diff --git a/GFramework.Game.Tests/Config/YamlConfigLoaderTests.cs b/GFramework.Game.Tests/Config/YamlConfigLoaderTests.cs
index a8f45394..1757e22d 100644
--- a/GFramework.Game.Tests/Config/YamlConfigLoaderTests.cs
+++ b/GFramework.Game.Tests/Config/YamlConfigLoaderTests.cs
@@ -559,10 +559,10 @@ public class YamlConfigLoaderTests
     }
 
     /// <summary>
-    ///     验证大数值配合十进制步进时，会沿用 JS 工具侧的 <c>multipleOf</c> 容差策略。
+    ///     验证大数值配合十进制步进时，会按十进制精确整倍数规则被运行时接受。
     /// </summary>
     [Test]
-    public async Task LoadAsync_Should_Accept_Large_Decimal_Number_When_MultipleOf_Matches_Js_Tolerance()
+    public async Task LoadAsync_Should_Accept_Large_Decimal_Number_When_MultipleOf_Matches_Exact_Decimal_Step()
     {
         CreateConfigFile(
             "monster/slime.yaml",
@@ -602,6 +602,50 @@ public class YamlConfigLoaderTests
         });
     }
 
+    /// <summary>
+    ///     验证大数量级但实际不满足 <c>multipleOf</c> 的数值会被运行时拒绝。
+    /// </summary>
+    [Test]
+    public void LoadAsync_Should_Throw_When_Large_Number_Is_Not_Actually_MultipleOf()
+    {
+        CreateConfigFile(
+            "monster/slime.yaml",
+            """
+            id: 1
+            dropRate: 1000000000000.4
+            """);
+        CreateSchemaFile(
+            "schemas/monster.schema.json",
+            """
+            {
+              "type": "object",
+              "required": ["id", "dropRate"],
+              "properties": {
+                "id": { "type": "integer" },
+                "dropRate": {
+                  "type": "number",
+                  "multipleOf": 1
+                }
+              }
+            }
+            """);
+
+        var loader = new YamlConfigLoader(_rootPath)
+            .RegisterTable<int, MonsterNumberConfigStub>("monster", "monster", "schemas/monster.schema.json",
+                static config => config.Id);
+        var registry = new ConfigRegistry();
+
+        var exception = Assert.ThrowsAsync<ConfigLoadException>(async () => await loader.LoadAsync(registry));
+
+        Assert.Multiple(() =>
+        {
+            Assert.That(exception, Is.Not.Null);
+            Assert.That(exception!.Diagnostic.FailureKind, Is.EqualTo(ConfigLoadFailureKind.ConstraintViolation));
+            Assert.That(exception.Diagnostic.DisplayPath, Is.EqualTo("dropRate"));
+            Assert.That(registry.Count, Is.EqualTo(0));
+        });
+    }
+
     /// <summary>
     ///     验证科学计数法数值会按 <c>number</c> 类型被运行时接受。
     /// </summary>
@@ -1702,7 +1746,7 @@ public class YamlConfigLoaderTests
 
         Assert.That(exception!.ParamName, Is.EqualTo("options"));
     }
-
+    
     /// <summary>
     ///     验证热重载失败时会保留旧表状态，并通过失败回调暴露诊断信息。
     /// </summary>
diff --git a/GFramework.Game/Config/YamlConfigSchemaValidator.cs b/GFramework.Game/Config/YamlConfigSchemaValidator.cs
index efeda8fa..7af85af0 100644
--- a/GFramework.Game/Config/YamlConfigSchemaValidator.cs
+++ b/GFramework.Game/Config/YamlConfigSchemaValidator.cs
@@ -1,3 +1,4 @@
+using System.Numerics;
 using System.Text.RegularExpressions;
 using GFramework.Game.Abstractions.Config;
 
@@ -15,6 +16,9 @@ internal static class YamlConfigSchemaValidator
     // The runtime intentionally uses the same culture-invariant regex semantics as the
     // JS tooling so grouping and backreferences behave consistently across environments.
     private const RegexOptions SupportedPatternRegexOptions = RegexOptions.CultureInvariant;
+    private static readonly Regex ExactDecimalPattern = new(
+        @"^(?<sign>[+-]?)(?:(?<integer>\d+)(?:\.(?<fraction>\d*))?|\.(?<fractionOnly>\d+))(?:[eE](?<exponent>[+-]?\d+))?$",
+        RegexOptions.CultureInvariant | RegexOptions.Compiled);
 
     /// <summary>
     ///     从磁盘加载并解析一个 JSON Schema 文件。
@@ -1409,7 +1413,7 @@ internal static class YamlConfigSchemaValidator
         }
 
         if (constraints.MultipleOf.HasValue &&
-            !IsMultipleOf(numericValue, constraints.MultipleOf.Value))
+            !IsMultipleOf(normalizedValue, numericValue, constraints.MultipleOf.Value))
         {
             throw ConfigLoadExceptionFactory.Create(
                 ConfigLoadFailureKind.ConstraintViolation,
@@ -1745,21 +1749,139 @@ internal static class YamlConfigSchemaValidator
 
     /// <summary>
     ///     判断数值是否满足 <c>multipleOf</c>。
-    ///     双精度浮点比较会在商空间保留一个与商量级相关的微小容差，
-    ///     以对齐 JS 工具侧对 0.1 / 0.01 这类十进制步进的判定方式，
-    ///     避免出现“编辑器通过、运行时拒绝”的跨环境漂移。
+    ///     优先按十进制字面量做精确整倍数判断，
+    ///     以同时避免 0.1 / 0.01 这类十进制步进的伪失败和大数量级非整倍数的伪通过；
+    ///     只有当值超出精确十进制路径时才退回双精度容差比较。
     /// </summary>
+    /// <param name="normalizedValue">用于数值比较的规范化 YAML 标量文本。</param>
     /// <param name="value">当前值。</param>
     /// <param name="divisor">步进约束。</param>
     /// <returns>是否满足整倍数关系。</returns>
-    private static bool IsMultipleOf(double value, double divisor)
+    private static bool IsMultipleOf(string normalizedValue, double value, double divisor)
     {
+        if (TryIsExactDecimalMultiple(normalizedValue, divisor, out var exactResult))
+        {
+            return exactResult;
+        }
+
         var quotient = value / divisor;
         var nearestInteger = Math.Round(quotient);
         var tolerance = 1e-9 * Math.Max(1d, Math.Abs(quotient));
         return Math.Abs(quotient - nearestInteger) <= tolerance;
     }
 
+    /// <summary>
+    ///     尝试按十进制字面量精确判断 <c>multipleOf</c>。
+    ///     该路径直接对齐 YAML / JSON 中常见的有限十进制写法，
+    ///     避免双精度舍入把明显的非整倍数误判为合法。
+    /// </summary>
+    /// <param name="valueText">规范化后的 YAML 数值文本。</param>
+    /// <param name="divisor">Schema 声明的步进约束。</param>
+    /// <param name="isMultiple">精确路径下的判断结果。</param>
+    /// <returns>是否成功进入精确十进制判断路径。</returns>
+    private static bool TryIsExactDecimalMultiple(string valueText, double divisor, out bool isMultiple)
+    {
+        var divisorText = divisor.ToString("R", CultureInfo.InvariantCulture);
+        if (!TryParseExactDecimal(valueText, out var valueSignificand, out var valueScale) ||
+            !TryParseExactDecimal(divisorText, out var divisorSignificand, out var divisorScale) ||
+            divisorSignificand.IsZero)
+        {
+            isMultiple = false;
+            return false;
+        }
+
+        var commonScale = Math.Max(valueScale, divisorScale);
+        var scaledValue = ScaleDecimalSignificand(valueSignificand, valueScale, commonScale);
+        var scaledDivisor = ScaleDecimalSignificand(divisorSignificand, divisorScale, commonScale);
+        isMultiple = scaledValue % scaledDivisor == BigInteger.Zero;
+        return true;
+    }
+
+    /// <summary>
+    ///     将有限十进制或科学计数法文本拆成“整数有效数字 + 十进制位数”形式。
+    ///     这样可以把整倍数判断转成同一尺度下的整数取模，避免浮点误差参与计算。
+    /// </summary>
+    /// <param name="text">待解析的数值文本。</param>
+    /// <param name="significand">去掉小数点后的有效数字。</param>
+    /// <param name="scale">十进制缩放位数；原值等于 <paramref name="significand" /> / 10^<paramref name="scale" />。</param>
+    /// <returns>是否成功解析为有限十进制数。</returns>
+    private static bool TryParseExactDecimal(string text, out BigInteger significand, out int scale)
+    {
+        var match = ExactDecimalPattern.Match(text);
+        if (!match.Success)
+        {
+            significand = BigInteger.Zero;
+            scale = 0;
+            return false;
+        }
+
+        var exponentGroup = match.Groups["exponent"].Value;
+        var exponent = 0;
+        if (!string.IsNullOrEmpty(exponentGroup) &&
+            !int.TryParse(exponentGroup, NumberStyles.Integer, CultureInfo.InvariantCulture, out exponent))
+        {
+            significand = BigInteger.Zero;
+            scale = 0;
+            return false;
+        }
+
+        var integerDigits = match.Groups["integer"].Value;
+        var fractionDigits = match.Groups["fraction"].Success
+            ? match.Groups["fraction"].Value
+            : match.Groups["fractionOnly"].Value;
+        var digits = string.Concat(integerDigits, fractionDigits);
+        if (digits.Length == 0)
+        {
+            digits = "0";
+        }
+
+        digits = digits.TrimStart('0');
+        if (digits.Length == 0)
+        {
+            significand = BigInteger.Zero;
+            scale = 0;
+            return true;
+        }
+
+        scale = checked(fractionDigits.Length - exponent);
+        if (scale < 0)
+        {
+            digits = string.Concat(digits, new string('0', -scale));
+            scale = 0;
+        }
+
+        while (scale > 0 && digits[^1] == '0')
+        {
+            digits = digits[..^1];
+            scale--;
+        }
+
+        significand = BigInteger.Parse(digits, CultureInfo.InvariantCulture);
+        if (match.Groups["sign"].Value == "-")
+        {
+            significand = BigInteger.Negate(significand);
+        }
+
+        return true;
+    }
+
+    /// <summary>
+    ///     将十进制有效数字放大到目标尺度，便于在同一量纲下执行整数取模。
+    /// </summary>
+    /// <param name="significand">原始有效数字。</param>
+    /// <param name="currentScale">当前十进制位数。</param>
+    /// <param name="targetScale">目标十进制位数。</param>
+    /// <returns>放大到目标尺度后的有效数字。</returns>
+    private static BigInteger ScaleDecimalSignificand(BigInteger significand, int currentScale, int targetScale)
+    {
+        if (currentScale == targetScale)
+        {
+            return significand;
+        }
+
+        return significand * BigInteger.Pow(10, targetScale - currentScale);
+    }
+
     /// <summary>
     ///     解析跨表引用目标表名称。
     /// </summary>
diff --git a/docs/zh-CN/game/config-system.md b/docs/zh-CN/game/config-system.md
index 327f8c27..e156bfd2 100644
--- a/docs/zh-CN/game/config-system.md
+++ b/docs/zh-CN/game/config-system.md
@@ -706,7 +706,7 @@ if (MonsterConfigBindings.References.TryGetByDisplayPath("dropItems", out var re
 - `enum`：供运行时校验、VS Code 校验和表单枚举选择复用
 - `minimum` / `maximum`：供运行时校验、VS Code 校验和生成代码 XML 文档复用
 - `exclusiveMinimum` / `exclusiveMaximum`：供运行时校验、VS Code 校验和生成代码 XML 文档复用
-- `multipleOf`：供运行时校验、VS Code 校验、表单 hint 和生成代码 XML 文档复用；当前按运行时与 JS 共用的浮点容差策略判断十进制步进
+- `multipleOf`：供运行时校验、VS Code 校验、表单 hint 和生成代码 XML 文档复用；当前优先按运行时与 JS 共用的十进制精确整倍数判定处理常见十进制步进，并在必要时退回浮点容差兜底
 - `minLength` / `maxLength`：供运行时校验、VS Code 校验和生成代码 XML 文档复用
 - `pattern`：供运行时校验、VS Code 校验、表单提示和生成代码 XML 文档复用；当前按 C# `CultureInvariant` 与 JS Unicode `u` 模式解释，非法模式会在 schema 解析阶段直接报错
 - `minItems` / `maxItems`：供运行时校验、VS Code 校验、表单提示和生成代码 XML 文档复用
diff --git a/tools/gframework-config-tool/src/configValidation.js b/tools/gframework-config-tool/src/configValidation.js
index d9d7d0a8..39dc7786 100644
--- a/tools/gframework-config-tool/src/configValidation.js
+++ b/tools/gframework-config-tool/src/configValidation.js
@@ -486,6 +486,11 @@ function matchesSchemaMultipleOf(scalarValue, multipleOf) {
         return true;
     }
 
+    const exactDecimalResult = tryMatchesExactDecimalMultiple(scalarValue, String(multipleOf));
+    if (exactDecimalResult !== null) {
+        return exactDecimalResult;
+    }
+
     const numericValue = Number(scalarValue);
     const quotient = numericValue / multipleOf;
     const nearestInteger = Math.round(quotient);
@@ -493,6 +498,88 @@ function matchesSchemaMultipleOf(scalarValue, multipleOf) {
     return Math.abs(quotient - nearestInteger) <= tolerance;
 }
 
+/**
+ * Try to evaluate one multipleOf constraint using exact decimal arithmetic.
+ * This keeps common YAML / JSON decimal literals aligned with the runtime and
+ * avoids large-number false positives that a pure floating-point quotient check can miss.
+ *
+ * @param {string} valueText YAML scalar text.
+ * @param {string} divisorText Schema multipleOf text.
+ * @returns {boolean | null} Exact result, or null when the inputs cannot be normalized exactly.
+ */
+function tryMatchesExactDecimalMultiple(valueText, divisorText) {
+    const valueParts = tryParseExactDecimal(valueText);
+    const divisorParts = tryParseExactDecimal(divisorText);
+    if (!valueParts || !divisorParts || divisorParts.significand === 0n) {
+        return null;
+    }
+
+    const commonScale = Math.max(valueParts.scale, divisorParts.scale);
+    const scaledValue = scaleDecimalSignificand(valueParts.significand, valueParts.scale, commonScale);
+    const scaledDivisor = scaleDecimalSignificand(divisorParts.significand, divisorParts.scale, commonScale);
+    return scaledValue % scaledDivisor === 0n;
+}
+
+/**
+ * Normalize a finite decimal literal into an integer significand plus decimal scale.
+ * The normalized form lets multipleOf checks run as integer modulo instead of floating-point math.
+ *
+ * @param {string} text Numeric text to normalize.
+ * @returns {{significand: bigint, scale: number} | null} Normalized parts, or null for unsupported input.
+ */
+function tryParseExactDecimal(text) {
+    const match = /^([+-]?)(?:(\d+)(?:\.(\d*))?|\.(\d+))(?:[eE]([+-]?\d+))?$/u.exec(String(text).trim());
+    if (!match) {
+        return null;
+    }
+
+    const exponent = match[5] ? Number.parseInt(match[5], 10) : 0;
+    if (!Number.isSafeInteger(exponent)) {
+        return null;
+    }
+
+    const integerDigits = match[2] ?? "";
+    const fractionDigits = match[3] !== undefined ? match[3] : (match[4] ?? "");
+    let digits = `${integerDigits}${fractionDigits}`.replace(/^0+/u, "");
+    if (digits.length === 0) {
+        return {significand: 0n, scale: 0};
+    }
+
+    let scale = fractionDigits.length - exponent;
+    if (scale < 0) {
+        digits += "0".repeat(-scale);
+        scale = 0;
+    }
+
+    while (scale > 0 && digits.endsWith("0")) {
+        digits = digits.slice(0, -1);
+        scale -= 1;
+    }
+
+    let significand = BigInt(digits);
+    if (match[1] === "-") {
+        significand = -significand;
+    }
+
+    return {significand, scale};
+}
+
+/**
+ * Scale one normalized decimal significand to a larger decimal precision.
+ *
+ * @param {bigint} significand Integer significand.
+ * @param {number} currentScale Current decimal scale.
+ * @param {number} targetScale Target decimal scale.
+ * @returns {bigint} Scaled significand.
+ */
+function scaleDecimalSignificand(significand, currentScale, targetScale) {
+    if (currentScale === targetScale) {
+        return significand;
+    }
+
+    return significand * (10n ** BigInt(targetScale - currentScale));
+}
+
 /**
  * Format a scalar value for YAML output.
  *
diff --git a/tools/gframework-config-tool/test/configValidation.test.js b/tools/gframework-config-tool/test/configValidation.test.js
index ea105c9b..3b793650 100644
--- a/tools/gframework-config-tool/test/configValidation.test.js
+++ b/tools/gframework-config-tool/test/configValidation.test.js
@@ -349,6 +349,47 @@ phases:
     assert.match(diagnostics[1].message, /phases\[1\]|uniqueItems|元素唯一/u);
 });
 
+test("validateParsedConfig should accept large decimal multiples without floating-point drift", () => {
+    const schema = parseSchemaContent(`
+        {
+          "type": "object",
+          "properties": {
+            "dropRate": {
+              "type": "number",
+              "multipleOf": 0.1
+            }
+          }
+        }
+    `);
+    const yaml = parseTopLevelYaml(`
+dropRate: 10000000.2
+`);
+
+    assert.deepEqual(validateParsedConfig(schema, yaml), []);
+});
+
+test("validateParsedConfig should reject large numbers that are not actually multiples", () => {
+    const schema = parseSchemaContent(`
+        {
+          "type": "object",
+          "properties": {
+            "dropRate": {
+              "type": "number",
+              "multipleOf": 1
+            }
+          }
+        }
+    `);
+    const yaml = parseTopLevelYaml(`
+dropRate: 1000000000000.4
+`);
+
+    const diagnostics = validateParsedConfig(schema, yaml);
+
+    assert.equal(diagnostics.length, 1);
+    assert.match(diagnostics[0].message, /multiple of 1|1 的整数倍/u);
+});
+
 test("validateParsedConfig should accept scientific-notation numbers", () => {
     const schema = parseSchemaContent(`
         {