diff --git a/backend/app/services/news_service.py b/backend/app/services/news_service.py
index bba85b0..bd1b835 100755
--- a/backend/app/services/news_service.py
+++ b/backend/app/services/news_service.py
@@ -508,7 +508,7 @@ class NewsPlatformService:
             logger.warning(f"[广场新闻-今日] {user.account} 请求异常: {e}")
 
         if today_articles:
-            # 今日文章：从新到旧排序（createTime 降序）
+            # 今日文章：从新到旧排序
             today_articles.sort(
                 key=lambda x: x.get("publishTime") or x.get("createTime") or x.get("pushTime") or "",
                 reverse=True
@@ -520,24 +520,93 @@ class NewsPlatformService:
                 if aid and aid not in seen:
                     seen.add(aid)
                     unique.append(a)
-            logger.info(
-                f"[广场新闻] {user.account} 今日新文章 {len(unique)} 篇，"
-                f"顺序轮询（最多取 {count} 篇）"
-            )
-            candidates = unique[:max(count * 2, 20)]   # 取候选池，稍多于 count 以备校验失败
 
-            # 有效性校验后返回
-            valid = []
-            for a in candidates:
-                if len(valid) >= count:
-                    break
+            # 有效性校验今日文章
+            today_valid = []
+            for a in unique:
                 aid = str(a.get("recordId") or a.get("id", ""))
                 if await self.validate_article(db, user, aid):
-                    valid.append(a)
-            if valid and not force_history:
-                return valid
-            if not valid:
-                logger.info(f"[广场新闻] {user.account} 今日文章校验后全部无效，转历史")
+                    today_valid.append(a)
+
+            if today_valid:
+                # ── 混合策略：今日文章不足时，补充历史文章，避免所有用户扎堆同一篇 ──
+                # 每个用户最多分配 1 篇今日文章，其余名额从历史文章随机补充
+                today_quota = min(len(today_valid), max(1, count // 3))
+                today_pick  = today_valid[:today_quota]
+                today_ids   = {str(a.get("recordId") or a.get("id","")) for a in today_pick}
+                hist_needed = count - today_quota
+
+                logger.info(
+                    f"[广场新闻] {user.account} 今日有效 {len(today_valid)} 篇，"
+                    f"取 {today_quota} 篇今日 + {hist_needed} 篇历史"
+                )
+
+                hist_valid = []
+                if hist_needed > 0:
+                    import math as _math2
+                    # 拉历史页（与 Phase 2 相同的小时分页逻辑）
+                    total_pages_h = 1
+                    page1_items_h = []
+                    try:
+                        async with httpx.AsyncClient(timeout=10) as _hc:
+                            _hr = await _hc.get(
+                                f"{biz}/business/square/list",
+                                headers=self._bearer(token),
+                                params=_build(1),
+                            )
+                        _hd = _hr.json()
+                        if _hd.get("code") in [0, 200]:
+                            _hnd = _hd.get("data", {})
+                            total_pages_h = max(1, _math2.ceil(_hnd.get("totalSize", 0) / 50))
+                            page1_items_h = _filter(
+                                _hnd.get("data") or _hnd.get("list") or _hnd.get("records") or []
+                            )
+                    except Exception:
+                        pass
+
+                    max_pages_h = min(total_pages_h, 10)
+                    hour_page_h = (_dt.now().hour % max_pages_h) + 1
+                    hist_pool   = page1_items_h if hour_page_h == 1 else []
+
+                    if hour_page_h != 1:
+                        try:
+                            async with httpx.AsyncClient(timeout=12) as _hc2:
+                                _hr2 = await _hc2.get(
+                                    f"{biz}/business/square/list",
+                                    headers=self._bearer(token),
+                                    params=_build(hour_page_h),
+                                )
+                            if _hr2.status_code == 200:
+                                _hd2 = _hr2.json()
+                                if _hd2.get("code") in [0, 200]:
+                                    _hnd2 = _hd2.get("data", {})
+                                    hist_pool = _filter(
+                                        _hnd2.get("data") or _hnd2.get("list") or _hnd2.get("records") or []
+                                    )
+                        except Exception:
+                            hist_pool = page1_items_h  # fallback to page 1
+
+                    # 排除今日文章，避免重复
+                    hist_pool = [x for x in hist_pool
+                                 if (x.get("recordId") or x.get("id","")) not in today_ids
+                                 and not _is_today(x)]
+                    _rand.shuffle(hist_pool)
+
+                    for a in hist_pool:
+                        if len(hist_valid) >= hist_needed:
+                            break
+                        aid = str(a.get("recordId") or a.get("id", ""))
+                        if await self.validate_article(db, user, aid):
+                            hist_valid.append(a)
+
+                result = today_pick + hist_valid
+                logger.info(
+                    f"[广场新闻] {user.account} 返回 {len(result)} 篇"
+                    f"（今日 {len(today_pick)} + 历史 {len(hist_valid)}）"
+                )
+                return result
+
+            logger.info(f"[广场新闻] {user.account} 今日文章校验后全部无效，转历史")
 
         # ── Phase 2: 无今日新文章 → 从最新(第1页)开始往旧顺序遍历 ────
         # 规则：始终从第1页（最新）开始，按页顺序 1→2→3...往旧方向走