Refocus organization card on Fide AI
Browse files- README.md +15 -17
- index.html +124 -61
README.md
CHANGED
|
@@ -10,30 +10,28 @@ short_description: Open research and benchmarks for faith-facing AI evaluation.
|
|
| 10 |
|
| 11 |
**Public standards for faith-facing AI evaluation.**
|
| 12 |
|
| 13 |
-
Fide AI
|
| 14 |
-
|
| 15 |
questions.
|
| 16 |
|
| 17 |
-
|
|
|
|
| 18 |
|
| 19 |
-
|
| 20 |
-
- Faith and moral guidance benchmarks
|
| 21 |
-
- Structured harness design for safer model behavior
|
| 22 |
-
- Responsible-use standards for spiritually sensitive AI contexts
|
| 23 |
-
|
| 24 |
-
## Featured
|
| 25 |
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
|
|
|
| 30 |
|
| 31 |
## Links
|
| 32 |
|
| 33 |
-
[Website](https://fideai.org) · [GitHub](https://github.com/FideAI) · [FMG-Bench
|
| 34 |
|
| 35 |
## Responsible Use
|
| 36 |
|
| 37 |
-
|
| 38 |
-
comparable, and reviewable.
|
| 39 |
-
safety, legal adequacy, denominational standing, or deployment
|
|
|
|
|
|
| 10 |
|
| 11 |
**Public standards for faith-facing AI evaluation.**
|
| 12 |
|
| 13 |
+
Fide AI is an independent research initiative developing public standards for AI
|
| 14 |
+
systems that operate near faith, moral guidance, and pastoral-care-adjacent
|
| 15 |
questions.
|
| 16 |
|
| 17 |
+
We study how models behave when questions are not merely informational, but
|
| 18 |
+
touch conscience, community, doctrine, identity, authority, and care.
|
| 19 |
|
| 20 |
+
## What We Work On
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
|
| 22 |
+
- Theological triage and tradition-aware evaluation
|
| 23 |
+
- Structured harnesses for safer faith-facing AI behavior
|
| 24 |
+
- Grounding, preference fidelity, comparison, and escalation boundaries
|
| 25 |
+
- Responsible-use standards for spiritually sensitive contexts
|
| 26 |
+
- Open benchmarks and reproducible evaluation methods
|
| 27 |
|
| 28 |
## Links
|
| 29 |
|
| 30 |
+
[Website](https://fideai.org) · [GitHub](https://github.com/FideAI) · [FMG-Bench](https://huggingface.co/datasets/FideAI/fmg-bench)
|
| 31 |
|
| 32 |
## Responsible Use
|
| 33 |
|
| 34 |
+
Our work is intended to make AI behavior in spiritually sensitive contexts
|
| 35 |
+
visible, comparable, and reviewable. It does not certify pastoral authority,
|
| 36 |
+
clinical safety, legal adequacy, denominational standing, or deployment
|
| 37 |
+
readiness.
|
index.html
CHANGED
|
@@ -7,74 +7,137 @@
|
|
| 7 |
<style>
|
| 8 |
:root { color-scheme: light dark; }
|
| 9 |
* { box-sizing: border-box; }
|
| 10 |
-
body {
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
.
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
.
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
.
|
| 37 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
</style>
|
| 39 |
</head>
|
| 40 |
<body>
|
| 41 |
-
<main class="
|
| 42 |
-
<
|
| 43 |
-
<
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
|
|
|
|
|
|
|
|
|
| 48 |
</div>
|
| 49 |
-
<
|
| 50 |
-
<h2>
|
| 51 |
-
<
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
<
|
| 55 |
-
|
| 56 |
-
</section>
|
| 57 |
-
<section class="project">
|
| 58 |
-
<div class="tile">
|
| 59 |
-
<h2>FMG-Bench</h2>
|
| 60 |
-
<p>The Faith & Moral Guidance Benchmark evaluates large language models in theological triage and pastoral guidance contexts.</p>
|
| 61 |
-
<div class="links">
|
| 62 |
-
<a class="btn" href="https://huggingface.co/datasets/FideAI/fmg-bench">Dataset</a>
|
| 63 |
-
<a class="btn" href="https://github.com/FideAI/fmg-bench">GitHub</a>
|
| 64 |
-
<a class="btn" href="https://fideai.org">Website</a>
|
| 65 |
-
</div>
|
| 66 |
</div>
|
| 67 |
-
<div class="
|
| 68 |
-
<h2>
|
| 69 |
-
<
|
| 70 |
-
<li>Theological triage and tradition-aware evaluation</li>
|
| 71 |
-
<li>Structured harness design for safer model behavior</li>
|
| 72 |
-
<li>Grounding, preference fidelity, comparison, and escalation boundaries</li>
|
| 73 |
-
<li>Responsible-use standards for spiritually sensitive AI contexts</li>
|
| 74 |
-
</ul>
|
| 75 |
</div>
|
| 76 |
</section>
|
| 77 |
-
<
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 78 |
</main>
|
| 79 |
</body>
|
| 80 |
</html>
|
|
|
|
| 7 |
<style>
|
| 8 |
:root { color-scheme: light dark; }
|
| 9 |
* { box-sizing: border-box; }
|
| 10 |
+
body {
|
| 11 |
+
margin: 0;
|
| 12 |
+
font-family: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif;
|
| 13 |
+
background: transparent;
|
| 14 |
+
color: inherit;
|
| 15 |
+
}
|
| 16 |
+
.wrap {
|
| 17 |
+
max-width: 1040px;
|
| 18 |
+
padding: 34px 42px 38px;
|
| 19 |
+
}
|
| 20 |
+
.kicker {
|
| 21 |
+
margin: 0 0 18px;
|
| 22 |
+
font-size: 12px;
|
| 23 |
+
font-weight: 760;
|
| 24 |
+
letter-spacing: .11em;
|
| 25 |
+
text-transform: uppercase;
|
| 26 |
+
color: #8ab4ff;
|
| 27 |
+
}
|
| 28 |
+
h1 {
|
| 29 |
+
margin: 0;
|
| 30 |
+
max-width: 900px;
|
| 31 |
+
font-size: clamp(42px, 5.6vw, 76px);
|
| 32 |
+
line-height: .95;
|
| 33 |
+
letter-spacing: 0;
|
| 34 |
+
font-weight: 820;
|
| 35 |
+
}
|
| 36 |
+
.lead {
|
| 37 |
+
margin: 24px 0 0;
|
| 38 |
+
max-width: 850px;
|
| 39 |
+
font-size: clamp(18px, 2.1vw, 25px);
|
| 40 |
+
line-height: 1.36;
|
| 41 |
+
font-weight: 620;
|
| 42 |
+
color: color-mix(in srgb, currentColor 88%, transparent);
|
| 43 |
+
}
|
| 44 |
+
.body {
|
| 45 |
+
margin: 18px 0 0;
|
| 46 |
+
max-width: 780px;
|
| 47 |
+
font-size: 15.5px;
|
| 48 |
+
line-height: 1.62;
|
| 49 |
+
color: color-mix(in srgb, currentColor 72%, transparent);
|
| 50 |
+
}
|
| 51 |
+
.rule {
|
| 52 |
+
height: 1px;
|
| 53 |
+
width: min(100%, 820px);
|
| 54 |
+
margin: 28px 0 22px;
|
| 55 |
+
background: linear-gradient(90deg, rgba(96,165,250,.55), rgba(168,85,247,.35), transparent);
|
| 56 |
+
}
|
| 57 |
+
.grid {
|
| 58 |
+
display: grid;
|
| 59 |
+
grid-template-columns: repeat(4, minmax(0, 1fr));
|
| 60 |
+
gap: 18px;
|
| 61 |
+
max-width: 980px;
|
| 62 |
+
margin: 0 0 26px;
|
| 63 |
+
}
|
| 64 |
+
.principle {
|
| 65 |
+
min-width: 0;
|
| 66 |
+
}
|
| 67 |
+
.principle h2 {
|
| 68 |
+
margin: 0 0 7px;
|
| 69 |
+
font-size: 14px;
|
| 70 |
+
line-height: 1.25;
|
| 71 |
+
font-weight: 780;
|
| 72 |
+
color: color-mix(in srgb, currentColor 90%, transparent);
|
| 73 |
+
}
|
| 74 |
+
.principle p {
|
| 75 |
+
margin: 0;
|
| 76 |
+
font-size: 13.5px;
|
| 77 |
+
line-height: 1.48;
|
| 78 |
+
color: color-mix(in srgb, currentColor 62%, transparent);
|
| 79 |
+
}
|
| 80 |
+
.links {
|
| 81 |
+
display: flex;
|
| 82 |
+
flex-wrap: wrap;
|
| 83 |
+
gap: 18px;
|
| 84 |
+
margin: 4px 0 0;
|
| 85 |
+
font-size: 14px;
|
| 86 |
+
}
|
| 87 |
+
a {
|
| 88 |
+
color: #8ab4ff;
|
| 89 |
+
font-weight: 720;
|
| 90 |
+
text-decoration: none;
|
| 91 |
+
}
|
| 92 |
+
a:hover { text-decoration: underline; }
|
| 93 |
+
.note {
|
| 94 |
+
margin: 24px 0 0;
|
| 95 |
+
max-width: 850px;
|
| 96 |
+
font-size: 13px;
|
| 97 |
+
line-height: 1.5;
|
| 98 |
+
color: color-mix(in srgb, currentColor 56%, transparent);
|
| 99 |
+
}
|
| 100 |
+
@media (max-width: 820px) {
|
| 101 |
+
.wrap { padding: 26px 22px 30px; }
|
| 102 |
+
.grid { grid-template-columns: repeat(2, minmax(0, 1fr)); gap: 16px; }
|
| 103 |
+
}
|
| 104 |
+
@media (max-width: 520px) {
|
| 105 |
+
.grid { grid-template-columns: 1fr; }
|
| 106 |
+
.links { gap: 12px; }
|
| 107 |
+
}
|
| 108 |
</style>
|
| 109 |
</head>
|
| 110 |
<body>
|
| 111 |
+
<main class="wrap">
|
| 112 |
+
<p class="kicker">Fide AI</p>
|
| 113 |
+
<h1>Public standards for faith-facing AI.</h1>
|
| 114 |
+
<p class="lead">We study how AI systems behave when questions touch conscience, community, doctrine, identity, authority, and care.</p>
|
| 115 |
+
<p class="body">Fide AI is an independent research initiative developing benchmarks, evaluation methods, and responsible-use standards for models operating near faith, moral guidance, and pastoral-care-adjacent contexts.</p>
|
| 116 |
+
<div class="rule" aria-hidden="true"></div>
|
| 117 |
+
<section class="grid" aria-label="Fide AI focus areas">
|
| 118 |
+
<div class="principle">
|
| 119 |
+
<h2>Theological triage</h2>
|
| 120 |
+
<p>Distinguishing creedal claims, tradition-specific disagreements, lower-certainty questions, and pastoral application.</p>
|
| 121 |
</div>
|
| 122 |
+
<div class="principle">
|
| 123 |
+
<h2>Tradition-aware evaluation</h2>
|
| 124 |
+
<p>Measuring whether systems represent Christian traditions carefully without flattening meaningful disagreement.</p>
|
| 125 |
+
</div>
|
| 126 |
+
<div class="principle">
|
| 127 |
+
<h2>Structured harnesses</h2>
|
| 128 |
+
<p>Studying instruction-layer scaffolding for grounding, preference handling, comparison, and escalation boundaries.</p>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 129 |
</div>
|
| 130 |
+
<div class="principle">
|
| 131 |
+
<h2>Responsible use</h2>
|
| 132 |
+
<p>Making behavior visible and reviewable without certifying pastoral authority or deployment readiness.</p>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 133 |
</div>
|
| 134 |
</section>
|
| 135 |
+
<nav class="links" aria-label="Fide AI links">
|
| 136 |
+
<a href="https://fideai.org">Website</a>
|
| 137 |
+
<a href="https://github.com/FideAI">GitHub</a>
|
| 138 |
+
<a href="https://huggingface.co/datasets/FideAI/fmg-bench">FMG-Bench</a>
|
| 139 |
+
</nav>
|
| 140 |
+
<p class="note">Our work does not certify pastoral authority, clinical safety, legal adequacy, denominational standing, or deployment readiness.</p>
|
| 141 |
</main>
|
| 142 |
</body>
|
| 143 |
</html>
|