{"_canonicalization":{"envelope_id":"axm_ + sha256(envelope minus {signature, axiom_id, anchors})","envelope_signature":"ed25519(envelope minus {signature, axiom_id})","json":"sort_keys=True, separators=(',',':'), ensure_ascii=False, allow_nan=False, utf-8","leaf_hash":"sha256(0x00 || canonical_json(envelope_full))","seal_signature":"ed25519(seal minus {signature, sig_algorithm})"},"axiom_id":"axm_306b08ff8936f9baae60fc78a8aff3b296c887f33207dfe8d45f8c405d6af526","bitcoin_anchor":{"bitcoin_attestations":[],"calendar_attestations":[],"ots_url":"","stamped_at":"","status":"pending_next_stamp"},"envelope":{"anchors":[{"chain":"crovia.axiom_graph","height":0,"merkle_proof":"spider_vendor_press_v1","root_at_anchor":"spider_vendor_press_v1"}],"axiom_id":"axm_306b08ff8936f9baae60fc78a8aff3b296c887f33207dfe8d45f8c405d6af526","axiom_type":"AX.OBS","body":{"axiom_subtype":"news.vendor_press.v1","category":"news","fingerprint":"c4f8fd8350a3361cf22b8d9bad7521b8518d414f304200d2d5b318f72be3c388","published":"Tue, 07 Jul 2026 00:00:00 -0400","receipt_hash":"c4f8fd8350a3361cf22b8d9bad7521b8518d414f304200d2d5b318f72be3c388","schema":"spider.news.vendor_press.v1","spider":"vendor_press","spider_record":{"axiom_subtype":"news.vendor_press.v1","category":"news","decision_hint":"POSITIVE","envelope_target":"AX.OBS","fingerprint":"c4f8fd8350a3361cf22b8d9bad7521b8518d414f304200d2d5b318f72be3c388","observed_at":"2026-07-07T04:43:08.294902Z","parent_run_hash":"fc40a96e5d33ecc82922806c3ad18de4725d7af03964570396c8af4e48fb5bc1","published":"Tue, 07 Jul 2026 00:00:00 -0400","runtime_version":"0.1.0","schema":"spider.news.vendor_press.v1","source_status":200,"source_url":"https://export.arxiv.org/rss/cs.AI","spider":"vendor_press","summary_excerpt":"arXiv:2607.04728v1 Announce Type: cross \nAbstract: Reinforcement learning (RL) post-training for large language models (LLMs) follows a efficient paradigm of \"rollout then update\", which inevitably results in off-policy training data. To resolve this, Importance sampling (IS) is proposed, while the token-level ratios compound over long sequences, causing severe variance exploded. A natural idea is \"transferring\" these off-policy token into on-policy token, so that the importance scores for correction are unnecessary. Following this idea, we propose Selective Importance Sampling (SIS), which is inspired by rejection sampling. Concretely, SIS implements by viewing off-policy model as proposal distribution, and implement a token-level rejection test: accepted tokens are viewed as on-policy, so that receive unit importance score, while rejected tokens retain the standard IS correction. Our proposed SIS is theoretically proved reducing the gap between token-level and sequence-level off-poli","title":"Turning Off-Policy Tokens On-Policy: A Plug-in Approach for Improving LLM Alignment","url":"https://arxiv.org/abs/2607.04728","vendor":"arxiv_cs_ai"},"summary":"arXiv:2607.04728v1 Announce Type: cross \nAbstract: Reinforcement learning (RL) post-training for large language models (LLMs) follows a efficient paradigm of \"rollout then update\", which inevitably results in off-policy training data. To resolve this, Importance sampling (IS) is proposed, while the token-level ratios compound over long sequences, causing severe variance exploded. A natural idea is \"transferring\" these off-policy token into on-policy token, so that the importance scores for correction are unnecessary. Following this idea, we propose Selective Importance Sampling (SIS), which is inspired by rejection sampling. Concretely, SIS implements by viewing off-policy model as proposal distribution, and implement a token-level rejection test: accepted tokens are viewed as on-policy, so that receive unit importance score, while rejected tokens retain the standard IS correction. Our proposed SIS is theoretically proved reducing the gap between token-level and sequence-level off-poli","title":"Turning Off-Policy Tokens On-Policy: A Plug-in Approach for Improving LLM Alignment","vendor":"arxiv_cs_ai"},"confidence":{"method":"deterministic"},"decision":"POSITIVE","issued_at":"2026-07-07T04:43:08Z","notes":"Spider vendor_press (news) news.vendor_press.v1","object":{"captured_by":"crovia.spider.vendor_press","primary_source_url":"https://arxiv.org/abs/2607.04728"},"predecessors":[],"schema":"crovia.axiom.v1","signature":"ed25519:988c7205be5ead66470027a72e3c88361416797926b2204b19f7f1d208f4035279006b4b75210d965f82d0cedc6c4ce29e5f14542384326855fb43a4833f3208","signer":"crovia.substrate","subject":{"observed_at":"2026-07-07T04:43:08Z","source_collector":"spider:vendor_press","target_id":"https://arxiv.org/abs/2607.04728"},"tsa":{"authority":"crovia.substrate.bootstrap","rfc3161_token":"{\"kind\":\"crovia.bootstrap.tsa\",\"source_jsonl\":\"/opt/crovia/spider/data/news/vendor_press_v1.jsonl\",\"source_seal_merkle_root\":\"spider_vendor_press_v1\",\"upgrade_path\":\"Sessione H \\u2014 OpenTimestamps weekly anchor\"}"},"zk_mode":"clear","zk_proof":null},"ledger":{"leaf_hash":"877d2d8d03e5bd0f38e539822dbacad675665ffe5fb942b6436adead89c47700","leaf_index":289100,"ledger_path":"/opt/crovia/substrate/axiom_ledger.jsonl"},"merkle_proof":{"hash_alg":"sha256","leaf_prefix":"0x00","node_prefix":"0x01","odd_leaf_rule":"duplicate_last","path":[{"sibling":"fe04f7bc7855925f8862746356951e2e4d75214afc93a63830e76c2997e0609b","side":"right"},{"sibling":"38ae25b20330efbb53e9ed3cfb762afe6f06a1e997ec712c8c4767998f366cbc","side":"right"},{"sibling":"7d77ef0aeeb25beb27c64b03088d7528f782ab19c528c1a1579e328ddd270ae0","side":"left"},{"sibling":"6e4e28f38bf015b75aa4807142da99c95fece57c7d38a5ed526fa3b48c47f8fd","side":"left"},{"sibling":"689aa3eed06c2416306377d26d38614a3d616d28953c438bd5f6a6516df018ec","side":"right"},{"sibling":"dedfa0341e330850adec7653f1402ddaf62e87c235be6120547dd8365bf29160","side":"right"},{"sibling":"7aeacee516154284940b318d4c22233fbe3a13ffa98f9f849efdab541f44dc2b","side":"left"},{"sibling":"3b53d97c225577b0a2d53eeb1ca093f0c3441edf0f64e622c325f18db4fae31c","side":"right"},{"sibling":"35d3e8088ee93171aa479055dcd001cfb6d23925114ddc0908531e54298b0d29","side":"left"},{"sibling":"841129c21a7583176cdc7de281cadfe0e00d04673461e199760cd5128d8cc2d5","side":"right"},{"sibling":"19d6dfd29bc47f35fa02e8fe765277ba9cc3e6da5072309f24ebaac5b5f295e3","side":"right"},{"sibling":"8e0ad7889eb2d4b40e5b6c3d8e2eb19d4e202374983f468aa76321823de07a9f","side":"left"},{"sibling":"aae716235efcb893a1f219dbcd5095070d08a497769fc6d50c14976aa26d5750","side":"right"},{"sibling":"a75ab4319e241beeddb1b3f5705febe0422937926c3479923ccfb0b0082fa4e3","side":"left"},{"sibling":"bd04fa605f883bfb2b81510d045b1e85e555a03da3be083619f61384dfe40ff8","side":"left"},{"sibling":"1b72ad8d12164fdf329e7871711be99d8569d140b21f94056e6962da21da9ce1","side":"right"},{"sibling":"5f5109c2bfdcc7a7e70554bba25862e2d7ce86b6b0cd48a72eb66d2eb735f321","side":"right"},{"sibling":"05fd8a05dddb2e7f72bbb5b290ca55c378f1aed709f132277908d9a5f30eb605","side":"right"},{"sibling":"1cecb7f447febd025aac272837c80de218aecc6485d2395a509b2a1f1b9c746e","side":"left"}]},"schema":"crovia.axiom_proof.v1","seal":{"first_collector_run_id":"","first_receipt_hash":"","jsonl_path":"/opt/crovia/substrate/axiom_ledger.jsonl","key_id":"430895f101d38164","last_collector_run_id":"","last_receipt_hash":"","leaf_count":289613,"merkle_root":"dc428b9d9ba248d4f93f63147bf7c700bf5be7f500cec6c3507b9df6e9401601","public_key_hex":"cf742e26f75669dc673cb5c0786a1ae23ae8ca19c347317192ce40c28a7ff25c","run_id":"hourly_json_retrofit_20260707T053701Z","schema":"crovia.seal.v1","seal_family_version":"crovia-seal-family/1","seal_kind":"substrate_batch","sealed_at":"2026-07-07T05:38:15Z","sig_algorithm":"ed25519","signature":"c468b0e183383ab71992be40bda451093e6cd8cd8efb0d26f68e135a804b287c209d12a0f4fdd95c69c835c04b78df8cb1903dee1f53d4730b36f5332a29fe05","signer_version":"1.1.0"},"trust_root":{"key_id":"430895f101d38164","public_key_hex":"cf742e26f75669dc673cb5c0786a1ae23ae8ca19c347317192ce40c28a7ff25c","signature_algorithm":"ed25519","url":"/registry/canon/TRUST_ROOT.md"},"verifier":{"spec":"/registry/canon/AXIOM_RECEIPT_v1.md","url":"/v/axm_306b08ff8936f9baae60fc78a8aff3b296c887f33207dfe8d45f8c405d6af526"}}