Ainglish An English dialect for AI agents
Menu

Evidence

Compare two experiments

Check what changed in the question, wording or readers before comparing the numbers.

This view keeps both results separate. It does not calculate a combined score or decide whether they reproduce each other.

Change the selected experimentsSearch by proposal, select a result or enter an exact identifier

Find experiments by proposal

Search for ordinary words from a proposal, then choose a match. Searching alone does not change the results below.

Choose two experiments by title, measurement and date. The choices include up to 50 newest public completed results, plus your current selections. Historical results stay labelled. For older records, use the evidence explorer or exact entry below.

Use an exact experiment identifier instead

A full attempt UUID entered here replaces the corresponding choice above. Content hashes are not result identifiers.

Clear comparison

Compare the essentials

Read each question across both results. A matching description is not proof that the studies are scientifically comparable.

QuestionFirst resultSecond result
Which proposal?must-as-rule / must-as-inference — does ‘must’ impose a requirement or report a conclusion?2026-09-02 11:27 UTCmust-as-rule / must-as-inference — does ‘must’ impose a requirement or report a conclusion?2026-09-04 03:11 UTC
What was measured?comprehension accuracyHow does the wording change correct answers from the declared reader panel?comprehension accuracyHow does the wording change correct answers from the declared reader panel?
What was the result?-24.05 percentage pointsReported interval: -32.3558 to -15.1961Not yet counting in evidence decisions-10.68 percentage pointsReported interval: -28.1868 to 8.4156Counts in current evidence decisions
Compared with what?Complete, careful EnglishReader exposure not recorded as a structured labelComplete, careful EnglishReader exposure not recorded as a structured label
Independently settled?DisputedEligible replications disagree and this original does not hold a settlement majority.Disagrees with the named originalThis eligible row adds one disagreement. An adverse or null direction is a valid result and remains visible.

Inspect first resultInspect second result

What differs in the declared methods?

These are literal recorded values, not a compatibility test. Matching declarations do not prove equivalent inputs or fair scoring. A missing structured field may be described elsewhere in the immutable specification.

What was measured · Same recorded value
First result
comprehension_accuracy_delta
Second result
comprehension_accuracy_delta
English comparison declarations · Same recorded value
First result
[
    "complete-careful-english-v1"
]
Second result
[
    "complete-careful-english-v1"
]
Tested population · Not recorded on either side
First result
Not recorded in this structured field
Second result
Not recorded in this structured field
Unit tested · Not recorded on either side
First result
Not recorded in this structured field
Second result
Not recorded in this structured field
Named instruments (including order) · Recorded values differ
First result
[
    "mistral-small3.2-24b-opaque-choice-q4_k_m@q4_k_m",
    "gemma3-12b-opaque-choice-q4_k_m@q4_k_m"
]
Second result
[
    "Sat-Qwen7-Q4@q4_k_m",
    "Sat-Gemma12-Q4@q4_k_m",
    "Sat-Mistral24-Q4@q4_k_m"
]
Reader exposure · Not recorded on either side
First result
Not recorded in this structured field
Second result
Not recorded in this structured field
Exposure window · Not recorded on either side
First result
Not recorded in this structured field
Second result
Not recorded in this structured field
How results combine · Not recorded on either side
First result
Not recorded in this structured field
Second result
Not recorded in this structured field
Scoring declaration · Not recorded on either side
First result
Not recorded in this structured field
Second result
Not recorded in this structured field
Conditions (including order) · Same recorded value
First result
[
    {
        "id": "must-as-rule",
        "weight": 1
    },
    {
        "id": "must-as-inference",
        "weight": 1
    }
]
Second result
[
    {
        "id": "must-as-rule",
        "weight": 1
    },
    {
        "id": "must-as-inference",
        "weight": 1
    }
]
First result: full recordcomprehension accuracy · -24.05 percentage points

First result · 2026-09-02 11:27 UTC

must-as-rule / must-as-inference — does ‘must’ impose a requirement or report a conclusion?

Not yet counting in evidence decisions. This row remains available for assessment, but does not currently carry a counting evidence result.

What was measured
comprehension accuracy · comprehension_accuracy_delta
How does the wording change correct answers from the declared reader panel?
Reported result
-24.05 percentage points
Reported interval: -32.3558 to -15.1961.

The value falls on the registered harmful side of this metric’s neutral point.

English comparison
Complete, careful English

Declared by the submitter; not a certification of equivalent information.

Every scientific compact form is compared with its complete registered careful-English meaning; ambiguous bare English is absent from the scalar.

Reader exposure
Reader exposure not recorded as a structured label
Named instruments
mistral-small3.2-24b-opaque-choice-q4_k_m@q4_k_m, gemma3-12b-opaque-choice-q4_k_m@q4_k_m

Reader population not separately declared.

Conditions covered
Separate outcomes retained for all 2 declared conditions

must-as-rule, must-as-inference

An overall average can hide a weak condition. A condition list is not proof that every form or claim in the proposal was tested.

Settlement role
Disputed

Eligible replications disagree and this original does not hold a settlement majority.

How often did each version lead to the right answer?

English comparison
96.77%
96.77%
Ainglish version
72.72%
72.72%

Reported real-item accuracy, not the separate calibration score. Both bars use the same 0–100% scale. The difference is measured in percentage points, not percent improvement. Any declared stratum weights are already applied.

Reported item-bootstrap interval: -32.3558 to -15.1961 percentage points.

This interval concerns the difference, not separate uncertainty bounds for either accuracy. It does not measure uncertainty across humans or future models.

At least one declared condition is resolution-limited. The overall interval does not settle every condition.

Real cases: 64 · Named readers: 2. These are different units; multiple answers to one case are not new cases.

Does the overall result hide differences between conditions?

Every stored condition, without new pooling. Differences and intervals use percentage points. Condition names come from the frozen experiment.
ConditionReported differenceReported intervalEnglish accuracyAinglish accuracy
must-as-rule0 Not recorded 100.00%100.00%
must-as-inference-48.1 Not recorded 93.55%45.45%

A missing condition interval is not zero uncertainty. An overall interval cannot substitute for agreement in every load-bearing condition.

Inspect externally stored inputs and recorded answers

The comparison label is the submitter’s declaration, not a semantic certification. Check that both versions preserve the information needed to answer the same question.

Numbers count only readable inputs attached to this receipt. They are not the experiment’s declared sample size or the number of reader calls.

The input material is linked externally. The number of study items and controls in that file has not been checked by this website. “External file” does not mean zero inputs.

Open the declared external input artifact. This is an unverified external link, not a hosted or inspected copy.

Declared input digest: 04a2ccb275d7ef9d865765cf37d1c1a2c3f9cd7c2a2b0d227107421ff9dc087d. A recorded digest alone does not establish that the linked file matches it.

The website does not fetch the file. Verify the declared digest recipe before relying on it: SDK item digests use canonical JSON of the item array, not the raw pretty-printed file bytes.

No readable study input pairs are stored inline in this receipt. This does not mean the experiment used none.

Prompts, reference material and other context can live elsewhere in the specification. Inputs and keys alone do not reconstruct every reader call or establish a fair comparison.

Declared population, method and retained outcomes

No structured study scope is declared here. Inspect the immutable manifest; do not infer a comparator or population from the headline.

Absolute arm results, reader-specific results and condition results below are retained values, not a newly pooled analysis. Accuracy arms use fractions from 0 to 1; their difference uses percentage points.

Absolute arm results

{
    "english": 0.9677000000000000046185277824406512081623077392578125,
    "ainglish": 0.72719999999999995754507153833401389420032501220703125,
    "chance": 0.25
}

Reader or tokenizer results

[
    {
        "model": "mistral-small3.2-24b-opaque-choice-q4_k_m",
        "value": -21.42999999999999971578290569595992565155029296875,
        "precision": "q4_k_m"
    },
    {
        "model": "gemma3-12b-opaque-choice-q4_k_m",
        "value": -27.059999999999998721023075631819665431976318359375,
        "precision": "q4_k_m"
    }
]

Condition results

[
    {
        "id": "must-as-rule",
        "weight": 1,
        "share": 0.5,
        "value": 0,
        "value_lo": null,
        "value_hi": null,
        "arms": {
            "english": 1,
            "ainglish": 1,
            "chance": 0.25
        },
        "resolution_bound": "ceiling"
    },
    {
        "id": "must-as-inference",
        "weight": 1,
        "share": 0.5,
        "value": -48.10000000000000142108547152020037174224853515625,
        "value_lo": null,
        "value_hi": null,
        "arms": {
            "english": 0.9354999999999999982236431605997495353221893310546875,
            "ainglish": 0.45450000000000001509903313490212894976139068603515625,
            "chance": 0.25
        },
        "resolution_bound": "resolvable"
    }
]

Exact result and immutable specificationExperiment history

Attempt dcdbafa8-9664-4267-ae94-919c612e899e
Content fa10a69200a49d993f82da7871741b4c3bb3cf6e32cd090322cb3a39f3f0f50d

Second result: full recordcomprehension accuracy · -10.68 percentage points

Second result · 2026-09-04 03:11 UTC

must-as-rule / must-as-inference — does ‘must’ impose a requirement or report a conclusion?

Counts in current evidence decisions. This row currently contributes to evidence decisions. Its direction is separate from whether the proposal is ready for adoption.

What was measured
comprehension accuracy · comprehension_accuracy_delta
How does the wording change correct answers from the declared reader panel?
Reported result
-10.68 percentage points
Reported interval: -28.1868 to 8.4156.

The value is neutral or does not resolve the registered direction.

English comparison
Complete, careful English

Declared by the submitter; not a certification of equivalent information.

Each marked form is compared with its complete registered careful-English meaning; bare ambiguous must is absent.

Reader exposure
Reader exposure not recorded as a structured label
Named instruments
Sat-Qwen7-Q4@q4_k_m, Sat-Gemma12-Q4@q4_k_m, Sat-Mistral24-Q4@q4_k_m

Reader population not separately declared.

Conditions covered
Separate outcomes retained for all 2 declared conditions

must-as-rule, must-as-inference

An overall average can hide a weak condition. A condition list is not proof that every form or claim in the proposal was tested.

Settlement role
Disagrees with the named original

This eligible row adds one disagreement. An adverse or null direction is a valid result and remains visible.

This result checks a named original, not every experiment on the proposal. Read its target original

Compare with the exact target attempt

Every declared condition must agree. Overlapping overall intervals alone do not confirm this original.

How much input text was reused?

Complete-pair freshness is not available for this receipt.

Separate-arm overlap is unavailable or has not been computed. This does not mean zero reuse.

Exact text comparisons only; repeated occurrences count separately. Shared text can deserve scrutiny even when each complete pair is new. These arm counts are descriptive and do not change settlement eligibility.

Declared target content identityfa10a69200a49d993f82da7871741b4c3bb3cf6e32cd090322cb3a39f3f0f50d

How often did each version lead to the right answer?

English comparison
67.18%
67.18%
Ainglish version
56.50%
56.50%

Reported real-item accuracy, not the separate calibration score. Both bars use the same 0–100% scale. The difference is measured in percentage points, not percent improvement. Any declared stratum weights are already applied.

Reported item-bootstrap interval: -28.1868 to 8.4156 percentage points.

This interval concerns the difference, not separate uncertainty bounds for either accuracy. It does not measure uncertainty across humans or future models.

Real cases: 24 · Named readers: 3. These are different units; multiple answers to one case are not new cases.

Does the overall result hide differences between conditions?

Every stored condition, without new pooling. Differences and intervals use percentage points. Condition names come from the frozen experiment.
ConditionReported differenceReported intervalEnglish accuracyAinglish accuracy
must-as-rule13 Not recorded 76.47%89.47%
must-as-inference-34.36 Not recorded 57.89%23.53%

A missing condition interval is not zero uncertainty. An overall interval cannot substitute for agreement in every load-bearing condition.

Inspect actual inputs and recorded answers

The comparison label is the submitter’s declaration, not a semantic certification. Check that both versions preserve the information needed to answer the same question.

Numbers count only readable inputs attached to this receipt. They are not the experiment’s declared sample size or the number of reader calls.

Showing 1–6 of 24 readable, inline study items, in stored order—not a selection of successes. 8 control items are kept separate.

Each result has its own input pages. Positions across the two studies do not imply matched cases.

Input 1 · mr01

English input
The release checklist requires the release bot to attach a provenance file; it does not say it happens.
Ainglish input
The release bot must-as-rule attach a provenance file.
Question
Observed: the published bundle has no provenance file. Result?
Recorded answer options
rule broken · inference wrong · both · neither
Submitted answer key
rule broken. This is the supplied key, not an independent validation of it.
Condition
must-as-rule

Filed correctness against that key (up to 12 reader/arm cells). These flags are not the reader’s verbatim output.

  • Sat-Gemma12-Q4 · English: did not match the submitted key.
  • Sat-Mistral24-Q4 · Ainglish: matched the submitted key.
  • Sat-Qwen7-Q4 · Ainglish: matched the submitted key.

Input 2 · mi01

English input
The signed bundle index supports the conclusion that the release bot attached a provenance file; it creates no duty.
Ainglish input
From the signed bundle index, the release bot must-as-inference have attached a provenance file.
Question
Observed: the published bundle has no provenance file. Result?
Recorded answer options
rule broken · inference wrong · both · neither
Submitted answer key
inference wrong. This is the supplied key, not an independent validation of it.
Condition
must-as-inference

Filed correctness against that key (up to 12 reader/arm cells). These flags are not the reader’s verbatim output.

  • Sat-Gemma12-Q4 · English: did not match the submitted key.
  • Sat-Mistral24-Q4 · Ainglish: did not match the submitted key.
  • Sat-Qwen7-Q4 · Ainglish: did not match the submitted key.

Input 3 · mr02

English input
The replay-protection policy requires the edge gateway to reject a request with an expired nonce; it does not say it happens.
Ainglish input
The edge gateway must-as-rule reject a request with an expired nonce.
Question
Observed: the gateway accepted that request. Result?
Recorded answer options
both · neither · rule broken · inference wrong
Submitted answer key
rule broken. This is the supplied key, not an independent validation of it.
Condition
must-as-rule

Filed correctness against that key (up to 12 reader/arm cells). These flags are not the reader’s verbatim output.

  • Sat-Gemma12-Q4 · Ainglish: matched the submitted key.
  • Sat-Mistral24-Q4 · Ainglish: matched the submitted key.
  • Sat-Qwen7-Q4 · English: matched the submitted key.

Input 4 · mi02

English input
The gateway decision log supports the conclusion that the edge gateway rejected the request with the expired nonce; it creates no duty.
Ainglish input
From the gateway decision log, the edge gateway must-as-inference have rejected the request with the expired nonce.
Question
Observed: the gateway accepted that request. Result?
Recorded answer options
both · neither · rule broken · inference wrong
Submitted answer key
inference wrong. This is the supplied key, not an independent validation of it.
Condition
must-as-inference

Filed correctness against that key (up to 12 reader/arm cells). These flags are not the reader’s verbatim output.

  • Sat-Gemma12-Q4 · English: matched the submitted key.
  • Sat-Mistral24-Q4 · Ainglish: matched the submitted key.
  • Sat-Qwen7-Q4 · English: did not match the submitted key.

Input 5 · mr03

English input
The recovery procedure requires the backup service to finish snapshot nine before key rotation; it does not say it happens.
Ainglish input
The backup service must-as-rule finish snapshot nine before key rotation.
Question
Observed: the key rotated while snapshot nine was still open. Result?
Recorded answer options
rule broken · inference wrong · both · neither
Submitted answer key
rule broken. This is the supplied key, not an independent validation of it.
Condition
must-as-rule

Filed correctness against that key (up to 12 reader/arm cells). These flags are not the reader’s verbatim output.

  • Sat-Gemma12-Q4 · Ainglish: matched the submitted key.
  • Sat-Mistral24-Q4 · Ainglish: matched the submitted key.
  • Sat-Qwen7-Q4 · Ainglish: matched the submitted key.

Input 6 · mi03

English input
The snapshot completion receipt supports the conclusion that the backup service finished snapshot nine before key rotation; it creates no duty.
Ainglish input
From the snapshot completion receipt, the backup service must-as-inference have finished snapshot nine before key rotation.
Question
Observed: the key rotated while snapshot nine was still open. Result?
Recorded answer options
rule broken · inference wrong · both · neither
Submitted answer key
inference wrong. This is the supplied key, not an independent validation of it.
Condition
must-as-inference

Filed correctness against that key (up to 12 reader/arm cells). These flags are not the reader’s verbatim output.

  • Sat-Gemma12-Q4 · English: matched the submitted key.
  • Sat-Mistral24-Q4 · Ainglish: did not match the submitted key.
  • Sat-Qwen7-Q4 · English: did not match the submitted key.

Recorded input digest: 12137cf96752fd5f4daec789c4399ee6bd399f917af51196cc54abd95846d97e

Prompts, reference material and other context can live elsewhere in the specification. Inputs and keys alone do not reconstruct every reader call or establish a fair comparison.

Declared population, method and retained outcomes

No structured study scope is declared here. Inspect the immutable manifest; do not infer a comparator or population from the headline.

Absolute arm results, reader-specific results and condition results below are retained values, not a newly pooled analysis. Accuracy arms use fractions from 0 to 1; their difference uses percentage points.

Absolute arm results

{
    "english": 0.6717999999999999527489080719533376395702362060546875,
    "ainglish": 0.564999999999999946709294817992486059665679931640625,
    "chance": 0.25
}

Reader or tokenizer results

[
    {
        "model": "Sat-Qwen7-Q4",
        "value": -11.42999999999999971578290569595992565155029296875,
        "precision": "q4_k_m"
    },
    {
        "model": "Sat-Gemma12-Q4",
        "value": -24.995000000000000994759830064140260219573974609375,
        "precision": "q4_k_m"
    },
    {
        "model": "Sat-Mistral24-Q4",
        "value": 0,
        "precision": "q4_k_m"
    }
]

Condition results

[
    {
        "id": "must-as-rule",
        "weight": 1,
        "share": 0.5,
        "value": 13,
        "value_lo": null,
        "value_hi": null,
        "arms": {
            "english": 0.76470000000000004636291350834653712809085845947265625,
            "ainglish": 0.8947000000000000508038056068471632897853851318359375,
            "chance": 0.25
        },
        "resolution_bound": "resolvable"
    },
    {
        "id": "must-as-inference",
        "weight": 1,
        "share": 0.5,
        "value": -34.3599999999999994315658113919198513031005859375,
        "value_lo": null,
        "value_hi": null,
        "arms": {
            "english": 0.57889999999999997015720509807579219341278076171875,
            "ainglish": 0.235300000000000009148237722911289893090724945068359375,
            "chance": 0.25
        },
        "resolution_bound": "resolvable"
    }
]

Exact result and immutable specificationExperiment history

Attempt da037138-e559-4ec9-a247-94ba2c3352c3
Content f3857f4a2f36f9da5fd9b78e6be49da43447772244dcbba95a1cd5965e0ebcc6

Different wording, readers, exposure or populations can legitimately produce different results. A visible reference is not training the model’s weights. Current models and tokenizers have learned English; future Ainglish-trained performance remains a research question.