency, higher input information capacity, and larger inference throughput and concurrency when compared to other linear complexity models. We provide Mojo/MAX inference implementations of SRMs exhibiting 12x the throughput and 170x the concurrency of similarly powerful Transformers inferenced on vLLM, increases characteristic of Pytorch implementations resulting in a 30% increase in compute-constant GSM8k Pass@k.","one_line_summary":"Structured Recurrent Mixers enable algebraic switching between parallel training and recurrent inference representations, delivering higher efficiency, information capacity, and throughput than other linear-complexity models.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The algebraic conversion between parallel and recurrent representations preserves full model capacity and performance without information loss or the need for device-specific optimizations, and that experimental comparisons to other models are conducted under equivalent conditions.","pith_extraction_headline":"The Structured Recurrent Mixer enables algebraic conversion between parallel training and recurrent inference representations."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.08696/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-19T22:36:03.809794Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T14:31:17.774680Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T10:51:48.364729Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"430b7a7423b357fe81a7995652ffbfd6ee893fd9bd59943b4603e0c68c02f0fe"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"a91b0ece17bbb2418ac58bf8044fa8232ca3f0f1232c1ba651be30f18bb157e8"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"f2b97cbf-0b4e-40c3-97a7-b38020b6aa58"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T01:05:16Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"HbyDS5SciMTFG8XhLYArCYBY0OBGSpOR/Jn8vB6BCLcR+1TLIvi8E3tA67ITr2K24v7rwqwQo22PM1NN0czoCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-21T09:42:06.878242Z"},"content_sha256":"54941683cda440f95c55fa9d4337d5fdd2b9e25c0ba7f612d89e52393a6e6823","schema_version":"1.0","event_id":"sha256:54941683cda440f95c55fa9d4337d5fdd2b9e25c0ba7f612d89e52393a6e6823"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/WJA6YFXLX2N6HKH6R63RJH7IFL/bundle.json","state_url":"https://pith.science/pith/WJA6YFXLX2N6HKH6R63RJH7IFL/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/WJA6YFXLX2N6HKH6R63RJH7IFL/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":"2026-08-21T09:42:06Z","links":{"resolver":"https://pith.science/pith/WJA6YFXLX2N6HKH6R63RJH7IFL","bundle":"https://pith.science/pith/WJA6YFXLX2N6HKH6R63RJH7IFL/bundle.json","state":"https://pith.science/pith/WJA6YFXLX2N6HKH6R63RJH7IFL/state.json","well_known_bundle":"https://pith.science/.well-known/pith/WJA6YFXLX2N6HKH6R63RJH7IFL/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:WJA6YFXLX2N6HKH6R63RJH7IFL","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":"7e92b75adf681edfe482f7f5acb99e9eb94aa2f8b206a21ff666d0a158e62d8e","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-09T05:07:55Z","title_canon_sha256":"5cbab1c908ea664ea4ff0043f88adf9a69b600dd9686e2ee6a27b00510649139"},"schema_version":"1.0","source":{"id":"2605.08696","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.08696","created_at":"2026-05-20T01:05:16Z"},{"alias_kind":"arxiv_version","alias_value":"2605.08696v2","created_at":"2026-05-20T01:05:16Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.08696","created_at":"2026-05-20T01:05:16Z"},{"alias_kind":"pith_short_12","alias_value":"WJA6YFXLX2N6","created_at":"2026-05-20T01:05:16Z"},{"alias_kind":"pith_short_16","alias_value":"WJA6YFXLX2N6HKH6","created_at":"2026-05-20T01:05:16Z"},{"alias_kind":"pith_short_8","alias_value":"WJA6YFXL","created_at":"2026-05-20T01:05:16Z"}],"graph_snapshots":[{"event_id":"sha256:54941683cda440f95c55fa9d4337d5fdd2b9e25c0ba7f612d89e52393a6e6823","target":"graph","created_at":"2026-05-20T01:05:16Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"We show experimentally that this dual representation allows for greater training efficiency, higher input information capacity, and larger inference throughput and concurrency when compared to other linear complexity models. We provide Mojo/MAX inference implementations of SRMs exhibiting 12x the throughput and 170x the concurrency of similarly powerful Transformers inferenced on vLLM, increases characteristic of Pytorch implementations resulting in a 30% increase in compute-constant GSM8k Pass@k."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The algebraic conversion between parallel and recurrent representations preserves full model capacity and performance without information loss or the need for device-specific optimizations, and that experimental comparisons to other models are conducted under equivalent conditions."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Structured Recurrent Mixers enable algebraic switching between parallel training and recurrent inference representations, delivering higher efficiency, information capacity, and throughput than other linear-complexity models."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"The Structured Recurrent Mixer enables algebraic conversion between parallel training and recurrent inference representations."}],"snapshot_sha256":"167de5f88db1ec1d62421f01e15f30104922e03c6417396d31b094ff25a0d5d2"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"a91b0ece17bbb2418ac58bf8044fa8232ca3f0f1232c1ba651be30f18bb157e8"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T22:36:03.809794Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T14:31:17.774680Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T10:51:48.364729Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.08696/integrity.json","findings":[],"snapshot_sha256":"430b7a7423b357fe81a7995652ffbfd6ee893fd9bd59943b4603e0c68c02f0fe","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Over the last two decades, language modeling has experienced a shift from the use of predominantly recurrent architectures that process tokens sequentially during training and inference to non-recurrent models that process sequence elements in parallel during training, which results in greater training efficiency and stability at the expense of lower inference throughput. Here we introduce the Structured Recurrent Mixer, an architecture that allows for algebraic conversion between a sequence parallel representation at train time and a recurrent representation at inference, notably without the ","authors_text":"Benjamin L. Badger","cross_cats":["cs.LG"],"headline":"The Structured Recurrent Mixer enables algebraic conversion between parallel training and recurrent inference representations.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-09T05:07:55Z","title":"Structured Recurrent Mixers for Massively Parallelized Sequence Generation"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.08696","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-12T01:24:07.994150Z","id":"f2b97cbf-0b4e-40c3-97a7-b38020b6aa58","model_set":{"reader":"grok-4.3"},"one_line_summary":"Structured Recurrent Mixers enable algebraic switching between parallel training and recurrent inference representations, delivering higher efficiency, information capacity, and throughput than other linear-complexity models.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"The Structured Recurrent Mixer enables algebraic conversion between parallel training and recurrent inference representations.","strongest_claim":"We show experimentally that this dual representation allows for greater training efficiency, higher input information capacity, and larger inference throughput and concurrency when compared to other linear complexity models. We provide Mojo/MAX inference implementations of SRMs exhibiting 12x the throughput and 170x the concurrency of similarly powerful Transformers inferenced on vLLM, increases characteristic of Pytorch implementations resulting in a 30% increase in compute-constant GSM8k Pass@k.","weakest_assumption":"The algebraic conversion between parallel and recurrent representations preserves full model capacity and performance without information loss or the need for device-specific optimizations, and that experimental comparisons to other models are conducted under equivalent conditions."}},"verdict_id":"f2b97cbf-0b4e-40c3-97a7-b38020b6aa58"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:4540f1017a20aa6074cc4227191171a34e15c0dd7e258f5443b6c48a0682bf1a","target":"record","created_at":"2026-05-20T01:05:16Z","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":"7e92b75adf681edfe482f7f5acb99e9eb94aa2f8b206a21ff666d0a158e62d8e","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-09T05:07:55Z","title_canon_sha256":"5cbab1c908ea664ea4ff0043f88adf9a69b600dd9686e2ee6a27b00510649139"},"schema_version":"1.0","source":{"id":"2605.08696","kind":"arxiv","version":2}},"canonical_sha256":"b241ec16ebbe9be3a8fe8fb7149fe82ae20d57b4d6c3f4fff95c03f6c22c370a","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"b241ec16ebbe9be3a8fe8fb7149fe82ae20d57b4d6c3f4fff95c03f6c22c370a","first_computed_at":"2026-05-20T01:05:16.081779Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T01:05:16.081779Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"lglLOxDNLhTEx5dQNBANps5N5KLmjvZ9kPeCGJYsCI4hW6OTZ7juhxZL/oFhWprEnZX71O98hM69+2UaCwTtDQ==","signature_status":"signed_v1","signed_at":"2026-05-20T01:05:16.082626Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.08696","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:4540f1017a20aa6074cc4227191171a34e15c0dd7e258f5443b6c48a0682bf1a","sha256:54941683cda440f95c55fa9d4337d5fdd2b9e25c0ba7f612d89e52393a6e6823"],"state_sha256":"9f7969afced0c7f2a455d77b2ba9312d2a80e45867594d7a9696cebe79503937"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"63HAz9g8uMoEuvxAq4XPirZxIUmTbTjbWPS4BgdvSBqg1n+oEjJOKFVEDdx28tYGdutI8yur8AdJdo3SPCi9Dw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-21T09:42:06.883356Z","bundle_sha256":"92fc752bb35b58271fd0c5a9aeecbc0f60d57c2dca63c15e0c7d59ed972170b3"}}