dsh-livebench-panel 0.1.16 → 0.1.18

package/lib/client.js

@@ -13,6 +13,7 @@ window.__ModuleLoader__.load({
 .dlb_head h2{margin:0;font-size:16px;font-weight:600}
 .dlb_sub{color:var(--dsw-alias-label-tertiary);font-size:12px}
 .dlb_card{border:1px solid var(--dsw-alias-border-l2);background:var(--dsw-alias-bg-layer-2);border-radius:10px;padding:14px;display:flex;flex-direction:column;gap:12px}
+.dlb_fields{display:flex;flex-direction:column;gap:12px}
 .dlb_grid{display:grid;grid-template-columns:repeat(12,1fr);gap:10px}
 @media (max-width:860px){.dlb_grid>.dlb_field{grid-column:1/-1 !important}}
 .dlb_field{display:flex;flex-direction:column;gap:4px;min-width:0}
@@ -51,10 +52,9 @@ window.__ModuleLoader__.load({
 .dlb_chip{appearance:none;font:inherit;cursor:pointer;color:var(--dsw-alias-label-secondary);background:transparent;border:1px solid var(--dsw-alias-border-l2);border-radius:999px;padding:3px 12px;font-size:12px;line-height:18px}
 .dlb_chip:disabled{opacity:.4;cursor:not-allowed}
 .dlb_chipOn{color:var(--dsw-alias-label-primary);border-color:var(--dsw-alias-state-business-primary);background:color-mix(in srgb,var(--dsw-alias-state-business-primary) 14%,transparent)}
-.dlb_actionWrap{display:flex;gap:12px;align-items:stretch}
-.dlb_actionCol{display:flex;flex-direction:column;gap:8px;width:130px;flex-shrink:0}
-.dlb_btnBar{flex:1;min-height:44px;font-weight:600;white-space:normal}
-@media (max-width:760px){.dlb_actionWrap{flex-direction:column}.dlb_actionCol{flex-direction:row;width:100%}.dlb_btnBar{min-height:40px;flex:1}}
+.dlb_btnRow{display:flex;gap:10px;justify-content:center;margin-top:2px}
+.dlb_btnBar{min-width:190px;min-height:40px;font-weight:600;text-align:center}
+@media (max-width:700px){.dlb_btnRow{flex-wrap:wrap}.dlb_btnBar{max-width:none}}
 .dlb_drag{cursor:grab;color:var(--dsw-alias-label-tertiary);text-align:center;user-select:none}`;
 		const tagId = "dsh-livebench-panel/panel.css";
 		if (typeof document !== "undefined" && document.querySelector("style[data-plugin-css=" + JSON.stringify(tagId) + "]") === null) {
@@ -95,537 +95,530 @@ window.__ModuleLoader__.load({
 		}
 
 		function LiveBenchView() {
-			const [config, setConfig] = useState(null);
-			const [configError, setConfigError] = useState(null);
-			const [sel, setSel] = useState({ models: [], efforts: {}, cats: [], tasks: [], release: "2024-11-25", begin: "", end: "", maxTokens: "32000" });
-			const [modelDdOpen, setModelDdOpen] = useState(false);
+					const [config, setConfig] = useState(null);
+					const [configError, setConfigError] = useState(null);
+					const [sel, setSel] = useState({ models: [], efforts: {}, cats: [], tasks: [], release: "2024-11-25", begin: "", end: "", maxTokens: "32000" });
+					const [modelDdOpen, setModelDdOpen] = useState(false);
+					const [busy, setBusy] = useState(false);
+					const [running, setRunning] = useState(false);
+					const [runsList, setRunsList] = useState([]);
+					const [startError, setStartError] = useState(null);
+					const [homeInput, setHomeInput] = useState("");
+					const [homeBusy, setHomeBusy] = useState(false);
+					const [results, setResults] = useState(null);
+					const [pickedModels, setPickedModels] = useState(() => new Set());
+					const [modelOrder, setModelOrder] = useState(() => {
+						try { return JSON.parse(localStorage.getItem("dlb_model_order_v1")) ?? []; } catch { return []; }
+					});
+					const dragKey = useRef(null);
 			const modelDdRef = useRef(null);
-			const [busy, setBusy] = useState(false);
-			const [running, setRunning] = useState(false);
-			const [runsList, setRunsList] = useState([]);
-			const [startError, setStartError] = useState(null);
-			const [homeInput, setHomeInput] = useState("");
-			const [homeBusy, setHomeBusy] = useState(false);
-			const [results, setResults] = useState(null);
-			const [pickedModels, setPickedModels] = useState(() => new Set());
-			const [modelOrder, setModelOrder] = useState(() => {
-				try { return JSON.parse(localStorage.getItem("dlb_model_order_v1")) ?? []; } catch { return []; }
-			});
-			const dragKey = useRef(null);
-			const logRef = useRef(null);
 
-			// 点击面板外部时关闭模型下拉
-			useEffect(() => {
-				if (!modelDdOpen) return;
-				const onDocDown = (event) => {
-					if (modelDdRef.current && !modelDdRef.current.contains(event.target)) setModelDdOpen(false);
-				};
-				document.addEventListener("mousedown", onDocDown);
-				return () => document.removeEventListener("mousedown", onDocDown);
-			}, [modelDdOpen]);
+					// 点击面板外部时关闭模型下拉
+					useEffect(() => {
+						if (!modelDdOpen) return undefined;
+						const onDocDown = (event) => {
+							if (modelDdRef.current && !modelDdRef.current.contains(event.target)) setModelDdOpen(false);
+						};
+						document.addEventListener("mousedown", onDocDown);
+						return () => document.removeEventListener("mousedown", onDocDown);
+					}, [modelDdOpen]);
 
-			const loadConfig = useCallback(async () => {
-				const payload = await api("/config");
-				if (payload.ok) {
-					setConfig(payload);
-					setConfigError(null);
-					setSel((prev) => {
-						const next = { ...prev, release: payload.releases.includes(prev.release) ? prev.release : "2024-11-25" };
-						// 保留仍存在的选择，剔除失效项
-						next.models = prev.models.filter((v) => {
-							const [pid] = v.split("::");
-							const p = payload.providers.find((x) => x.id === pid);
-							return p ? true : false;
-						});
-						return next;
-					});
-				} else {
-					setConfigError(payload.error ?? "config unavailable");
-				}
-			}, []);
-
-			const loadResults = useCallback(async () => {
-				const payload = await api("/results");
-				if (payload.ok) setResults(payload);
-			}, []);
+					const loadConfig = useCallback(async () => {
+						const payload = await api("/config");
+						if (payload.ok) {
+							setConfig(payload);
+							setConfigError(null);
+							setSel((prev) => {
+								const next = { ...prev, release: payload.releases.includes(prev.release) ? prev.release : "2024-11-25" };
+								next.models = prev.models.filter((value) => {
+									const pid = value.split("::")[0];
+									return payload.providers.some((p) => p.id === pid);
+								});
+								return next;
+							});
+						} else {
+							setConfigError(payload.error ?? "config unavailable");
+						}
+					}, []);
 
-			const refreshStatus = useCallback(async () => {
-				const payload = await api("/status");
-				if (payload.ok) {
-					setRunsList(payload.runs ?? []);
-					setRunning(payload.running === true);
-					return payload.running === true;
-				}
-				return false;
-			}, []);
+					const loadResults = useCallback(async () => {
+						const payload = await api("/results");
+						if (payload.ok) setResults(payload);
+					}, []);
 
-			useEffect(() => {
-				loadConfig();
-				loadResults();
-				refreshStatus();
-			}, [loadConfig, loadResults, refreshStatus]);
+					const refreshStatus = useCallback(async () => {
+						const payload = await api("/status");
+						if (payload.ok) {
+							setRunsList(payload.runs ?? []);
+							setRunning(payload.running === true);
+							return payload.running === true;
+						}
+						return false;
+					}, []);
 
-			// poll while a run is active; refresh results when it ends
-			const wasRunning = useRef(false);
-			useEffect(() => {
-				const timer = setInterval(async () => {
-					const active = await refreshStatus();
-					if (wasRunning.current && !active) loadResults();
-					wasRunning.current = active;
-				}, 2500);
-				return () => clearInterval(timer);
-			}, [refreshStatus, loadResults]);
+					useEffect(() => {
+						loadConfig();
+						loadResults();
+						refreshStatus();
+					}, [loadConfig, loadResults, refreshStatus]);
 
-			const categories = useMemo(() => (config ? Object.keys(config.tasks) : []), [config]);
-			// 有效题数：LiveBench 会丢弃「发布晚于所选 release」和「在所选 release
-			// 前已退役」的题目，这里按题目桶 (发布日, 移除日) 精确复算。
-			const countFor = useCallback((category, task) => {
-				if (!config || category.length === 0) return null;
-				const meta = (config.tasks[category] ?? {})[task];
-				if (!meta || !Array.isArray(meta.buckets)) return null;
-				const option = sel.release;
-				return meta.buckets.reduce((sum, b) => {
-					const releasedOk = b.r !== "" && b.r <= option;
-					const notRemoved = b.rm === "" || b.rm > option;
-					return sum + (releasedOk && notRemoved ? b.n : 0);
-				}, 0);
-			}, [config, sel.release]);
-			const catCount = useCallback((category) => {
-				if (!config) return 0;
-				return Object.keys(config.tasks[category] ?? {}).reduce((sum, task) => sum + (countFor(category, task) ?? 0), 0);
-			}, [config, countFor]);
-			// 多选任务 chips：未选分类时展示全部分类的任务
-			const taskChips = useMemo(() => {
-				const cats = sel.cats.length > 0 ? sel.cats : categories;
-				return cats.flatMap((cat) => Object.keys(config?.tasks[cat] ?? {}).map((task) => `${cat}/${task}`));
-			}, [sel.cats, categories, config]);
-			const toggleList = (key, value) => setSel((prev) => {
-				const list = prev[key];
-				const next = list.includes(value) ? list.filter((v) => v !== value) : [...list, value];
-				return { ...prev, [key]: next };
-			});
-			// 已选模型（provider::model 值列表）的解析与 per-model effort
-			const selectedModelEntries = useMemo(() => sel.models.map((value) => {
-				const [pid, ...rest] = value.split("::");
-				const mid = rest.join("::");
-				const provider = config?.providers.find((p) => p.id === pid) ?? null;
-				const model = provider?.models.find((m) => m.id === mid) ?? null;
-				return { providerId: pid, modelId: mid, modelValue: value, efforts: model?.efforts ?? [] };
-			}).filter((entry) => entry.modelId !== ""), [sel.models, config]);
+					// 轮询评测状态；从运行中转为结束后自动刷新成绩
+					const wasRunning = useRef(false);
+					useEffect(() => {
+						const timer = setInterval(async () => {
+							const active = await refreshStatus();
+							if (wasRunning.current && !active) loadResults();
+							wasRunning.current = active;
+						}, 2500);
+						return () => clearInterval(timer);
+					}, [refreshStatus, loadResults]);
 
-			const onStart = async () => {
-				setStartError(null);
-				if (selectedModelEntries.length === 0) {
-					setStartError("请先在下拉框中至少选择一个模型。");
-					return;
-				}
-				const benchNames = [];
-				if (sel.cats.length === 0) {
-					benchNames.push("live_bench");
-				} else if (sel.tasks.length === 0) {
-					for (const cat of sel.cats) benchNames.push(`live_bench/${cat}`);
-				} else {
-					for (const t of sel.tasks) benchNames.push(`live_bench/${t}`);
-				}
-				const zeroTask = sel.tasks.find((t) => countFor(t.split("/")[0], t.split("/")[1]) === 0);
-				if (zeroTask) {
-					setStartError(`任务 ${zeroTask.split("/")[1]} 在 release ${sel.release} 下没有可用题目（该批题目已退役），请取消勾选或换 release。`);
-					return;
-				}
-				setBusy(true);
-				try {
-					// 每个 selected 模型并发启动一个评测（服务端有并发上限保护），
-					// 推理强度按各模型在次级下拉中的选择逐个下发
-					const failures = [];
-					const launches = selectedModelEntries.map(async (entry) => {
-						const payload = await api("/start", {
-							method: "POST",
-							headers: { "content-type": "application/json" },
-							body: JSON.stringify({
-								provider: entry.providerId,
-								model: entry.modelId,
-								reasoningEffort: sel.efforts[entry.modelValue] ?? "default",
-								benchNames,
-								release: sel.release,
-								begin: sel.begin,
-								end: sel.end,
-								maxTokens: sel.maxTokens,
-							}),
-						});
-						if (!payload.ok) failures.push(`${entry.modelId}: ${payload.error ?? "启动失败"}`);
+					const categories = useMemo(() => (config ? Object.keys(config.tasks) : []), [config]);
+					// 有效题数：LiveBench 丢弃「发布晚于所选 release」与「在所选 release 前已退役」的题目
+					const countFor = useCallback((category, task) => {
+						if (!config || category.length === 0) return null;
+						const meta = (config.tasks[category] ?? {})[task];
+						if (!meta || !Array.isArray(meta.buckets)) return null;
+						const option = sel.release;
+						return meta.buckets.reduce((sum, b) => {
+							const releasedOk = b.r !== "" && b.r <= option;
+							const notRemoved = b.rm === "" || b.rm > option;
+							return sum + (releasedOk && notRemoved ? b.n : 0);
+						}, 0);
+					}, [config, sel.release]);
+					const catCount = useCallback((category) => {
+						if (!config) return 0;
+						return Object.keys(config.tasks[category] ?? {}).reduce((sum, task) => sum + (countFor(category, task) ?? 0), 0);
+					}, [config, countFor]);
+					const taskChips = useMemo(() => {
+						const cats = sel.cats.length > 0 ? sel.cats : categories;
+						return cats.flatMap((cat) => Object.keys(config?.tasks[cat] ?? {}).map((task) => `${cat}/${task}`));
+					}, [sel.cats, categories, config]);
+					const toggleList = (key, value) => setSel((prev) => {
+						const list = prev[key];
+						const next = list.includes(value) ? list.filter((v) => v !== value) : [...list, value];
+						return { ...prev, [key]: next };
 					});
-					await Promise.all(launches);
-					if (failures.length > 0) setStartError(failures.join("；"));
-					await refreshStatus();
-				} catch (error) {
-					setStartError(String(error.message ?? error));
-				} finally {
-					setBusy(false);
-				}
-			};
 
-			const onStop = async () => {
-				await api("/stop", { method: "POST" });
-				await refreshStatus();
-			};
+					// 已选模型解析（provider::model）
+					const selectedModelEntries = useMemo(() => sel.models.map((value) => {
+						const sep = value.indexOf("::");
+						const pid = value.slice(0, sep);
+						const mid = value.slice(sep + 2);
+						const provider = config?.providers.find((p) => p.id === pid) ?? null;
+						const model = provider?.models.find((m) => m.id === mid) ?? null;
+						return { providerId: pid, modelId: mid, modelValue: value, efforts: model?.efforts ?? [] };
+					}), [sel.models, config]);
 
-			const saveHome = async () => {
-				setHomeBusy(true);
-				setStartError(null);
-				try {
-					const payload = await api("/home", {
-						method: "POST",
-						headers: { "content-type": "application/json" },
-						body: JSON.stringify({ path: homeInput }),
-					});
-					if (!payload.ok) setStartError(payload.error ?? "设置失败");
-					else { setHomeInput(""); await loadConfig(); }
-				} catch (error) {
-					setStartError(String(error.message ?? error));
-				} finally {
-					setHomeBusy(false);
-				}
-			};
+					const onStart = async () => {
+						setStartError(null);
+						if (selectedModelEntries.length === 0) {
+							setStartError("请先在模型下拉中至少勾选一个模型。");
+							return;
+						}
+						const benchNames = [];
+						if (sel.cats.length === 0) {
+							benchNames.push("live_bench");
+						} else if (sel.tasks.length === 0) {
+							for (const cat of sel.cats) benchNames.push(`live_bench/${cat}`);
+						} else {
+							for (const t of sel.tasks) benchNames.push(`live_bench/${t}`);
+						}
+						const zeroTask = sel.tasks.find((t) => countFor(t.split("/")[0], t.split("/")[1]) === 0);
+						if (zeroTask) {
+							setStartError(`任务 ${zeroTask.split("/")[1]} 在 release ${sel.release} 下没有可用题目（该批题目已退役），请取消勾选或换 release。`);
+							return;
+						}
+						setBusy(true);
+						try {
+							// 每个模型并发启动一个评测（服务端并发上限 6）
+							const failures = [];
+							const launches = selectedModelEntries.map(async (entry) => {
+								const payload = await api("/start", {
+									method: "POST",
+									headers: { "content-type": "application/json" },
+									body: JSON.stringify({
+										provider: entry.providerId,
+										model: entry.modelId,
+										reasoningEffort: sel.efforts[entry.modelValue] ?? "default",
+										benchNames,
+										release: sel.release,
+										begin: sel.begin,
+										end: sel.end,
+										maxTokens: sel.maxTokens,
+									}),
+								});
+								if (!payload.ok) failures.push(`${entry.modelId}: ${payload.error ?? "启动失败"}`);
+							});
+							await Promise.all(launches);
+							if (failures.length > 0) setStartError(failures.join("；"));
+							await refreshStatus();
+						} catch (error) {
+							setStartError(String(error.message ?? error));
+						} finally {
+							setBusy(false);
+						}
+					};
 
-			const setField = (key) => (event) => {
-				const value = event.target.value;
-				setSel((prev) => ({ ...prev, [key]: value }));
-			};
+					const onStop = async () => {
+						await api("/stop", { method: "POST" });
+						await refreshStatus();
+					};
 
-			// ---- 成绩矩阵：model 为行标识、category/task 为列标识 ----
-			// 附带：time 列（该模型评测开始时间 = 行内最早判分时间）、
-			// 行拖拽排序（localStorage 持久化）、行选择与单行/批量删除。
-			const taskKey = (category, task) => `${category}/${task}`;
-			const matrix = useMemo(() => {
-				if (!results) return null;
-				const models = [...new Set(results.rows.map((r) => r.model))].sort();
-				const tasks = [...new Set(results.rows.map((r) => taskKey(r.category, r.task)))].sort();
-				const cells = new Map(); // `${model}\u0000${taskKey}` -> row
-				const startTimes = new Map(); // model -> min tstamp
-				for (const row of results.rows) {
-					cells.set(`${row.model}\u0000${taskKey(row.category, row.task)}`, row);
-					const prev = startTimes.get(row.model);
-					if (row.time && (prev === undefined || row.time < prev)) startTimes.set(row.model, row.time);
-				}
-				return { models, tasks, cells, startTimes };
-			}, [results]);
-			const sortedModels = useMemo(() => {
-				if (!matrix) return [];
-				const indexOf = new Map(modelOrder.map((key, index) => [key, index]));
-				return [...matrix.models].sort((a, b) => {
-					const ia = indexOf.has(a) ? indexOf.get(a) : Number.MAX_SAFE_INTEGER;
-					const ib = indexOf.has(b) ? indexOf.get(b) : Number.MAX_SAFE_INTEGER;
-					if (ia !== ib) return ia - ib;
-					return a.localeCompare(b);
-				});
-			}, [matrix, modelOrder]);
+					const saveHome = async () => {
+						setHomeBusy(true);
+						setStartError(null);
+						try {
+							const payload = await api("/home", {
+								method: "POST",
+								headers: { "content-type": "application/json" },
+								body: JSON.stringify({ path: homeInput }),
+							});
+							if (!payload.ok) setStartError(payload.error ?? "设置失败");
+							else { setHomeInput(""); await loadConfig(); }
+						} catch (error) {
+							setStartError(String(error.message ?? error));
+						} finally {
+							setHomeBusy(false);
+						}
+					};
 
-			const persistModelOrder = (keys) => {
-				setModelOrder(keys);
-				try { localStorage.setItem("dlb_model_order_v1", JSON.stringify(keys)); } catch { /* private mode */ }
-			};
+					const setField = (key) => (event) => {
+						const value = event.target.value;
+						setSel((prev) => ({ ...prev, [key]: value }));
+					};
+					const toggleModel = (value) => setSel((prev) => {
+						const set = new Set(prev.models);
+						const nextEfforts = { ...prev.efforts };
+						if (set.has(value)) { set.delete(value); delete nextEfforts[value]; }
+						else set.add(value);
+						return { ...prev, models: [...set], efforts: nextEfforts };
+					});
+					const setModelEffort = (value) => (event) => {
+						const v = event.target.value;
+						setSel((prev) => ({ ...prev, efforts: { ...prev.efforts, [value]: v } }));
+					};
 
-			const onRowDrop = (targetModel) => {
-				const sourceModel = dragKey.current;
-				dragKey.current = null;
-				if (!sourceModel || sourceModel === targetModel) return;
-				const keys = sortedModels.slice();
-				const from = keys.indexOf(sourceModel);
-				const to = keys.indexOf(targetModel);
-				if (from < 0 || to < 0) return;
-				keys.splice(to, 0, keys.splice(from, 1)[0]);
-				persistModelOrder(keys);
-			};
+					// ---- 成绩矩阵：model 为行标识、category/task 为列标识 ----
+					// time 列 = 该模型评测开始时间（行内最早判分时间）
+					const taskKey = (category, task) => `${category}/${task}`;
+					const matrix = useMemo(() => {
+						if (!results) return null;
+						const models = [...new Set(results.rows.map((r) => r.model))].sort();
+						const tasks = [...new Set(results.rows.map((r) => taskKey(r.category, r.task)))].sort();
+						const cells = new Map();
+						const startTimes = new Map();
+						for (const row of results.rows) {
+							cells.set(`${row.model}__@__${taskKey(row.category, row.task)}`, row);
+							const prev = startTimes.get(row.model);
+							if (row.time && (prev === undefined || row.time < prev)) startTimes.set(row.model, row.time);
+						}
+						return { models, tasks, cells, startTimes };
+					}, [results]);
+					const cellOf = (model, key) => matrix.cells.get(`${model}__@__${key}`);
+					const sortedModels = useMemo(() => {
+						if (!matrix) return [];
+						const indexOf = new Map(modelOrder.map((key, index) => [key, index]));
+						return [...matrix.models].sort((a, b) => {
+							const ia = indexOf.has(a) ? indexOf.get(a) : Number.MAX_SAFE_INTEGER;
+							const ib = indexOf.has(b) ? indexOf.get(b) : Number.MAX_SAFE_INTEGER;
+							if (ia !== ib) return ia - ib;
+							return a.localeCompare(b);
+						});
+					}, [matrix, modelOrder]);
 
-			const togglePicked = (model) => setPickedModels((prev) => {
-				const next = new Set(prev);
-				if (next.has(model)) next.delete(model); else next.add(model);
-				return next;
-			});
+					const persistModelOrder = (keys) => {
+						setModelOrder(keys);
+						try { localStorage.setItem("dlb_model_order_v1", JSON.stringify(keys)); } catch { /* ignore */ }
+					};
 
-			// 删除一个模型行 = 删除该模型在所有任务下的判分与答案记录
-			const deleteModels = async (modelsToDelete) => {
-				if (!results || modelsToDelete.length === 0) return;
-				const serverRows = [];
-				for (const row of results.rows) {
-					if (modelsToDelete.includes(row.model)) {
-						serverRows.push({ model: row.model, category: row.category, task: row.task });
-					}
-				}
-				if (serverRows.length === 0) return;
-				if (!window.confirm(`确认删除所选 ${modelsToDelete.length} 个模型的全部评测成绩（共 ${serverRows.length} 条记录，含答案）？不可恢复。`)) return;
-				await api("/delete", {
-					method: "POST",
-					headers: { "content-type": "application/json" },
-					body: JSON.stringify({ rows: serverRows }),
-				});
-				setPickedModels(new Set());
-				persistModelOrder(modelOrder.filter((model) => !modelsToDelete.includes(model)));
-				await loadResults();
-			};
+					const onRowDrop = (targetModel) => {
+						const sourceModel = dragKey.current;
+						dragKey.current = null;
+						if (!sourceModel || sourceModel === targetModel) return;
+						const keys = sortedModels.slice();
+						const from = keys.indexOf(sourceModel);
+						const to = keys.indexOf(targetModel);
+						if (from < 0 || to < 0) return;
+						keys.splice(to, 0, keys.splice(from, 1)[0]);
+						persistModelOrder(keys);
+					};
 
-			const fmtTime = (t) => {
-				if (!t) return "—";
-				const dt = new Date(t * 1000);
-				const pad = (n) => String(n).padStart(2, "0");
-				return `${dt.getFullYear()}-${pad(dt.getMonth() + 1)}-${pad(dt.getDate())} ${pad(dt.getHours())}:${pad(dt.getMinutes())}`;
-			};
+					const togglePicked = (model) => setPickedModels((prev) => {
+						const next = new Set(prev);
+						if (next.has(model)) next.delete(model); else next.add(model);
+						return next;
+					});
 
-			return h("div", { className: c("root") },
-				h("div", { className: c("head") },
-					h("h2", null, "LiveBench"),
-					h("span", { className: c("sub") }, config?.root ? `LiveBench 评测面板 · ${config.root}` : "LiveBench 评测面板"),
-					h("span", { className: c("badge"), "data-ok": config?.available ? "1" : "0" },
-						config === null ? "检测中…" : config.available ? "LiveBench 就绪" : "未找到 LiveBench"),
-				),
-				h("div", { className: c("row") },
-					h("span", { className: c("label") }, "LiveBench 路径"),
-					h("input", {
-						className: c("input"), style: { flex: "1", minWidth: "240px" },
-						placeholder: config?.root ?? "V:\\…\\LiveBench 或 ~/LiveBench",
-						value: homeInput, onChange: (event) => setHomeInput(event.target.value),
-					}),
-					h("button", { className: c("btnGhost") + " " + c("btn"), disabled: homeBusy || homeInput.trim().length === 0, onClick: saveHome },
-						homeBusy ? "保存中…" : "设置路径"),
-					h("span", { className: c("hint") }, "需含 livebench\\run_livebench.py 与 .venv；也可用环境变量 DSH_LIVEBENCH_HOME"),
-				),
-				configError !== null && h("p", { className: c("error") }, `加载配置失败：${configError}`),
-				h("div", { className: c("card") },
-					h("div", { className: c("actionWrap") },
-						h("div", { className: c("grid"), style: { flex: 1, minWidth: 0 } },
-						h("div", { className: c("field"), ref: modelDdRef, style: { gridColumn: "span 6" } },
-							h("label", { className: c("label") }, "模型（harness 全部模型，可多选；勾选后可在行内单独设定推理强度）"),
-							h("button", {
-								type: "button", className: c("select"), style: { textAlign: "left", width: "100%", overflow: "hidden", textOverflow: "ellipsis", whiteSpace: "nowrap" },
-								onClick: () => setModelDdOpen((v) => !v),
-							}, sel.models.length === 0
-								? "点击选择模型…"
-								: sel.models.length === 1
-									? sel.models[0].split("::")[1]
-									: `已选 ${sel.models.length} 个模型`),
-							modelDdOpen && h("div", {
-								style: {
-									marginTop: "6px", maxHeight: "340px", overflowY: "auto", border: "1px solid var(--dsw-alias-border-l2)",
-									borderRadius: "8px", background: "var(--dsw-alias-bg-layer-1)", padding: "6px",
-								},
-							},
-								(config?.providers ?? []).map((p) => h("div", { key: p.id, style: { marginBottom: "4px" } },
-									h("div", { style: { fontSize: "11px", color: "var(--dsw-alias-label-tertiary)", padding: "3px 4px" } },
-										p.name + (p.routable ? "" : " ·未接LiveBench")),
-									p.models.map((m) => {
-										const value = p.id + "::" + m.id;
-										const checked = sel.models.includes(value);
-										return h("div", { key: value, style: { borderBottom: "1px solid var(--dsw-alias-border-l2)", padding: "3px 4px" } },
-											h("div", {
-												style: { display: "flex", alignItems: "center", gap: "8px", cursor: "pointer" },
-												onClick: () => {
-													setSel((prev) => {
-														const set = new Set(prev.models);
-														const nextEfforts = { ...prev.efforts };
-														if (set.has(value)) { set.delete(value); delete nextEfforts[value]; }
-														else set.add(value);
-														return { ...prev, models: [...set], efforts: nextEfforts };
-													});
-												},
-											},
-												h("input", { type: "checkbox", checked, readOnly: true, style: { cursor: "pointer" } }),
-												h("span", { style: { fontSize: "12.5px", flex: 1 } }, m.name)),
-											checked && m.efforts.length > 0 && h("div", { style: { display: "flex", alignItems: "center", gap: "6px", padding: "4px 0 2px 26px" } },
-												h("span", { style: { fontSize: "11px", color: "var(--dsw-alias-label-tertiary)", whiteSpace: "nowrap" } }, "推理强度"),
-												h("select", {
-													className: c("select"), style: { height: "28px", fontSize: "12px", flex: 1 },
-													value: sel.efforts[value] ?? "default",
-													onClick: (event) => event.stopPropagation(),
-													onChange: (event) => {
-														const v = event.target.value;
-														setSel((prev) => ({ ...prev, efforts: { ...prev.efforts, [value]: v } }));
-													},
-												},
-													h("option", { value: "default" }, "（模型默认）"),
-													m.efforts.map((e) => h("option", { key: e, value: e }, e === "off" ? "off（关闭思考）" : e)))));
-									}))),
+					// 删除一个模型行 = 删除该模型在所有任务下的判分与答案记录
+					const deleteModels = async (modelsToDelete) => {
+						if (!results || modelsToDelete.length === 0) return;
+						const serverRows = [];
+						for (const row of results.rows) {
+							if (modelsToDelete.includes(row.model)) {
+								serverRows.push({ model: row.model, category: row.category, task: row.task });
+							}
+						}
+						if (serverRows.length === 0) return;
+						if (!window.confirm(`确认删除所选 ${modelsToDelete.length} 个模型的全部评测成绩（共 ${serverRows.length} 条记录，含答案）？不可恢复。`)) return;
+						await api("/delete", {
+							method: "POST",
+							headers: { "content-type": "application/json" },
+							body: JSON.stringify({ rows: serverRows }),
+						});
+						setPickedModels(new Set());
+						persistModelOrder(modelOrder.filter((model) => !modelsToDelete.includes(model)));
+						await loadResults();
+					};
+
+					const fmtTime = (t) => {
+						if (!t) return "—";
+						const dt = new Date(t * 1000);
+						const pad = (n) => String(n).padStart(2, "0");
+						return `${dt.getFullYear()}-${pad(dt.getMonth() + 1)}-${pad(dt.getDate())} ${pad(dt.getHours())}:${pad(dt.getMinutes())}`;
+					};
+
+					return h("div", { className: c("root") },
+						h("div", { className: c("head") },
+							h("h2", null, "LiveBench"),
+							h("span", { className: c("sub") }, config?.root ? `LiveBench 评测面板 · ${config.root}` : "LiveBench 评测面板"),
+							h("span", { className: c("badge"), "data-ok": config?.available ? "1" : "0" },
+								config === null ? "检测中…" : config.available ? "LiveBench 就绪" : "未找到 LiveBench"),
 						),
-						h("div", { className: c("field"), style: { gridColumn: "span 2" } },
-							h("label", { className: c("label") }, "题集 release"),
-							h("select", { className: c("select"), value: sel.release, onChange: setField("release") },
-								(config?.releases ?? ["2024-11-25"]).map((r) => h("option", { key: r, value: r }, r))),
+						h("div", { className: c("row") },
+							h("span", { className: c("label") }, "LiveBench 路径"),
+							h("input", {
+								className: c("input"), style: { flex: "1", minWidth: "240px" },
+								placeholder: config?.root ?? "选择或输入 LiveBench 目录",
+								value: homeInput, onChange: (event) => setHomeInput(event.target.value),
+							}),
+							h("button", { className: c("btnGhost") + " " + c("btn"), disabled: homeBusy || homeInput.trim().length === 0, onClick: saveHome },
+								homeBusy ? "保存中…" : "设置路径"),
+							h("span", { className: c("hint") }, "需含 livebench\\run_livebench.py 与 .venv；也可用环境变量 DSH_LIVEBENCH_HOME"),
 						),
-						h("div", { className: c("field"), style: { gridColumn: "span 2" } },
-							h("label", { className: c("label") }, "题目序号范围（可选）"),
-							h("div", { className: c("row") },
-								h("input", { className: c("input"), type: "number", min: 0, placeholder: "起", value: sel.begin, onChange: setField("begin"), style: { width: "50%" } }),
-								h("input", { className: c("input"), type: "number", min: 0, placeholder: "止", value: sel.end, onChange: setField("end"), style: { width: "50%" } }),
+						configError !== null && h("p", { className: c("error") }, `加载配置失败：${configError}`),
+						h("div", { className: c("card") },
+							h("div", { className: c("fields") },
+								h("div", { className: c("field"), ref: modelDdRef },
+									h("label", { className: c("label") }, "模型（harness 全部模型，可多选；勾选后可在行内单独设定推理强度）"),
+									h("button", {
+										type: "button", className: c("select"), style: { textAlign: "left", width: "100%", overflow: "hidden", textOverflow: "ellipsis", whiteSpace: "nowrap" },
+										onClick: () => setModelDdOpen((v) => !v),
+									}, sel.models.length === 0
+										? "点击选择模型…"
+										: sel.models.length === 1
+											? sel.models[0].split("::")[1]
+											: `已选 ${sel.models.length} 个模型`),
+									modelDdOpen && h("div", {
+										style: {
+											marginTop: "6px", maxHeight: "340px", overflowY: "auto", border: "1px solid var(--dsw-alias-border-l2)",
+											borderRadius: "8px", background: "var(--dsw-alias-bg-layer-1)", padding: "6px",
+										},
+									},
+										(config?.providers ?? []).map((p) => h("div", { key: p.id, style: { marginBottom: "4px" } },
+											h("div", { style: { fontSize: "11px", color: "var(--dsw-alias-label-tertiary)", padding: "3px 4px" } },
+												p.name + (p.routable ? "" : " ·未接LiveBench")),
+											p.models.map((m) => {
+												const value = p.id + "::" + m.id;
+												const checked = sel.models.includes(value);
+												return h("div", { key: value, style: { borderBottom: "1px solid var(--dsw-alias-border-l2)", padding: "3px 4px" } },
+													h("div", {
+														style: { display: "flex", alignItems: "center", gap: "8px", cursor: "pointer" },
+														onClick: () => toggleModel(value),
+													},
+														h("input", { type: "checkbox", checked, readOnly: true, style: { cursor: "pointer" } }),
+														h("span", { style: { fontSize: "12.5px", flex: 1 } }, m.name)),
+													checked && m.efforts.length > 0 && h("div", { style: { display: "flex", alignItems: "center", gap: "6px", padding: "4px 0 2px 26px" } },
+														h("span", { style: { fontSize: "11px", color: "var(--dsw-alias-label-tertiary)", whiteSpace: "nowrap" } }, "推理强度"),
+														h("select", {
+															className: c("select"), style: { height: "28px", fontSize: "12px", flex: 1 },
+															value: sel.efforts[value] ?? "default",
+															onClick: (event) => event.stopPropagation(),
+															onChange: setModelEffort(value),
+														},
+															h("option", { value: "default" }, "（模型默认）"),
+															m.efforts.map((e) => h("option", { key: e, value: e }, e === "off" ? "off（关闭思考）" : e)))));
+											}))),
+								),
+								h("div", { className: c("field") },
+									h("label", { className: c("label") }, "题集 release"),
+									h("select", { className: c("select"), value: sel.release, onChange: setField("release") },
+										(config?.releases ?? ["2024-11-25"]).map((r) => h("option", { key: r, value: r }, r))),
+								),
+								),
+								h("div", { className: c("field") },
+									h("label", { className: c("label") }, "max-tokens（默认 32000）"),
+									h("input", { className: c("input"), type: "number", min: 256, max: 32768, value: sel.maxTokens, onChange: setField("maxTokens") }),
+								),
+								h("div", { className: c("field") },
+									h("label", { className: c("label") }, "题目序号范围（可选）"),
+									h("div", { className: c("row") },
+										h("input", { className: c("input"), type: "number", min: 0, placeholder: "起", value: sel.begin, onChange: setField("begin"), style: { width: "50%" } }),
+										h("input", { className: c("input"), type: "number", min: 0, placeholder: "止", value: sel.end, onChange: setField("end"), style: { width: "50%" } }),
+									),
+								h("div", { className: c("field") },
+									h("label", { className: c("label") }, "分类（可多选，不选 = 全部分类；括号内为该 release 下有效题数）"),
+									h("div", { className: c("chips") },
+										categories.map((cat) => {
+											const n = catCount(cat);
+											const active = sel.cats.includes(cat);
+											return h("button", {
+												key: cat, type: "button", className: c("chip") + (active ? " " + c("chipOn") : ""),
+												onClick: () => toggleList("cats", cat), disabled: n === 0,
+											}, `${cat}（${n}）`);
+										})),
+								),
+								sel.cats.length > 0 && h("div", { className: c("field") },
+									h("label", { className: c("label") }, "任务（可多选，不选 = 所选分类的全部任务）"),
+									h("div", { className: c("chips") },
+										taskChips.map((key) => {
+											const [cat, task] = key.split("/");
+											const n = countFor(cat, task) ?? 0;
+											const active = sel.tasks.includes(key);
+											return h("button", {
+												key, type: "button", className: c("chip") + (active ? " " + c("chipOn") : ""),
+												onClick: () => toggleList("tasks", key), disabled: n === 0,
+												title: n === 0 ? "该任务在此 release 下已无可用题目" : `${n} 题有效`,
+											}, `${task}（${n}）`);
+										})),
+								),
 							),
+							h("div", { className: c("btnRow") },
+								h("button", {
+									className: c("btn") + " " + c("btnBar"), onClick: onStart,
+									disabled: busy || !config?.available || sel.models.length === 0,
+								}, `开始评测（${sel.models.length} 个模型并发）`),
+								running && h("button", { className: c("btn") + " " + c("btnDanger") + " " + c("btnBar"), onClick: onStop }, "停止全部"),
+								h("button", { className: c("btnGhost") + " " + c("btn") + " " + c("btnBar"), onClick: () => { loadConfig(); loadResults(); } }, "刷新"),
+							),
+							startError !== null && h("p", { className: c("error") }, startError),
 						),
-						h("div", { className: c("field"), style: { gridColumn: "span 2" } },
-							h("label", { className: c("label") }, "max-tokens（默认 32000）"),
-							h("input", { className: c("input"), type: "number", min: 256, max: 32768, value: sel.maxTokens, onChange: setField("maxTokens") }),
-						),
-						h("div", { className: c("field"), style: { gridColumn: "1 / -1" } },
-							h("label", { className: c("label") }, "分类（可多选，不选 = 全部分类；括号内为该 release 下有效题数）"),
-							h("div", { className: c("chips") },
-								categories.map((cat) => {
-									const n = catCount(cat);
-									const active = sel.cats.includes(cat);
-									return h("button", {
-										key: cat, type: "button", className: c("chip") + (active ? " " + c("chipOn") : ""),
-										onClick: () => toggleList("cats", cat), disabled: n === 0,
-									}, `${cat}（${n}）`);
-								})),
-						),
-						sel.cats.length > 0 && h("div", { className: c("field"), style: { gridColumn: "1 / -1" } },
-							h("label", { className: c("label") }, "任务（可多选，不选 = 所选分类的全部任务）"),
-							h("div", { className: c("chips") },
-								taskChips.map((key) => {
-									const [cat, task] = key.split("/");
-									const n = countFor(cat, task) ?? 0;
-									const active = sel.tasks.includes(key);
-									return h("button", {
-										key, type: "button", className: c("chip") + (active ? " " + c("chipOn") : ""),
-										onClick: () => toggleList("tasks", key), disabled: n === 0,
-										title: n === 0 ? "该任务在此 release 下已无可用题目" : `${n} 题有效`,
-									}, `${task}（${n}）`);
-								})),
-						),
-					),
-						h("div", { className: c("actionCol") },
-							h("button", {
-								className: c("btn") + " " + c("btnBar"), onClick: onStart,
-								disabled: busy || !config?.available || sel.models.length === 0,
-							}, `开始评测（${sel.models.length}）`),
-							running && h("button", { className: c("btn") + " " + c("btnDanger") + " " + c("btnBar"), onClick: onStop }, "停止全部"),
-							h("button", { className: c("btnGhost") + " " + c("btn") + " " + c("btnBar"), onClick: () => { loadConfig(); loadResults(); } }, "刷新"),
+						runsList.length > 0 && h("div", { className: c("card") },
+							h("div", { className: c("row") },
+								h("span", { className: c("badge"), "data-ok": running ? "1" : "0" },
+									running ? "有评测运行中" : "无运行中的评测"),
+							),
+							runsList.map((r) => h("details", { key: r.runId, open: r.running },
+								h("summary", { style: { cursor: "pointer", fontSize: "12px", color: "var(--dsw-alias-label-secondary)" } },
+									h("span", { className: c("badge"), "data-ok": r.running ? "1" : "0" }, r.running ? "运行中" : r.exitCode === 0 ? "完成" : `退出(${r.exitCode})`),
+									" ",
+									r.displayName),
+								h("pre", { className: c("log") }, r.log || "（暂无输出）"),
+							))),
 						),
-					),
-					startError !== null && h("p", { className: c("error") }, startError),
-				),
-				runsList.length > 0 && h("div", { className: c("card") },
-					h("div", { className: c("row") },
-						h("span", { className: c("badge"), "data-ok": running ? "1" : "0" },
-							running ? "有评测运行中" : "无运行中的评测"),
-					),
-					runsList.map((r) => h("details", { key: r.runId, open: r.running },
-						h("summary", { style: { cursor: "pointer", fontSize: "12px", color: "var(--dsw-alias-label-secondary)" } },
-							h("span", { className: c("badge"), "data-ok": r.running ? "1" : "0" }, r.running ? "运行中" : r.exitCode === 0 ? "完成" : `退出(${r.exitCode})`),
-							" ",
-							r.displayName),
-						h("pre", { className: c("log"), ref: r.running ? logRef : null }, r.log || "（暂无输出）"),
-					)))),
-				sortedModels.length > 0 && h("div", { className: c("card") },
-					h("div", { className: c("row") },
-						h("span", { className: c("label") }, "评测成绩（model 为行；分数 = 平均分 ×100；行首可拖动排序，顺序保存在本机）"),
-						h("button", {
-							className: c("btn") + " " + c("btnDanger"),
-							disabled: pickedModels.size === 0,
-							onClick: () => deleteModels([...pickedModels]),
-						}, `删除所选（${pickedModels.size}）`),
-					),
-					h("table", { className: c("table") },
-						h("thead", null, h("tr", null,
-							h("th", { style: { width: "26px" } }, "⠿"),
-							h("th", { style: { width: "30px" } },
-								h("input", { type: "checkbox", checked: pickedModels.size > 0 && pickedModels.size === sortedModels.length,
-									onChange: (event) => setPickedModels(event.target.checked ? new Set(sortedModels) : new Set()) })),
-							h("th", null, "model"),
-							matrix.tasks.map((taskKeyCol) => {
-								const [cat, task] = taskKeyCol.split("/");
-								return h("th", { key: taskKeyCol, title: `${cat} / ${task}` }, task);
-							}),
-							h("th", null, "删除"),
-							h("th", null, "time"),
-						)),
-						h("tbody", null,
-							sortedModels.map((model) => {
-								const checked = pickedModels.has(model);
-								const startTime = matrix.startTimes.get(model);
-								return h("tr", {
-									key: model,
-									draggable: true,
-									onDragStart: () => { dragKey.current = model; },
-									onDragOver: (event) => event.preventDefault(),
-									onDrop: () => onRowDrop(model),
-								},
-									h("td", { className: c("drag"), title: "拖动排序" }, "⠿"),
-									h("td", null, h("input", { type: "checkbox", checked, onChange: () => togglePicked(model) })),
-									h("td", null, model),
+						sortedModels.length > 0 && h("div", { className: c("card") },
+							h("div", { className: c("row") },
+								h("span", { className: c("label") }, "评测成绩（model 为行；分数 = 平均分 ×100；行首可拖动排序，顺序保存在本机）"),
+								h("button", {
+									className: c("btn") + " " + c("btnDanger"),
+									disabled: pickedModels.size === 0,
+									onClick: () => deleteModels([...pickedModels]),
+								}, `删除所选（${pickedModels.size}）`),
+							),
+							h("table", { className: c("table") },
+								h("thead", null, h("tr", null,
+									h("th", { style: { width: "26px" } }, "⠿"),
+									h("th", { style: { width: "30px" } },
+										h("input", { type: "checkbox", checked: pickedModels.size > 0 && pickedModels.size === sortedModels.length,
+											onChange: (event) => setPickedModels(event.target.checked ? new Set(sortedModels) : new Set()) })),
+									h("th", null, "model"),
 									matrix.tasks.map((taskKeyCol) => {
-										const row = matrix.cells.get(`${model}\u0000${taskKeyCol}`);
-										// ×= 该模型未做过此任务；E = 做了但全部 API 失败
-										let cellText = "×";
-										let title = "该模型未执行此任务";
-										if (row) {
-											if (row.answered > 0 && row.errors >= row.answered) {
-												cellText = `E (${row.errors})`;
-												title = "该任务全部回答均 API 失败（$ERROR$），计 0 分";
-											} else if (row.judged === 0) {
-												cellText = "未判分";
-												title = "有答案但尚未判分";
-											} else {
-												cellText = `${row.score.toFixed(1)} (${row.judged}/${row.total})`;
-												if (row.errors > 0) title += ` · 其中 ${row.errors} 题 API 失败计 0 分`;
-											}
-										}
-										return h("td", { key: taskKeyCol, className: c("score"), title }, cellText);
+										const [cat, task] = taskKeyCol.split("/");
+										return h("th", { key: taskKeyCol, title: `${cat} / ${task}` }, task);
 									}),
-									h("td", null, h("button", {
-										className: c("btnGhost") + " " + c("btn"), title: "删除该模型的所有评测成绩",
-										onClick: () => deleteModels([model]),
-									}, "删除")),
-									h("td", null, fmtTime(startTime)),
-								);
-							})),
-					),
-				),
-				config !== null && config.available === false && h("div", { className: c("card") + " " + c("notice") },
-					h("p", { className: c("noticeTitle") }, "⚠ 本面板需要本地安装 LiveBench（当前未检测到）"),
-					h("p", { className: c("hint") },
-						"dsh-livebench-panel 只是控制台，真正的评测由 LiveBench 项目完成。面板会在以下位置查找它：环境变量 DSH_LIVEBENCH_HOME，或默认路径 V:\\PythonProject\\C_UtilizeSpace\\LiveBench（需含 .venv 与 livebench\\run_livebench.py）。按下面步骤安装："),
-					h("ol", { className: c("steps") },
-						h("li", null, "克隆 LiveBench 仓库（放到默认路径可免配置环境变量）：",
-							h("code", { className: c("code") }, "git clone https://github.com/LiveBench/LiveBench V:\\PythonProject\\C_UtilizeSpace\\LiveBench")),
-						h("li", null, "用 Python 3.11 建虚拟环境并安装（3.10 会因 litellm 报 NotRequired 错误）：",
-							h("code", { className: c("code") }, "cd /d V:\\PythonProject\\C_UtilizeSpace\\LiveBench && py -3.11 -m venv .venv"),
-							h("code", { className: c("code") }, ".venv\\Scripts\\python -m pip install -e .")),
-						h("li", null, "（评测 coding 类才需要）安装评分依赖：",
-							h("code", { className: c("code") }, ".venv\\Scripts\\python -m pip install -r livebench\\code_runner\\requirements_eval.txt")),
-						h("li", null, "下载题目数据：",
-							h("code", { className: c("code") }, "cd livebench && ..\\.venv\\Scripts\\python download_questions.py")),
-						h("li", null, "重启 dsh web，回到本页点「刷新」。",
-							h("br", null),
-							"若装在其它目录：设置系统环境变量 ", h("code", { className: c("code") }, "DSH_LIVEBENCH_HOME=你的LiveBench目录"), " 后再重启 dsh web。"),
-					),
-					h("p", { className: c("hint") },
-						"除 LiveBench 外无需其它配置：评测所需的 API Key 会自动从 harness 凭据库读取并注入；npm 安装本插件时可一并执行 dsh plugin --profile web add dsh-livebench-panel。完整说明见插件目录内 README：~\\.dsh\\plugins\\dsh-livebench-panel\\README.md"),
-				),
-				h("details", { className: c("details") },
-					h("summary", null, "❓ 使用说明 · 选项含义与选择建议（点击展开）"),
-					h("p", { className: c("helpP") },
-						"操作流程：选择参数 → 点「开始评测」→ 日志区实时显示运行进度 → 结束后成绩表自动刷新（运行中可「停止」）。全部判分均为客观比对（文本/符号执行/测试用例），不使用 AI 评分。"),
-					h("table", { className: c("helpTable") },
-						h("thead", null, h("tr", null, h("th", null, "选项"), h("th", null, "含义与选择建议"))),
-						h("tbody", null,
-							h("tr", null, h("th", null, "模型"), h("td", null,
-								"harness 全部 provider 的全部模型（含内置 DeepSeek 官方）。API Key 自动从 harness 凭据库读取并注入，无需手工配置；名字带「·未接LiveBench」的 provider 无法自动路由，评测会按模型名原生尝试。")),
-							h("tr", null, h("th", null, "推理强度"), h("td", null,
-								"模型的思考深度（off=关闭思考）。强度会编码进条目名（如 glm-5.3@max），不同强度在成绩表中是独立条目，方便对比。想省成本选 low，追求质量选 high/max。")),
-							h("tr", null, h("th", null, "题集 release"), h("td", null,
-								"题目发布批次。推荐 2024-11-25（公开题目最全）。LiveBench 每月换题：新批次下老任务会陆续退役，任务下拉括号内就是该批次下的有效题数。")),
-							h("tr", null, h("th", null, "分类 / 任务"), h("td", null,
-								"六大类共 18 个任务：coding（代码生成/补全）、math（竞赛数学等）、reasoning（空间推理/逻辑谜题）、language（拼写/连线/语义）、data_analysis（表格操作）、instruction_following（指令遵循）。首次验证推荐 language → typos。")),
-							h("tr", null, h("th", null, "题目序号范围"), h("td", null,
-								"从 0 起。冒烟测试填 0 到 2（只跑 3 题）；2 题得分噪声很大（对一题就是 0↔100 的波动），想看真实水平建议 20 题以上。")),
-							h("tr", null, h("th", null, "max-tokens"), h("td", null,
-								"单题回答的 token 上限，默认 32000。推理模型思考也占 token，不要低于 8192，否则思考被截断、答案为空会记 0 分。")),
+									h("th", null, "删除"),
+									h("th", null, "time"),
+								)),
+								h("tbody", null,
+									sortedModels.map((model) => {
+										const checked = pickedModels.has(model);
+										const startTime = matrix.startTimes.get(model);
+										return h("tr", {
+											key: model,
+											draggable: true,
+											onDragStart: () => { dragKey.current = model; },
+											onDragOver: (event) => event.preventDefault(),
+											onDrop: () => onRowDrop(model),
+										},
+											h("td", { className: c("drag"), title: "拖动排序" }, "⠿"),
+											h("td", null, h("input", { type: "checkbox", checked, onChange: () => togglePicked(model) })),
+											h("td", null, model),
+											matrix.tasks.map((taskKeyCol) => {
+												const row = matrix.cells.get(`${model}__@__${taskKeyCol}`);
+												let cellText = "×";
+												let title = "该模型未执行此任务";
+												if (row) {
+													if (row.answered > 0 && row.errors >= row.answered) {
+														cellText = `E (${row.errors})`;
+														title = "该任务全部回答均 API 失败（$ERROR$），计 0 分";
+													} else if (row.judged === 0) {
+														cellText = "未判分";
+														title = "有答案但尚未判分";
+													} else {
+														cellText = `${row.score.toFixed(1)} (${row.judged}/${row.total})`;
+														if (row.errors > 0) title += ` · 其中 ${row.errors} 题 API 失败计 0 分`;
+													}
+												}
+												return h("td", { key: taskKeyCol, className: c("score"), title }, cellText);
+											}),
+											h("td", null, h("button", {
+												className: c("btnGhost") + " " + c("btn"), title: "删除该模型的所有评测成绩",
+												onClick: () => deleteModels([model]),
+											}, "删除")),
+											h("td", null, fmtTime(startTime)),
+										);
+									})),
+							),
 						),
-					),
-					h("p", { className: c("helpP"), style: { marginTop: "10px" } },
-						"提示：回答为 $ERROR$ 表示该题 API 调用失败（网络/鉴权/参数问题）计 0 分，可在上方日志区查看具体错误；zebra_puzzle（逻辑谜题）是公认最难的任务，低分属正常现象。"),
-				),
-				h("p", { className: c("hint") },
-					"评分读取 LiveBench 的 ground_truth_judgment.jsonl；正式榜单可用 release 2024-11-25。多选题集请在「分类/任务」中缩小范围，避免长时间运行。"),
-			);
-		}
-
+						config !== null && config.available === false && h("div", { className: c("card") + " " + c("notice") },
+							h("p", { className: c("noticeTitle") }, "⚠ 本面板需要本地安装 LiveBench（当前未检测到）"),
+							h("p", { className: c("hint") },
+								"dsh-livebench-panel 只是控制台，真正的评测由 LiveBench 项目完成。面板会依次查找：本页「LiveBench 路径」输入框保存的目录、环境变量 DSH_LIVEBENCH_HOME、用户目录 ~/LiveBench（需含 .venv 与 livebench\\run_livebench.py）。按下面步骤安装："),
+							h("ol", { className: c("steps") },
+								h("li", null, "克隆 LiveBench 仓库：",
+									h("code", { className: c("code") }, "git clone https://github.com/LiveBench/LiveBench V:\\PythonProject\\C_UtilizeSpace\\LiveBench")),
+								h("li", null, "用 Python 3.11 建虚拟环境并安装（3.10 会因 litellm 报 NotRequired 错误）：",
+									h("code", { className: c("code") }, "cd /d V:\\PythonProject\\C_UtilizeSpace\\LiveBench && py -3.11 -m venv .venv"),
+									h("code", { className: c("code") }, ".venv\\Scripts\\python -m pip install -e .")),
+								h("li", null, "（评测 coding 类才需要）安装评分依赖：",
+									h("code", { className: c("code") }, ".venv\\Scripts\\python -m pip install -r livebench\\code_runner\\requirements_eval.txt")),
+								h("li", null, "下载题目数据：",
+									h("code", { className: c("code") }, "cd livebench && ..\\.venv\\Scripts\\python download_questions.py")),
+								h("li", null, "重启 dsh web，回到本页点「刷新」。",
+									h("br", null),
+									"若装在其它目录：在本页顶部「LiveBench 路径」输入框填入目录并点「设置路径」即可，也可用环境变量 DSH_LIVEBENCH_HOME。"),
+							),
+							h("p", { className: c("hint") },
+								"除 LiveBench 外无需其它配置：评测所需的 API Key 会自动从 harness 凭据库读取并注入；npm 安装本插件时可一并执行 dsh plugin --profile web add dsh-livebench-panel。完整说明见插件目录内 README：~\\.dsh\\plugins\\dsh-livebench-panel\\README.md"),
+						),
+						h("details", { className: c("details") },
+							h("summary", null, "❓ 使用说明 · 选项含义与选择建议（点击展开）"),
+							h("p", { className: c("helpP") },
+								"操作流程：选择参数 → 点「开始评测」→ 日志区实时显示运行进度 → 结束后成绩表自动刷新（运行中可「停止」）。全部判分均为客观比对（文本/符号执行/测试用例），不使用 AI 评分。"),
+							h("table", { className: c("helpTable") },
+								h("thead", null, h("tr", null, h("th", null, "选项"), h("th", null, "含义与选择建议"))),
+								h("tbody", null,
+									h("tr", null, h("th", null, "模型"), h("td", null,
+										"harness 全部 provider 的全部模型（含内置 DeepSeek 官方）。API Key 自动从 harness 凭据库读取并注入，无需手工配置；名字带「·未接LiveBench」的 provider 无法自动路由，评测会按模型名原生尝试。")),
+									h("tr", null, h("th", null, "推理强度"), h("td", null,
+										"在模型下拉中勾选模型后，于该行内的次级下拉单独设定；强度会编码进条目名（如 glm-5.3@max），不同强度在成绩表中是独立条目，方便对比。")),
+									h("tr", null, h("th", null, "题集 release"), h("td", null,
+										"题目发布批次。推荐 2024-11-25（公开题目最全）。LiveBench 每月换题：新批次下老任务会陆续退役，任务下拉括号内就是该批次下的有效题数。")),
+									h("tr", null, h("th", null, "分类 / 任务"), h("td", null,
+										"六大类共 18 个任务：coding（代码生成/补全）、math（竞赛数学等）、reasoning（空间推理/逻辑谜题）、language（拼写/连线/语义）、data_analysis（表格操作）、instruction_following（指令遵循）。首次验证推荐 language → typos。")),
+									h("tr", null, h("th", null, "题目序号范围"), h("td", null,
+										"从 0 起。冒烟测试填 0 到 2（只跑 3 题）；2 题得分噪声很大（对一题就是 0↔100 的波动），想看真实水平建议 20 题以上。")),
+									h("tr", null, h("th", null, "max-tokens"), h("td", null,
+										"单题回答的 token 上限，默认 32000。推理模型思考也占 token，不要低于 8192，否则思考被截断、答案为空会记 0 分。")),
+								),
+							),
+							h("p", { className: c("helpP"), style: { marginTop: "10px" } },
+								"提示：回答为 $ERROR$ 表示该题 API 调用失败（网络/鉴权/参数问题）计 0 分，可在上方日志区查看具体错误；zebra_puzzle（逻辑谜题）是公认最难的任务，低分属正常现象。"),
+						),
+						h("p", { className: c("hint") },
+							"评分读取 LiveBench 的 ground_truth_judgment.jsonl；正式榜单可用 release 2024-11-25。多选题集请在「分类/任务」中缩小范围，避免长时间运行。"),
+					);
+				}
 		/**
 		 * Mount the LiveBench tab into the Trajectory view.
 		 * @param ctx - the browser plugin context.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-livebench-panel",
-  "version": "0.1.16",
+  "version": "0.1.18",
   "description": "DSH web plugin: a LiveBench tab in the Trajectory view (right of 对话/轨迹). Run LiveBench evaluations against every model configured in the DeepSeek Harness — pick provider/model, category, task, release and question range from dropdowns, watch progress, and read scores in place.",
   "license": "MIT",
   "type": "module",