{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:2Q7GPC27AYWXVKYKW35LOQCDQX","short_pith_number":"pith:2Q7GPC27","schema_version":"1.0","canonical_sha256":"d43e678b5f062d7aab0ab6fab7404385fe248f65b783aefb69d02cfdf5617436","source":{"kind":"arxiv","id":"2605.31228","version":1},"attestation_state":"computed","paper":{"title":"EchoRL: Reinforcement Learning via Rollout Echoing","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Aniri, Jinhe Bi, Michael F\\\"arber, Minglai Yang, Sikuan Yan, Volker Tresp, Wenke Huang, Xingcheng Zhou, Xun Xiao, Yujun Wang, Yunpu Ma, Zixuan Cao","submitted_at":"2026-05-29T12:31:55Z","abstract_excerpt":"Reinforcement Learning with Verifiable Rewards is an effective route for post-training to strengthen the reasoning capability of large language models. However, as training proceeds, the learning signal can collapse thus makes the training gain become marginal and ineffective. Specifically, a growing fraction of prompts' rollouts become advantage-degenerated: all the self-generated rollouts show verified-success, making the standard deviation over their rewards be zero; accordingly each rollout's advantage becomes degenerated (zero) as well. Given such rollouts' advantages, the policy-gradient"},"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":"2605.31228","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-29T12:31:55Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"dbca9631691a8aaa2149af829a7019cdbcb9ee23193a9d14af3c70571a71ba66","abstract_canon_sha256":"7edcd298a37679da3035684a822b91735bc45a83c852725d9a0937d250a4450d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T01:04:05.116599Z","signature_b64":"F3x9iRzw5BoInOgjqxMHs7tOrwuZUaZw+AcrPqYj6zNOLK7eebZJR2GHtAXowD8jswk96OghPG82AtgSGgvzDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d43e678b5f062d7aab0ab6fab7404385fe248f65b783aefb69d02cfdf5617436","last_reissued_at":"2026-06-01T01:04:05.115819Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T01:04:05.115819Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"EchoRL: Reinforcement Learning via Rollout Echoing","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Aniri, Jinhe Bi, Michael F\\\"arber, Minglai Yang, Sikuan Yan, Volker Tresp, Wenke Huang, Xingcheng Zhou, Xun Xiao, Yujun Wang, Yunpu Ma, Zixuan Cao","submitted_at":"2026-05-29T12:31:55Z","abstract_excerpt":"Reinforcement Learning with Verifiable Rewards is an effective route for post-training to strengthen the reasoning capability of large language models. However, as training proceeds, the learning signal can collapse thus makes the training gain become marginal and ineffective. Specifically, a growing fraction of prompts' rollouts become advantage-degenerated: all the self-generated rollouts show verified-success, making the standard deviation over their rewards be zero; accordingly each rollout's advantage becomes degenerated (zero) as well. Given such rollouts' advantages, the policy-gradient"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.31228","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/2605.31228/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":"2605.31228","created_at":"2026-06-01T01:04:05.115966+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.31228v1","created_at":"2026-06-01T01:04:05.115966+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.31228","created_at":"2026-06-01T01:04:05.115966+00:00"},{"alias_kind":"pith_short_12","alias_value":"2Q7GPC27AYWX","created_at":"2026-06-01T01:04:05.115966+00:00"},{"alias_kind":"pith_short_16","alias_value":"2Q7GPC27AYWXVKYK","created_at":"2026-06-01T01:04:05.115966+00:00"},{"alias_kind":"pith_short_8","alias_value":"2Q7GPC27","created_at":"2026-06-01T01:04:05.115966+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/2Q7GPC27AYWXVKYKW35LOQCDQX","json":"https://pith.science/pith/2Q7GPC27AYWXVKYKW35LOQCDQX.json","graph_json":"https://pith.science/api/pith-number/2Q7GPC27AYWXVKYKW35LOQCDQX/graph.json","events_json":"https://pith.science/api/pith-number/2Q7GPC27AYWXVKYKW35LOQCDQX/events.json","paper":"https://pith.science/paper/2Q7GPC27"},"agent_actions":{"view_html":"https://pith.science/pith/2Q7GPC27AYWXVKYKW35LOQCDQX","download_json":"https://pith.science/pith/2Q7GPC27AYWXVKYKW35LOQCDQX.json","view_paper":"https://pith.science/paper/2Q7GPC27","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.31228&json=true","fetch_graph":"https://pith.science/api/pith-number/2Q7GPC27AYWXVKYKW35LOQCDQX/graph.json","fetch_events":"https://pith.science/api/pith-number/2Q7GPC27AYWXVKYKW35LOQCDQX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2Q7GPC27AYWXVKYKW35LOQCDQX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2Q7GPC27AYWXVKYKW35LOQCDQX/action/storage_attestation","attest_author":"https://pith.science/pith/2Q7GPC27AYWXVKYKW35LOQCDQX/action/author_attestation","sign_citation":"https://pith.science/pith/2Q7GPC27AYWXVKYKW35LOQCDQX/action/citation_signature","submit_replication":"https://pith.science/pith/2Q7GPC27AYWXVKYKW35LOQCDQX/action/replication_record"}},"created_at":"2026-06-01T01:04:05.115966+00:00","updated_at":"2026-06-01T01:04:05.115966+00:00"}