{"_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_48041e84bcba8fcce4e584e2aef8095245b6afa8aebff92158d2c92925016dd3","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_48041e84bcba8fcce4e584e2aef8095245b6afa8aebff92158d2c92925016dd3","axiom_type":"AX.OBS","body":{"axiom_subtype":"news.vendor_press.v1","category":"news","fingerprint":"f85854dc754286eaa825d078f9779a9590f8584146868e75058a9d7ea92b1924","published":"Mon, 20 Jul 2026 00:00:00 -0400","receipt_hash":"f85854dc754286eaa825d078f9779a9590f8584146868e75058a9d7ea92b1924","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":"f85854dc754286eaa825d078f9779a9590f8584146868e75058a9d7ea92b1924","observed_at":"2026-07-20T04:43:09.641409Z","parent_run_hash":"0fd83663f0f57da59b26313ca1a35283a3e9f06e3165d3e143d26f7174743aca","published":"Mon, 20 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:2603.26556v2 Announce Type: replace-cross \nAbstract: Converting a pretrained Transformer into a more efficient hybrid model through distillation offers a promising approach to reducing inference costs. However, achieving high-quality generation in distilled models requires careful joint design of both the student architecture and the distillation process. Many prior distillation works evaluate downstream multiple-choice benchmarks by ranking candidate answers with log-likelihood rather than requiring autoregressive generation, which can obscure important differences in model quality. For example, on overlapping benchmarks, we show that a 7B distilled model that nearly matches its teacher to within 0.2 pp under log-likelihood scoring falls behind by 20.8 pp when it must generate answers autoregressively.\n  We investigate this phenomenon with GenDistill, a multi-stage pipeline we designed for distilling a pretrained Transformer into an efficient Hybrid Kimi Delta Attention (Hybrid-","title":"When Perplexity Lies: Generation-Focused Distillation of Hybrid Sequence Models","url":"https://arxiv.org/abs/2603.26556","vendor":"arxiv_cs_ai"},"summary":"arXiv:2603.26556v2 Announce Type: replace-cross \nAbstract: Converting a pretrained Transformer into a more efficient hybrid model through distillation offers a promising approach to reducing inference costs. However, achieving high-quality generation in distilled models requires careful joint design of both the student architecture and the distillation process. Many prior distillation works evaluate downstream multiple-choice benchmarks by ranking candidate answers with log-likelihood rather than requiring autoregressive generation, which can obscure important differences in model quality. For example, on overlapping benchmarks, we show that a 7B distilled model that nearly matches its teacher to within 0.2 pp under log-likelihood scoring falls behind by 20.8 pp when it must generate answers autoregressively.\n  We investigate this phenomenon with GenDistill, a multi-stage pipeline we designed for distilling a pretrained Transformer into an efficient Hybrid Kimi Delta Attention (Hybrid-","title":"When Perplexity Lies: Generation-Focused Distillation of Hybrid Sequence Models","vendor":"arxiv_cs_ai"},"confidence":{"method":"deterministic"},"decision":"POSITIVE","issued_at":"2026-07-20T04:43:09Z","notes":"Spider vendor_press (news) news.vendor_press.v1","object":{"captured_by":"crovia.spider.vendor_press","primary_source_url":"https://arxiv.org/abs/2603.26556"},"predecessors":[],"schema":"crovia.axiom.v1","signature":"ed25519:15e527b5e483127d0c389bce2edb4e492d5e659b253ed919c608f78d53ab1feab13e915697340cc2a741f244b7c94b96577638aa1d306aaeaa18a9bf635f9007","signer":"crovia.substrate","subject":{"observed_at":"2026-07-20T04:43:09Z","source_collector":"spider:vendor_press","target_id":"https://arxiv.org/abs/2603.26556"},"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":"cdee10cad20f744840ea3815a7a5cfd172681515bee4bcbb652703db8a0a6562","leaf_index":333390,"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":"7653288911793ff935d4f989d78488a06e2622aabc1b6e148d861922994632f9","side":"right"},{"sibling":"024809ab98431168553afcd2a13d9e424634f7f3f00e8d8022c3d1bdec1ddb07","side":"left"},{"sibling":"25c7873d04b914326cf27fae151fc8874f2ff5f2943f403f86067b38b9a6856b","side":"left"},{"sibling":"63e22f3318998a551d7a8ff393eb84391039f3a818b0cd9d22d7aaa1d54d6883","side":"left"},{"sibling":"2757cb640ccb0d7ed3b3d1937a5ca6f9ad854fa7f38968d791c6a1931bd07e34","side":"right"},{"sibling":"cafc5529c4d98a0d1cf7ab89a7300fc6c7eaa68c576b882112c3382c257c9b70","side":"right"},{"sibling":"fd7a2f2e44dc3f058d65e59ceb20b3d20eb68b3d8d84208ea601209a4c4df8a9","side":"left"},{"sibling":"5f23b1a0a67d8fa1aa690680510620f249f8d6683d46c202fca306510fbe398e","side":"right"},{"sibling":"f720760992870795e6d2b913ad9162f9e144b821ea97e4dada744d0cac06e93b","side":"right"},{"sibling":"45c0e4431502711514503abd48e8b3d34ee2f4bffa994c883aaf2adecd0ad8e9","side":"left"},{"sibling":"dedd2da92d9447ddf1b1db68fe20109a426ed18359861ed746907ac820021a8d","side":"left"},{"sibling":"a1c43cc7cd9c775fac33940ee5124aece01596733f003fc53743f43483f9f597","side":"right"},{"sibling":"b5ad3eafd7eeb74c063261356fdd9bf6059ee6d0bf1e3c70e60731b394a5536e","side":"left"},{"sibling":"93e399d152203db688c6a5a58d25131205603504f5b79123a1f2b5a5ed9c1e54","side":"right"},{"sibling":"b6e0cad7f6eb9107f0edd276f1a9942635d8cd6d60d2a97e7daac08b110dc209","side":"right"},{"sibling":"80ec062e7e625dc3f9bb5865cb5198696bbec2608e48abae5670677b90695899","side":"right"},{"sibling":"77025bcb374a7ad74f520643e20a8ae1205a7ee78507b0beb93117760f1c29d3","side":"left"},{"sibling":"0aced6f0c9dec3e6cc9e89b68b70f5f8ce7e1eb13606d92db1917b76e57393c7","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":333540,"merkle_root":"ee60f62b8a724dd9bde638d638caf32cefec4440f832018b457ff47a0ec56a8c","public_key_hex":"cf742e26f75669dc673cb5c0786a1ae23ae8ca19c347317192ce40c28a7ff25c","run_id":"hourly_json_retrofit_20260720T053701Z","schema":"crovia.seal.v1","seal_family_version":"crovia-seal-family/1","seal_kind":"substrate_batch","sealed_at":"2026-07-20T05:38:36Z","sig_algorithm":"ed25519","signature":"82a3787e628bfab19c377d875220e1aaedfc498736c545f4708a1b887e8398afdf306995994493a36c864ff7139a2d436b905ce7081aaa540789aa4f707dc800","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_48041e84bcba8fcce4e584e2aef8095245b6afa8aebff92158d2c92925016dd3"}}