{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:F6QCW7GLFJAMDP5ZIWUITM533H","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":"7853f28c755e217f94a5d2a661ff5d73c45763cba9f4bf99300ba3cf39976d92","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-16T00:29:35Z","title_canon_sha256":"86dbf4174048556332b84ef277c240c97dde3a1c50fbb125074d6e5dc82c2512"},"schema_version":"1.0","source":{"id":"2605.16727","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.16727","created_at":"2026-05-20T00:02:38Z"},{"alias_kind":"arxiv_version","alias_value":"2605.16727v1","created_at":"2026-05-20T00:02:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.16727","created_at":"2026-05-20T00:02:38Z"},{"alias_kind":"pith_short_12","alias_value":"F6QCW7GLFJAM","created_at":"2026-05-20T00:02:38Z"},{"alias_kind":"pith_short_16","alias_value":"F6QCW7GLFJAMDP5Z","created_at":"2026-05-20T00:02:38Z"},{"alias_kind":"pith_short_8","alias_value":"F6QCW7GL","created_at":"2026-05-20T00:02:38Z"}],"graph_snapshots":[{"event_id":"sha256:612007736c27d7656d5f43bf6c9be1f8a0729de255ba40dc95f9dcfbfcea6901","target":"graph","created_at":"2026-05-20T00:02:38Z","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":"Despite lower training-time reward, the population mean outperforms the baseline on three code benchmarks (HumanEval+, MBPP+, LiveCodeBench) and seven math benchmarks (AIME 24/25, AMC 23, MATH-500, Minerva, GSM8K, OlympiadBench), and even the weakest member of the population beats the baseline on aggregate."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That cross-evaluation between sub-populations reliably prevents self-calibration and sustains an expanding problem-space arms race rather than allowing the population to converge on a narrow set of solvable problems."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"PopuLoRA shows that co-evolving populations of LoRA adapters through cross-evaluated self-play can outperform compute-matched single-agent baselines on multiple code and math reasoning benchmarks."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"A population of specialized LoRA adapters in asymmetric self-play creates a co-evolutionary arms race that improves LLM reasoning over single-agent baselines."}],"snapshot_sha256":"68213903ef3093e6a4ac822c77750a3aeb6726f19f02e3c7c6752ad8021cf605"},"formal_canon":{"evidence_count":1,"snapshot_sha256":"2facb1975cfb3e53f1f86762d6dc880fdb2895583df84123cd1954a66a89ed47"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T22:01:23.082836Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T21:41:13.976443Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-19T19:01:56.346329Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T18:33:26.471815Z","status":"skipped","version":"1.0.0"}],"endpoint":"/pith/2605.16727/integrity.json","findings":[],"snapshot_sha256":"0d052aa4fd4e5a29937f12b459e123d0d37810d7f4d92c469629f916a12185cd","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We introduce PopuLoRA, a population-based asymmetric self-play framework for reinforcement learning with verifiable rewards (RLVR) post-training of LLMs. Teachers and students are specialised LoRA adapters on a shared frozen base: teachers propose problems, matched students solve them under a programmatic verifier, and cross-evaluation between sub-populations replaces the self-calibration that limits single-agent self-play. A family of LoRA weight-space evolution operators (mutations and crossovers that produce same-rank population members in seconds) serves as the replacement step of a popula","authors_text":"Augustine N. Mavor-Parker, Geoffrey Bradway, Lorenz Wolf, Matthew James Sargent, Maxwill Lin, Roger Creus Castanyer","cross_cats":[],"headline":"A population of specialized LoRA adapters in asymmetric self-play creates a co-evolutionary arms race that improves LLM reasoning over single-agent baselines.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-16T00:29:35Z","title":"PopuLoRA: Co-Evolving LLM Populations for Reasoning Self-Play"},"references":{"count":67,"internal_anchors":16,"resolved_work":67,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Vinyals, Oriol and Babuschkin, Igor and Czarnecki, Wojciech M. and Mathieu, Micha. Grandmaster Level in. Nature , volume =. 2019 , doi =","work_id":"ee92cbd6-af5b-4ba7-a386-fa02f0285ffa","year":2019},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"International Conference on Learning Representations , year =","work_id":"bedd69d1-1a3c-4eaf-905b-ceb510e077e2","year":null},{"cited_arxiv_id":"2402.03300","doi":"10.48550/arxiv.2402.03300","is_internal_anchor":true,"ref_index":3,"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","year":null},{"cited_arxiv_id":"1707.06347","doi":"10.48550/arxiv.1707.06347","is_internal_anchor":true,"ref_index":4,"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","year":2017},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Machine Learning , volume =","work_id":"71efaacf-fbce-4277-80c9-97d1cec481a0","year":1992}],"snapshot_sha256":"b763dc216a8b4e8da68cb3ff19d43fbbf8d86d182a0c637715d86df12d565dba"},"source":{"id":"2605.16727","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-19T21:33:52.891464Z","id":"da00c2b2-9610-46cd-8978-5de6a7220c0e","model_set":{"reader":"grok-4.3"},"one_line_summary":"PopuLoRA shows that co-evolving populations of LoRA adapters through cross-evaluated self-play can outperform compute-matched single-agent baselines on multiple code and math reasoning benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"A population of specialized LoRA adapters in asymmetric self-play creates a co-evolutionary arms race that improves LLM reasoning over single-agent baselines.","strongest_claim":"Despite lower training-time reward, the population mean outperforms the baseline on three code benchmarks (HumanEval+, MBPP+, LiveCodeBench) and seven math benchmarks (AIME 24/25, AMC 23, MATH-500, Minerva, GSM8K, OlympiadBench), and even the weakest member of the population beats the baseline on aggregate.","weakest_assumption":"That cross-evaluation between sub-populations reliably prevents self-calibration and sustains an expanding problem-space arms race rather than allowing the population to converge on a narrow set of solvable problems."}},"verdict_id":"da00c2b2-9610-46cd-8978-5de6a7220c0e"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:4ed265096500155b3a92969653853e5805b3f4f2bd6240f737b3e27bb5e442c0","target":"record","created_at":"2026-05-20T00:02:38Z","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":"7853f28c755e217f94a5d2a661ff5d73c45763cba9f4bf99300ba3cf39976d92","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-16T00:29:35Z","title_canon_sha256":"86dbf4174048556332b84ef277c240c97dde3a1c50fbb125074d6e5dc82c2512"},"schema_version":"1.0","source":{"id":"2605.16727","kind":"arxiv","version":1}},"canonical_sha256":"2fa02b7ccb2a40c1bfb945a889b3bbd9c0a801c5b32458052d00a67bb04eb446","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2fa02b7ccb2a40c1bfb945a889b3bbd9c0a801c5b32458052d00a67bb04eb446","first_computed_at":"2026-05-20T00:02:38.695064Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:02:38.695064Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"xPuYyBEKSX+lykzgGPdATptz2kVprwUcVBpdfT+Cjy6ZJBnzuKoJwQLa/HSHI/C9E5gq0DoQjULwdHNrudB9Dw==","signature_status":"signed_v1","signed_at":"2026-05-20T00:02:38.695932Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.16727","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:4ed265096500155b3a92969653853e5805b3f4f2bd6240f737b3e27bb5e442c0","sha256:612007736c27d7656d5f43bf6c9be1f8a0729de255ba40dc95f9dcfbfcea6901"],"state_sha256":"da49aaceea825c8772b422b9e18669533afe32b91577033a09586aafe5275e19"}