{"_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_c7166b58c6ee6ed9f13947e72b3eb254b63fe26871fa089362310dd965a45623","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_c7166b58c6ee6ed9f13947e72b3eb254b63fe26871fa089362310dd965a45623","axiom_type":"AX.OBS","body":{"axiom_subtype":"news.vendor_press.v1","category":"news","fingerprint":"e476bca93778bbd9a8a6cf126a40058bcbca4ed16f9401373ce6477fde2b1756","published":"Fri, 22 May 2026 00:00:00 -0400","receipt_hash":"e476bca93778bbd9a8a6cf126a40058bcbca4ed16f9401373ce6477fde2b1756","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":"e476bca93778bbd9a8a6cf126a40058bcbca4ed16f9401373ce6477fde2b1756","observed_at":"2026-05-22T04:43:12.593252Z","parent_run_hash":"dd4d56660d55b2d65dc84dc5e7c8f83487d90da2dd343b5707d6948a3bb0d917","published":"Fri, 22 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:2605.20730v1 Announce Type: cross \nAbstract: In-context learning (ICL) allows large language models (LLMs) to adapt to new tasks through demonstrations, yet it suffers from escalating inference costs as context length increases. While task vectors offer a promising alternative by compressing demonstrations into compact hidden-state representations, their quality has been evaluated only through downstream task accuracy. This indirect criterion provides limited insight into how to design more effective task vector extraction methods. In this paper, we posit that inference using task vectors should align their predictive distribution with that of ICL. To quantify this, we introduce $d_{\\text{NTP}}$, a metric that measures the discrepancy in next-token probabilities between task vector-based and ICL-based inference. Our empirical analysis reveals that $d_{\\text{NTP}}$ serves as a performance proxy, exhibiting a strong negative correlation with downstream accuracy. Motivated by this, ","title":"Distributional Alignment as a Criterion for Designing Task Vectors in In-Context Learning","url":"https://arxiv.org/abs/2605.20730","vendor":"arxiv_cs_ai"},"summary":"arXiv:2605.20730v1 Announce Type: cross \nAbstract: In-context learning (ICL) allows large language models (LLMs) to adapt to new tasks through demonstrations, yet it suffers from escalating inference costs as context length increases. While task vectors offer a promising alternative by compressing demonstrations into compact hidden-state representations, their quality has been evaluated only through downstream task accuracy. This indirect criterion provides limited insight into how to design more effective task vector extraction methods. In this paper, we posit that inference using task vectors should align their predictive distribution with that of ICL. To quantify this, we introduce $d_{\\text{NTP}}$, a metric that measures the discrepancy in next-token probabilities between task vector-based and ICL-based inference. Our empirical analysis reveals that $d_{\\text{NTP}}$ serves as a performance proxy, exhibiting a strong negative correlation with downstream accuracy. Motivated by this, ","title":"Distributional Alignment as a Criterion for Designing Task Vectors in In-Context Learning","vendor":"arxiv_cs_ai"},"confidence":{"method":"deterministic"},"decision":"POSITIVE","issued_at":"2026-05-22T04:43:12Z","notes":"Spider vendor_press (news) news.vendor_press.v1","object":{"captured_by":"crovia.spider.vendor_press","primary_source_url":"https://arxiv.org/abs/2605.20730"},"predecessors":[],"schema":"crovia.axiom.v1","signature":"ed25519:f7f69d4e81f248a36026acabbcc6b6778846189cef9d5265db30546e241ade83104f051774891715da12efdd65d5dabb73fa473e117f909497b9e1ce08b3850e","signer":"crovia.substrate","subject":{"observed_at":"2026-05-22T04:43:12Z","source_collector":"spider:vendor_press","target_id":"https://arxiv.org/abs/2605.20730"},"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":"01a3823e11664710fadd9b4e9e1ae0f1001cdd309be93e4f75d3dc76d9356e47","leaf_index":148268,"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":"54acbbb2c944b0f77785630dcf03b24a344f6fa67ef948b504f8204294d82b30","side":"right"},{"sibling":"20a70cf39eb9d6a394b3ee86545dfa08c4232e99c3ebe3f9be3abb17c6f99203","side":"right"},{"sibling":"140ac88b5f5c59030c808618b7ae1f3ba0cdc211aa9c7de32f6ae934836329cf","side":"left"},{"sibling":"5d8981a297da8fdc66fa26150948adc6e0a43989ed7222d0ab3cefdd2e897b19","side":"left"},{"sibling":"a85323a9dec6abf79196fd1bef8c923e06bdbdca7aa50a79c7e961497b94c3b4","side":"right"},{"sibling":"0e7e1f05ee2c16785b52fffc9d8a81fa99a3ecbade13700862d5e65f6d4b80a2","side":"left"},{"sibling":"2f67ce1ca4d405b588b0534baed360bc76fbba6b3678854914e0889a47207afb","side":"right"},{"sibling":"0fa191199f785899b5cc0f48838a3ffb2aa741ddcd9510fe1664869d36a8d31a","side":"right"},{"sibling":"e6aaada16cbe58b790855c6723700fd07821cc6df59048dc470d404b1ba60842","side":"left"},{"sibling":"d337a9fdcfc121e9691d8db9173af6a3fe0c33d4a6d5f0c8a7a01af04f9fb856","side":"left"},{"sibling":"8b39e07457f5cc5d687d2ae42284dbe705bb87084e7db4b626aff81e51dacd19","side":"right"},{"sibling":"79a713e1e345ccb99c5fe994a11708c8e9bcfa2e91f940d70421cb7d8d77ecc6","side":"right"},{"sibling":"249870fb494bef050c409081e5de45f9042938d7b2823524ee496f296aa63667","side":"right"},{"sibling":"96c48ee8328f1b7925a4cc4421df5cb0bd81c92a5d8354c93126fa5f0166d225","side":"right"},{"sibling":"35ca36cee447f0ef7064a25d55f59357c66901e31427729c4c1d8b14aa8adb6c","side":"left"},{"sibling":"4e13b4a3e69bb83d13913477a782913ce03937edd65046853c1964d3cbb6564b","side":"right"},{"sibling":"0f7b2df1c4580bf7bb7c24b9158ba20593a06af18a5f18d0973e5eff20c35cd8","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":148601,"merkle_root":"44900cffd986f40535c83f46a250e86fbf1019d41f27080a00fbf9b8d77ec33a","public_key_hex":"cf742e26f75669dc673cb5c0786a1ae23ae8ca19c347317192ce40c28a7ff25c","run_id":"hourly_json_retrofit_20260524T133701Z","schema":"crovia.seal.v1","seal_family_version":"crovia-seal-family/1","seal_kind":"substrate_batch","sealed_at":"2026-05-24T13:37:32Z","sig_algorithm":"ed25519","signature":"059e428c3c5241de303721ad6ac7b748758372180f3f0a717810312aecd6fab073abb3ea264157666de581a2b361c4c6e2aa8ca081b08ce9be090721f0e3400e","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_c7166b58c6ee6ed9f13947e72b3eb254b63fe26871fa089362310dd965a45623"}}