Evidence row 222241

What was observed

{
  "count": 4,
  "tools": [
    {
      "description": "Estimate the task-level energy change (ΔE% vs FP16) of weight-only quantization (NF4 or INT8) for a model of a given size on a GPU architecture class, from measured-anchored curves. The answer carries basis (measured/interpolated/estimated), confidence and evidence boundaries.",
      "inputSchema": {
        "properties": {
          "arch": {
            "description": "GPU architecture class",
            "enum": [
              "turing",
              "ada",
              "blackwell",
              "ampere",
              "hopper"
            ],
            "type": "string"
          },
          "batch": {
            "description": "batch size (default 1; >1 shrinks savings — modelled, not measured)",
            "type": "integer"
          },
          "ctx": {
            "description": "context length (default 2048)",
            "type": "integer"
          },
          "params_b": {
            "description": "model size in billions of parameters, e.g. 7",
            "type": "number"
          },
          "precision": {
            "description": "quantization format (default NF4)",
            "enum": [
              "NF4",
              "INT8"
            ],
            "type": "string"
          },
          "q": {
            "description": "alternative: plain-English query, e.g. '13B on A100 int8 batch 32'",
            "type": "string"
          }
        },
        "type": "object"
      },
      "name": "estimate_quantization_energy"
    },
    {
      "description": "Get the model-size crossover for NF4/INT8 per architecture: below it quantization costs energy, above it saves. Crossovers belong to a card + software stack, not an architecture — the answer says so.",
      "inputSchema": {
        "properties": {
          "arch": {
            "description": "optional: limit to one architecture",
            "enum": [
              "turing",
              "ada",
              "blackwell",
              "ampere",
              "hopper"
            ],
            "type": "string"
          }
        },
        "type": "object"
      },
      "name": "get_quantization_crossover"
    },
    {
      "description": "List what is actually measured (GPU architecture × precision × size range, NVML direct), what is borrowed/estimated, and where the gaps are. Use this before trusting an estimate at the edges of coverage, and to answer 'is there data for my GPU?' honestly.",
      "inputSchema": {
        "properties": {},
        "type": "object"
      },
      "name": "list_data_coverage"
    },
    {
      "description": "Search the measured anchors by model name, GPU name, architecture or size. Returns ΔE% vs FP16 per anchor with run counts — the raw evidence behind the curves.",
      "inputSchema": {
        "properties": {
          "query": {
            "description": "model ('TinyLlama'), GPU ('4090'), architecture ('ampere') or size ('7')",
            "type": "string"
          }
        },
        "required": [
          "query"
        ],
        "type": "object"
      },
      "name": "search_measurements"
    }
  ]
}

What the leaf hashes

{"body":"ffa6dac46489bac630eb14aa0a0c9f54a973451bb6b71cc6dda5c98ecc1e1697","id":222241,"kind":"mcp.tools","method":"probe.mcp/1","observed_us":1793027899338047,"prev":null,"subject":"c1b4980f-86b2-4104-890c-14079512d8bd","subject_kind":"endpoint"}

Proof

Fold these into the leaf in order, hashing each sorted pair with keccak256, to reach the root of 2026-10-26 (296695 rows).

1. 0xe84736e002967ebb6fca23088238a233d13bb09703ba6ef17f45cf6c9a7aad88
2. 0x0be451f78b03c064d962fe249219e26c0ed35143e78aa3a30d9c3ab457eba999
3. 0x1d3ca0c3beebdb68243a2cec90b40d4be645c6350a51b3388174f91d1fa018fe
4. 0x51e4dfeac80f557a874d8309f48dfa1beca1f8c3a71397ed26b211c2374e3b47
5. 0x629ccd27baf0f7f53309cf33f25b8a406db04fc09421c06c8bebe3fd56694e43
6. 0xf418ec29be28050ec505cab6f5cf98e8d950e5905eb3a986ce65871f112028a3
7. 0xb80bdf9a9f1d0091126919f444a5a2589c589e7a611b1821d6aa759bdc5b103b
8. 0x9d06fd031027de70d45cfea695f43b1360c91ce388c1cc046a207b0cba6d3c94
9. 0x5cef51db826b59f0c60148b2591fed88d0735b46ddbe89ba624a249c06bbaff6
10. 0xdcf8f946c4e700e867538a2fcf3231d4a1788831144ee814093e20884cc85b1a
11. 0xf9ef09a725d52af13e504e0eafd5a3073b223792ca665fa964297ee3fc3cee9b
12. 0x3d0b5282d5282c30fc9f86662367cab01d0dba65f13d2968391a7d087b79dfb3
13. 0xafcc1fe54cfe00a3a1de33df127340075f22433bb89848bbc81499cd7aca5b2b
14. 0x9bc5fc98278b9f6dc18c66c6e162aefd5244be03170d41bc5161f58ad11beed7
15. 0x95657a7d540ed6cac0bde9995782333b2562a9a6641853269ac95148827410f0
16. 0x5fb5dda563336cd1bc1b9f44c52b76237ae72ed2145cb9290f39f304ca7dd678
17. 0x74a0a08e4e7ab9aa5bfbead33dec7e1c7e8e364cff69a63720d764bbe987aadc

Day root

0xf43e8dcc3d451a4788b26be904c55e8c4c0fa75c9666796c51db38d37e2c0df8

Chained

0x702d5ac508f08b0d1accf36af25822d8e5660367494edd17a642fb61e158a57a

Published

On Arc Testnet, block 69251483: 0x495162f8936e19f0039918feebe6c307f327cd6bf8c37219620b1970041fa9ae