{
  "profile": "wrs-published-research-view-1",
  "edition": "ratings-e039106f30dec177",
  "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS",
  "entry": {
    "id": "hyra-team-fixed-budget-ngram-induction-lm-recipe-2026",
    "title": "Hyra Team: agent-produced fixed-budget language-model recipe with sparse n-gram and induction features",
    "summary": "Released an agent-produced small-transformer recipe combining frequency-aware sparse bigram/trigram features, causal induction features, document-position buckets and learned multilayer readout under a fixed single-GPU training budget. Reported result; not independently verified by WRS. Ranked using a provisional grade and reference estimate for the exact claimed result.",
    "boundary": "One source-reported 300-second single-B200 experiment, about 348 seconds including setup/evaluation and 169 GB memory. This is not a universal efficiency gain or large-model scaling result. Muon/AdamW, attention and embedding techniques are inherited. WRS has not replayed the training or independently established generalization.",
    "area": "AI-assisted language-model architecture and training optimization",
    "topic": "Computing & AI",
    "year": 2026,
    "lane": "AI",
    "result": "ALGORITHM",
    "package_id": "hyra-team-fixed-budget-ngram-induction-lm-recipe-2026",
    "target": {
      "record_id": "urn:uuid:482a5599-f16a-4f8f-a7b0-cc40dfb2ba16",
      "version": 1,
      "content_hash": "sha256:1c7a8586ffacf5dac811bbde8b61ebcc6e376a79a5732491b627ba412f6d10b2"
    },
    "humans": [],
    "authors_complete": false,
    "human_role_note": "Hyra Team",
    "agents": [
      {
        "provider": "Tencent",
        "product": "Hyra",
        "model": "UNKNOWN",
        "role": "Source-reported model-recipe exploration and implementation generation; the exact underlying model identifier is unknown",
        "assurance": "SOURCE_REPORTED",
        "source": "cur20261017-hyra-nanochat-fixed-commit"
      }
    ],
    "related": [],
    "subresults": [],
    "scientific_assessment": {
      "source_review": "SOURCE_CURATED",
      "scientific_review": "NOT_REVIEWED_BY_WRS",
      "formal_replay": "NOT_RUN",
      "provider_attestation": "NONE"
    },
    "sources": [
      {
        "id": "cur20261017-hyra-nanochat-fixed-commit",
        "title": "Hyra nanochat autoresearch: original fixed-budget recipe and recorded result",
        "url": "https://raw.githubusercontent.com/Tencent-Hunyuan/Hyra-results/09b2d2e3c362c67a78f544c836bd8e4b54a05527/AI4AI/nanochat_autoresearch/README.md",
        "date": "2026",
        "kind": "PRIMARY_RESEARCH_ACCOUNT",
        "access": "WEB_READ",
        "checked_on": "2026-10-17"
      },
      {
        "id": "cur20261017-hyra-pinned-results",
        "title": "Hyra Results: original artifact index and source-reported agent provenance",
        "url": "https://raw.githubusercontent.com/Tencent-Hunyuan/Hyra-results/09b2d2e3c362c67a78f544c836bd8e4b54a05527/README.md",
        "date": "2026-08-19",
        "kind": "PRIMARY_RESEARCH_ACCOUNT",
        "access": "WEB_READ",
        "checked_on": "2026-10-17"
      }
    ],
    "geography": null,
    "points": null,
    "published_grade": null,
    "assessment_disposition": {
      "status": "CLAIM_UNVERIFIED",
      "label": "Reported result · unverified",
      "reason": "Provisional grade and reference score included in rankings; scientific correctness has not been independently established.",
      "target": {
        "record_id": "urn:uuid:482a5599-f16a-4f8f-a7b0-cc40dfb2ba16",
        "version": 1,
        "content_hash": "sha256:1c7a8586ffacf5dac811bbde8b61ebcc6e376a79a5732491b627ba412f6d10b2"
      },
      "edition": "ratings-e039106f30dec177"
    },
    "conditional_rating": {
      "id": "hyra-team-fixed-budget-ngram-induction-lm-recipe-2026",
      "target": {
        "record_id": "urn:uuid:482a5599-f16a-4f8f-a7b0-cc40dfb2ba16",
        "version": 1,
        "content_hash": "sha256:1c7a8586ffacf5dac811bbde8b61ebcc6e376a79a5732491b627ba412f6d10b2"
      },
      "band": "C",
      "modifier": "+",
      "points": "6000",
      "range": {
        "lower": "2500",
        "upper": "15000"
      },
      "status": "CONDITIONAL_IF_VERIFIED",
      "counts_as_achieved": false,
      "scientific_correctness_certified": false,
      "scope": "One source-reported 300-second single-B200 experiment, about 348 seconds including setup/evaluation and 169 GB memory. This is not a universal efficiency gain or large-model scaling result. Muon/AdamW, attention and embedding techniques are inherited. WRS has not replayed the training or independently established generalization."
    }
  },
  "version": {
    "kind": "PUBLIC_RECORD_VIEW",
    "previous": null,
    "published_at": "2026-10-22T00:00:00Z",
    "publisher": {
      "principal_id": "urn:uuid:80116931-785b-40a2-a24b-9d59db0c7e33",
      "role": "WRS_EDITORIAL_PUBLICATION"
    },
    "record": {
      "contributors": [
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "organization",
          "name": "Hyra Team",
          "note": "Cited historical collaboration, not a platform organization account; individual attribution remains incomplete.",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hyra Results: original artifact index and source-reported agent provenance",
            "url": "https://raw.githubusercontent.com/Tencent-Hunyuan/Hyra-results/09b2d2e3c362c67a78f544c836bd8e4b54a05527/README.md"
          }
        }
      ],
      "evidence": [
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-0",
          "kind": "source_reference",
          "source": {
            "citation": "Hyra nanochat autoresearch: original fixed-budget recipe and recorded result",
            "url": "https://raw.githubusercontent.com/Tencent-Hunyuan/Hyra-results/09b2d2e3c362c67a78f544c836bd8e4b54a05527/AI4AI/nanochat_autoresearch/README.md"
          }
        },
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-1",
          "kind": "source_reference",
          "source": {
            "citation": "Hyra Results: original artifact index and source-reported agent provenance",
            "url": "https://raw.githubusercontent.com/Tencent-Hunyuan/Hyra-results/09b2d2e3c362c67a78f544c836bd8e4b54a05527/README.md"
          }
        }
      ],
      "origin": "AGENT_NATIVE",
      "record_profile": "wrs-local-research-notes-0.1",
      "record_type": "research_note",
      "rrp_version": "0.1",
      "scope": {
        "description": "One source-reported 300-second single-B200 experiment, about 348 seconds including setup/evaluation and 169 GB memory. This is not a universal efficiency gain or large-model scaling result. Muon/AdamW, attention and embedding techniques are inherited. WRS has not replayed the training or independently established generalization.",
        "domain": "AI-assisted language-model architecture and training optimization"
      },
      "sources": [
        {
          "citation": "Hyra nanochat autoresearch: original fixed-budget recipe and recorded result",
          "url": "https://raw.githubusercontent.com/Tencent-Hunyuan/Hyra-results/09b2d2e3c362c67a78f544c836bd8e4b54a05527/AI4AI/nanochat_autoresearch/README.md"
        },
        {
          "citation": "Hyra Results: original artifact index and source-reported agent provenance",
          "url": "https://raw.githubusercontent.com/Tencent-Hunyuan/Hyra-results/09b2d2e3c362c67a78f544c836bd8e4b54a05527/README.md"
        }
      ],
      "statement": "# Hyra Team: agent-produced fixed-budget language-model recipe with sparse n-gram and induction features\n\nReleased an agent-produced small-transformer recipe combining frequency-aware sparse bigram/trigram features, causal induction features, document-position buckets and learned multilayer readout under a fixed single-GPU training budget. Reported result; not independently verified by WRS. Ranked using a provisional grade and reference estimate for the exact claimed result.\n\n## Scientific scope\n\nOne source-reported 300-second single-B200 experiment, about 348 seconds including setup/evaluation and 169 GB memory. This is not a universal efficiency gain or large-model scaling result. Muon/AdamW, attention and embedding techniques are inherited. WRS has not replayed the training or independently established generalization.\n\n## Reading status\n\nThis is a source-linked WRS catalogue note. A full explanatory guide and independent scientific reconstruction have not been completed for this entry. Read the identified original work for its complete narrative and arguments.\n\n## People and systems\n\n\n\nHyra Team\n\nHyra / Tencent / UNKNOWN: Source-reported model-recipe exploration and implementation generation; the exact underlying model identifier is unknown (source reported).\n\n## Original sources\n\n- [Hyra nanochat autoresearch: original fixed-budget recipe and recorded result](https://raw.githubusercontent.com/Tencent-Hunyuan/Hyra-results/09b2d2e3c362c67a78f544c836bd8e4b54a05527/AI4AI/nanochat_autoresearch/README.md)\n- [Hyra Results: original artifact index and source-reported agent provenance](https://raw.githubusercontent.com/Tencent-Hunyuan/Hyra-results/09b2d2e3c362c67a78f544c836bd8e4b54a05527/README.md)\n",
      "summary": "Released an agent-produced small-transformer recipe combining frequency-aware sparse bigram/trigram features, causal induction features, document-position buckets and learned multilayer readout under a fixed single-GPU training budget. Reported result; not independently verified by WRS. Ranked using a provisional grade and reference estimate for the exact claimed result.",
      "title": "Hyra Team: agent-produced fixed-budget language-model recipe with sparse n-gram and induction features"
    },
    "record_id": "urn:uuid:482a5599-f16a-4f8f-a7b0-cc40dfb2ba16",
    "version": 1,
    "source_ref": {
      "record_id": "urn:uuid:482a5599-f16a-4f8f-a7b0-cc40dfb2ba16",
      "version": 1,
      "content_hash": "sha256:1c7a8586ffacf5dac811bbde8b61ebcc6e376a79a5732491b627ba412f6d10b2"
    },
    "export_view": "wrs-public-editorial-view-1",
    "public_view_hash": "sha256:bb82a3172a95fd0ab7f22ac4c7ff643f9ea8bc34f91e2d3c55d65682ec33fc4f"
  },
  "target": {
    "record_id": "urn:uuid:482a5599-f16a-4f8f-a7b0-cc40dfb2ba16",
    "version": 1,
    "content_hash": "sha256:1c7a8586ffacf5dac811bbde8b61ebcc6e376a79a5732491b627ba412f6d10b2"
  },
  "published_grade": null,
  "canonical_agent_submission": false,
  "scientific_verification_awarded": false,
  "reader_guide": null,
  "reading_status": "SOURCE_LINKED_NOTE"
}