{"_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_3af5361f7ebaa87d5eac835776075681e291294e51f1c8b5c7ab003c0d6b2da8","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_arxiv_retraction_v1","root_at_anchor":"spider_arxiv_retraction_v1"}],"axiom_id":"axm_3af5361f7ebaa87d5eac835776075681e291294e51f1c8b5c7ab003c0d6b2da8","axiom_type":"AX.OBS","body":{"axiom_subtype":"research.arxiv_retraction.v1","category":"research","fingerprint":"a60ae57e9bd064b0d05d1010c41faf13a3e79a1e5bf95932876500da820f1acb","published":"2026-02-26T01:54:24Z","receipt_hash":"a60ae57e9bd064b0d05d1010c41faf13a3e79a1e5bf95932876500da820f1acb","schema":"spider.research.arxiv_retraction.v1","spider":"arxiv_retraction","spider_record":{"axiom_subtype":"research.arxiv_retraction.v1","category":"research","decision_hint":"POSITIVE","envelope_target":"AX.OBS","fingerprint":"a60ae57e9bd064b0d05d1010c41faf13a3e79a1e5bf95932876500da820f1acb","observed_at":"2026-07-24T05:55:08.689522Z","parent_run_hash":"ca183106b0cf390894301ca030e66433437207e2af690031ecc20ccf9f077294","published":"2026-02-26T01:54:24Z","runtime_version":"0.1.0","schema":"spider.research.arxiv_retraction.v1","source_status":200,"source_url":"http://export.arxiv.org/api/query?search_query=cat:cs.AI+AND+%28abs:withdrawn+OR+abs:retracted%29&max_results=20&sortBy=submittedDate&sortOrder=descending","spider":"arxiv_retraction","summary_excerpt":"Test-Time Training (TTT) language models replace the KV-cache with fast weights updated during inference, achieving O(1) memory but suffering catastrophic failure on exact-recall tasks. Version 1 of this work proposed SR-TTT, which routes high-surprisal tokens to a sparse exact-attention Residual Cache, and reported large Needle-in-a-Haystack gains. We show those gains were evaluation artifacts: the loss and metric read logits at the answer positions rather than one position earlier, training both models to copy an answer already visible in their input (a model trained on retrieval-impossible data reaches 100% accuracy under the flawed metric); additionally, the cache attended non-causally over future tokens, including the answer itself. We release a corrected implementation with startup causality self-tests, then ask whether the SR-TTT hypothesis survives correction. It does not, and the failure decomposes into two independent, separately measured bottlenecks. Storage: surprisal gatin","title":"SR-TTT Does Not Learn Retrieval: A Correction and Mechanistic Post-Mortem of Surprisal-Aware Residual Test-Time Training","url":"https://arxiv.org/pdf/2603.06642v2","vendor":"arxiv"},"summary":"Test-Time Training (TTT) language models replace the KV-cache with fast weights updated during inference, achieving O(1) memory but suffering catastrophic failure on exact-recall tasks. Version 1 of this work proposed SR-TTT, which routes high-surprisal tokens to a sparse exact-attention Residual Cache, and reported large Needle-in-a-Haystack gains. We show those gains were evaluation artifacts: the loss and metric read logits at the answer positions rather than one position earlier, training both models to copy an answer already visible in their input (a model trained on retrieval-impossible data reaches 100% accuracy under the flawed metric); additionally, the cache attended non-causally over future tokens, including the answer itself. We release a corrected implementation with startup causality self-tests, then ask whether the SR-TTT hypothesis survives correction. It does not, and the failure decomposes into two independent, separately measured bottlenecks. Storage: surprisal gatin","title":"SR-TTT Does Not Learn Retrieval: A Correction and Mechanistic Post-Mortem of Surprisal-Aware Residual Test-Time Training","vendor":"arxiv"},"confidence":{"method":"deterministic"},"decision":"POSITIVE","issued_at":"2026-07-24T05:55:08Z","notes":"Spider arxiv_retraction (research) research.arxiv_retraction.v1","object":{"captured_by":"crovia.spider.arxiv_retraction","primary_source_url":"https://arxiv.org/pdf/2603.06642v2"},"predecessors":[],"schema":"crovia.axiom.v1","signature":"ed25519:970a863bad67a0b9efe2a108e82137240a98a18c3fff5ef387ab1fc410b85bdf5349748c776228b2499d458b71598b175a7a3ef7e2502f132e8b8fdf9319850d","signer":"crovia.substrate","subject":{"observed_at":"2026-07-24T05:55:08Z","source_collector":"spider:arxiv_retraction","target_id":"https://arxiv.org/pdf/2603.06642v2"},"tsa":{"authority":"crovia.substrate.bootstrap","rfc3161_token":"{\"kind\":\"crovia.bootstrap.tsa\",\"source_jsonl\":\"/opt/crovia/spider/data/research/arxiv_retraction_v1.jsonl\",\"source_seal_merkle_root\":\"spider_arxiv_retraction_v1\",\"upgrade_path\":\"Sessione H \\u2014 OpenTimestamps weekly anchor\"}"},"zk_mode":"clear","zk_proof":null},"ledger":{"leaf_hash":"edc0e87f46dbada272ce56d6705ddd0bf7498f26e32accebddff08f2d7970735","leaf_index":347518,"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":"6db9633b8be89e612fbce6ab68f7c8779c0c41e1d82c6793643064bf87463ded","side":"right"},{"sibling":"8e5e3008959d1547d97dd322d82b940505da5eb28e795ef4e476f3b16ba96c8a","side":"left"},{"sibling":"15c80c8cbe9374c6b41dff21079ea3f5f253f99f94b64c4512bbc0872df959de","side":"left"},{"sibling":"4a22814739823b6599d92476aff00713fd09cf78e5128838d95b88cd53a42af5","side":"left"},{"sibling":"d43f13e5e692221aa0a2afcff8e098448e5ac516ea03b248588fbf732e21c933","side":"left"},{"sibling":"a546d636b2b895dd2c7308f3c4c73e30e58a5ce9b047d3c1b3628c49114c4b71","side":"left"},{"sibling":"e5cf65492c62edba310f536b0b50f1dc8ede43fda638bf2d3fc5b75746819ff1","side":"left"},{"sibling":"f2f68c5d228459548da499339a15d31c060ead543526386ba2c8e6aa9c21460c","side":"right"},{"sibling":"a1d428ee2fccb7ef3d0f5469d21c6c9647446c6817f77fa6aa6d3e4846639bd8","side":"left"},{"sibling":"0e03e5b097a0dc88c67962b1d9b168c43326d5ed55bb9b0bf4a84729cf30806d","side":"right"},{"sibling":"e6c9083cd0939b38f691f619de2054068654e9c77f7c7ab051e0d48b77339e5d","side":"left"},{"sibling":"d3139af8c5ce235438e1c69e4b7afa44ba09129fd86674968434f23e546f423e","side":"left"},{"sibling":"a0d45efaf4b22735095323ce2b3134097dd3c43da10fd8c2d9ac270583d612e5","side":"right"},{"sibling":"40f01d4d8df42e716123d474dca7cf07500676d03b8c29dfa4d412431f0fd4ab","side":"right"},{"sibling":"f3e45bceed774d2402fa45d41ff5190f295823bd2f216eb90157884150034693","side":"left"},{"sibling":"32b06d058018452104980f93016dc123b51964332208c9553ec27f6b6e4f9ddb","side":"right"},{"sibling":"77025bcb374a7ad74f520643e20a8ae1205a7ee78507b0beb93117760f1c29d3","side":"left"},{"sibling":"85ad89aeee42094716cdcc4221fc8f01dfb8c989b24193beec402ac5ad9b5ce2","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":347633,"merkle_root":"cde404840c5ea5b08ad99dc1d20529583a1d3006be0c8b1083b0ca24b4756e81","public_key_hex":"cf742e26f75669dc673cb5c0786a1ae23ae8ca19c347317192ce40c28a7ff25c","run_id":"hourly_json_retrofit_20260724T063701Z","schema":"crovia.seal.v1","seal_family_version":"crovia-seal-family/1","seal_kind":"substrate_batch","sealed_at":"2026-07-24T06:38:43Z","sig_algorithm":"ed25519","signature":"86b59d5ba3f8e095f25274d895f4588497f73112b6db20038ebe7e15f8302994d3c5c649a546cad7b8904fc04d1995dccbf606f70c39861039aca39c8457780f","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_3af5361f7ebaa87d5eac835776075681e291294e51f1c8b5c7ab003c0d6b2da8"}}