{"_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_8f624195b4d09a6d50a226d64696f03ec924003b2e3a4ed56349818366102d82","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_8f624195b4d09a6d50a226d64696f03ec924003b2e3a4ed56349818366102d82","axiom_type":"AX.OBS","body":{"axiom_subtype":"news.vendor_press.v1","category":"news","fingerprint":"63ad53faa206ec2eebdd1a8f554e28aebcc9058bd1b4f731740f6f0070509851","published":"Tue, 26 May 2026 00:00:00 -0400","receipt_hash":"63ad53faa206ec2eebdd1a8f554e28aebcc9058bd1b4f731740f6f0070509851","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":"63ad53faa206ec2eebdd1a8f554e28aebcc9058bd1b4f731740f6f0070509851","observed_at":"2026-05-26T04:43:39.018238Z","parent_run_hash":"dca8dedd754ad6a1772113d6b97ee4f4ab9a0afbdeb44aaace5ff2d2446b164b","published":"Tue, 26 May 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:2605.23945v1 Announce Type: new \nAbstract: Reinforcement Learning from Human Feedback (RLHF) has become a key post-training paradigm for improving model quality. However, the synchronous three-stage RLHF pipeline is often bottlenecked by the generation stage, where response-length skew causes the effective batch size to shrink rapidly during decoding, leaving GPUs underutilized while a few long responses remain unfinished. Mainstream frameworks employ a static tensor parallelism (TP) configuration that cannot adapt to changing batch characteristics, leaving substantial performance headroom unexplored. We propose PAT, an adaptive TP method that dynamically reconfigures TP during the generation stage of each RLHF iteration. PAT introduces two key techniques. First, a predictor-guided online reconfiguration method decides both the reconfiguration point and the target TP configuration based on offline profiling, triggering reconfiguration only when the predicted latency benefit outwe","title":"Accelerating Long-Tail Generation in Synchronous RLHF Training via Adaptive Tensor Parallelism","url":"https://arxiv.org/abs/2605.23945","vendor":"arxiv_cs_ai"},"summary":"arXiv:2605.23945v1 Announce Type: new \nAbstract: Reinforcement Learning from Human Feedback (RLHF) has become a key post-training paradigm for improving model quality. However, the synchronous three-stage RLHF pipeline is often bottlenecked by the generation stage, where response-length skew causes the effective batch size to shrink rapidly during decoding, leaving GPUs underutilized while a few long responses remain unfinished. Mainstream frameworks employ a static tensor parallelism (TP) configuration that cannot adapt to changing batch characteristics, leaving substantial performance headroom unexplored. We propose PAT, an adaptive TP method that dynamically reconfigures TP during the generation stage of each RLHF iteration. PAT introduces two key techniques. First, a predictor-guided online reconfiguration method decides both the reconfiguration point and the target TP configuration based on offline profiling, triggering reconfiguration only when the predicted latency benefit outwe","title":"Accelerating Long-Tail Generation in Synchronous RLHF Training via Adaptive Tensor Parallelism","vendor":"arxiv_cs_ai"},"confidence":{"method":"deterministic"},"decision":"POSITIVE","issued_at":"2026-05-26T04:43:39Z","notes":"Spider vendor_press (news) news.vendor_press.v1","object":{"captured_by":"crovia.spider.vendor_press","primary_source_url":"https://arxiv.org/abs/2605.23945"},"predecessors":[],"schema":"crovia.axiom.v1","signature":"ed25519:3ddfc9c74953b3785d9807f36780421c9c19d1f71475a5a96f62b0b80f217b287bf467162a33cfdd850fe5d9e42ab31baf0ec21ad395d279c802a2866d32ee08","signer":"crovia.substrate","subject":{"observed_at":"2026-05-26T04:43:39Z","source_collector":"spider:vendor_press","target_id":"https://arxiv.org/abs/2605.23945"},"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":"227e826de2614cafad675a712a3de3b5ee73de2850b1c0701954ec7b808561fc","leaf_index":151308,"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":"4d70029d9982390de866b807c4e398ca91274f69647874068e5184d79494cbd2","side":"right"},{"sibling":"0c19ad19074a9d67e4a6b944cae72c06616006ac1dcd0a9e290d63f73e659efe","side":"right"},{"sibling":"ea7c4eede263c47e6679a39d12b3a4c199f97ed475196a972f6632d0b6712712","side":"left"},{"sibling":"771bfe09e035f93099d9fd26e5edba714d81f21346337c7deee59a59c469c2c6","side":"left"},{"sibling":"99632c966d574be17cbc599e6f8b42aaa8132f345760dc5757870f3918442546","side":"right"},{"sibling":"e1170472132a5bf945fa97169489cdcb989ae469470b68ca0b110228d0ebd22d","side":"right"},{"sibling":"e5bcb64fb5869d64722cefc3be31e82a01b365efdec3937cfb54d8b941368e56","side":"right"},{"sibling":"8d693959d4763b28b351e1bc64d35b03213077740e93193dee0f2ab1f6459eb7","side":"right"},{"sibling":"a808644c9a09dd75253c6c0ba9275dc94a6b6451ad5184ebdafda4a2722b36a5","side":"left"},{"sibling":"fdd22ebd87e5ba37d1153e47753a63818abb25df8fe45dbf735ba5c512c3355b","side":"left"},{"sibling":"cf593c482d17d202b94914915d0c65fb8713053548db64771d0ec5dbb404a07f","side":"left"},{"sibling":"134949308b15cffd6792ee2cf678119af34d69a64764d7c89cd47573c94e1cda","side":"left"},{"sibling":"e20a7391fed5b5f3b675af68344a3f5b050d6701e127b7db010af3941e59dfdb","side":"right"},{"sibling":"e5893793e3591ed7f5e58ca94ffcfba46bb30f69fb1c25d5ba8ef49eb99f9126","side":"right"},{"sibling":"35ca36cee447f0ef7064a25d55f59357c66901e31427729c4c1d8b14aa8adb6c","side":"left"},{"sibling":"e3eecaf996dbe7229a7bb1d234c97aea97a252c5f7c89f8547b6d091db0f0e40","side":"right"},{"sibling":"55bcbd4da3e20d93931f7e58673f10232e81a5b1514d7396cb4b71e8f95788d0","side":"right"},{"sibling":"d841ad93efda0869e5eb97678f348f03f5caab4353e05ff4bf18f47fb945b822","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":152106,"merkle_root":"ac5182c6f3dd09931f2a689df5f4be36df7b55e55bcf106e195671f5ed55fd8f","public_key_hex":"cf742e26f75669dc673cb5c0786a1ae23ae8ca19c347317192ce40c28a7ff25c","run_id":"hourly_json_retrofit_20260526T053701Z","schema":"crovia.seal.v1","seal_family_version":"crovia-seal-family/1","seal_kind":"substrate_batch","sealed_at":"2026-05-26T05:37:34Z","sig_algorithm":"ed25519","signature":"a401243fbd2c077d29a623c6ef616c78fbd5c8ce1930afa7af7165b2386e5a8cf15d5094983a1c962e71b27b911a3f3ce09c8cfab9393be2ce5ce8ee6513da06","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_8f624195b4d09a6d50a226d64696f03ec924003b2e3a4ed56349818366102d82"}}