计算机使用智能体分类English
R04.01.01.01 · 任务小类

多跳难找事实

Hard-to-find multi-hop facts

Find a precise fact that requires chaining several sources (BrowseComp-style).

20 条本项目编写的提案 · 分类 v1.1 · 新增基准运行 0 次。下方验证方式与部分得分规则是设计,不是执行结果。

典型应用
Chrome · Safari · search engines
小类 GUI 需求标签
AP · API 足够

任务提案 · 20

展开查看完整原文与建议检查方式;两种界面均保留提案英文原文。

R04.01.01.01-T01 Identify a film from oblique clues

原始指令

Find the feature film that: was a director's narrative debut after two documentaries, won a cinematography award at a major 2010s festival, and features a lead actor who had previously only done television. Write the film title and release year to answer.txt with the source URLs you used.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单 vis · 视觉判断
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
3 / 5

建议验证方式

answer.txt title+year equal the single film matching all three constraints (lab-fixed answer); at least two distinct supporting URLs listed. Graded by exact title match + citation presence.

建议部分得分

40% narrowed to correct candidate; 40% exact title+year; 20% evidence URLs

R04.01.01.01-T02 Find a paper by author alma-mater chain

原始指令

Find the NLP conference paper (2018-2023) whose first author did their undergraduate degree at a named small liberal-arts college and whose last author completed a PhD at a specific European university in a given year. Write the paper title and venue to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
long · 长流程
提案难度
5 / 5

建议验证方式

answer.txt title+venue match the unique lab-fixed paper satisfying both author constraints; citing URLs included. Graded by exact title match + citations.

建议部分得分

30% correct author pair identified; 30% paper candidate; 30% exact title+venue; 10% URLs

R04.01.01.01-T03 Pin down a sports match from constraints

原始指令

Identify the top-flight football match played between 1990 and 1999 that: was refereed by an official of a specified nationality, had exactly two red cards, and drew an attendance in a given range at a named stadium. Write the two teams and the exact date to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
4 / 5

建议验证方式

answer.txt teams+date equal the unique match meeting all constraints; supporting URLs present. Graded by exact match + citations.

建议部分得分

40% constraints combined; 40% exact teams+date; 20% URLs

R04.01.01.01-T04 Locate a town meeting several constraints

原始指令

Find the European town that sits on a named river, has a population under 10,000, hosts an annual festival founded in a given decade, and is the birthplace of a notable 19th-century composer. Write the town and country to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
3 / 5

建议验证方式

answer.txt town+country equal the unique lab-fixed match; citing URLs included. Graded by exact match + citations.

建议部分得分

40% candidate narrowed; 40% exact town+country; 20% URLs

R04.01.01.01-T05 Date an event via a chained fact

原始指令

A bridge was opened the same year a named explorer reached a particular landmark, in a city that later hosted a specified world's fair. Find the bridge's exact opening date. Write the date to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
3 / 5

建议验证方式

answer.txt date equals the lab-fixed opening date; the chain's intermediate facts are implied by the cited URLs. Graded by exact date + citations.

建议部分得分

30% explorer-year found; 30% city/bridge identified; 30% exact date; 10% URLs

R04.01.01.01-T06 Identify an album by personnel and chart

原始指令

Find the studio album that reached a specified chart position in a given year, was produced by a named producer, and featured a session drummer who also played on a famous earlier record. Write the album and artist to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
4 / 5

建议验证方式

answer.txt album+artist equal the unique match for all three constraints; citing URLs present. Graded by exact match + citations.

建议部分得分

40% constraints combined; 40% exact album+artist; 20% URLs

R04.01.01.01-T07 Find a railway station by constraints

原始指令

Identify the railway station that is the terminus of a named heritage line, opened in a given year, and is the closest station to a specific castle. Write the station name to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
3 / 5

建议验证方式

answer.txt station equals the unique lab-fixed match; citing URLs included. Graded by exact match + citations.

建议部分得分

50% station candidate; 30% exact name; 20% URLs

R04.01.01.01-T08 Identify a compound by discovery chain

原始指令

Find the chemical compound first synthesized by a chemist who studied under a named Nobel laureate, in a decade when that laureate held a chair at a specified university. Write the compound's common name and formula to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
3 / 5

建议验证方式

answer.txt compound name+formula equal the lab-fixed answer; citing URLs present. Graded by exact match + citations.

建议部分得分

40% chemist/era chain; 40% exact compound+formula; 20% URLs

R04.01.01.01-T09 Identify a TV drama from Chinese-language sources

原始指令

Using Chinese-language sources, identify the TV drama that aired in a given year, was filmed in a named city, starred a lead actor who later won a specified award, and ran for a stated number of episodes. Write the drama's Chinese and English titles to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单 vis · 视觉判断
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
4 / 5

建议验证方式

answer.txt titles match the unique lab-fixed drama; citing URLs included. Graded by exact title match (either language) + citations.

建议部分得分

40% constraints combined; 40% exact title; 20% URLs

R04.01.01.01-T10 Find a novel by translator and award chain

原始指令

Find the translated novel whose English translator also translated a famous poet's collected works, that won a translation prize in a given year, and whose original was published in a specified country. Write the novel's title and author to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
4 / 5

建议验证方式

answer.txt title+author equal the unique match; citing URLs present. Graded by exact match + citations.

建议部分得分

40% translator/prize chain; 40% exact title+author; 20% URLs

R04.01.01.01-T11 Identify a building by architect and feature

原始指令

Find the building designed by a named architect, completed in a given decade, that was the first in its city to use a specific structural feature. Write the building name and completion year to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
3 / 5

建议验证方式

answer.txt name+year equal the lab-fixed match; citing URLs included. Graded by exact match + citations.

建议部分得分

50% building candidate; 30% exact name+year; 20% URLs

R04.01.01.01-T12 Trace a footballer transfer via Spanish sources

原始指令

Using Spanish-language sources, identify the footballer who moved between two named clubs in a given transfer window, for a fee in a stated range, having scored a specific number of league goals the prior season. Write the player's full name and the fee to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
long · 长流程
提案难度
5 / 5

建议验证方式

answer.txt player name matches the unique lab-fixed transfer and the fee is within the cited range; citing URLs present. Graded by exact name + fee consistency + citations.

建议部分得分

30% clubs/window chain; 30% prior-season stat; 30% exact player; 10% URLs

R04.01.01.01-T13 Identify a species named after a person

原始指令

Find the animal species whose scientific name honors a named naturalist, was first described in a given decade, and is endemic to a specific island group. Write the binomial name to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
3 / 5

建议验证方式

answer.txt binomial equals the lab-fixed species; citing URLs included. Graded by exact binomial + citations.

建议部分得分

50% species candidate; 30% exact binomial; 20% URLs

R04.01.01.01-T14 Identify a planetary feature

原始指令

Find the named surface feature on a solar-system body that was imaged by a specific mission, lies within a given latitude band, and is named after a figure from a particular mythology. Write the feature name and the body to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
3 / 5

建议验证方式

answer.txt feature+body equal the lab-fixed match; citing URLs present. Graded by exact match + citations.

建议部分得分

50% feature candidate; 30% exact name+body; 20% URLs

R04.01.01.01-T15 Identify a treaty by signatory chain

原始指令

Find the 19th-century treaty signed in a named city, by a diplomat who had earlier negotiated a specified agreement, that ceded a particular territory. Write the treaty's name and year to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
3 / 5

建议验证方式

answer.txt treaty+year equal the lab-fixed match; citing URLs included. Graded by exact match + citations.

建议部分得分

40% signatory/city chain; 40% exact treaty+year; 20% URLs

R04.01.01.01-T16 Derive a number by combining pages

原始指令

Compute the difference in elevation (meters) between the highest peak of a named mountain range and the lowest point of a specified lake, where each value must be read from its own source page. Write the difference and the two underlying values to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
4 / 5

建议验证方式

answer.txt difference is within +/-1 m of the lab value and the two cited source values are consistent with it. Graded by numeric match + citation of both values.

建议部分得分

30% peak value; 30% lake value; 30% correct difference; 10% URLs

R04.01.01.01-T17 Capital and bordering-country fact

原始指令

Name the capital city of the only country that borders both a specified country and a named sea, then state that capital's founding century. Write both to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
2 / 5

建议验证方式

answer.txt capital and founding century equal the lab-fixed answer; citing URLs present. Graded by exact match + citations.

建议部分得分

50% correct country+capital; 30% founding century; 20% URLs

R04.01.01.01-T18 Olympic host and venue fact

原始指令

Identify the city that hosted the Summer Olympics in a given year and name the main stadium used, then state that stadium's seating capacity at the time. Write all three to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
2 / 5

建议验证方式

answer.txt city, stadium, and capacity equal the lab-fixed answer (capacity within tolerance); citing URLs present. Graded by exact match + citations.

建议部分得分

40% city+stadium; 40% capacity; 20% URLs

R04.01.01.01-T19 Author birthplace to landmark

原始指令

A famous author was born in a city; name the tallest building in that city as of 2020 and its height in meters. Write the author's birth city, the building, and its height to answer.txt with source URLs.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单
工作产物
WP12 · 答案 / 决策
条件
未声明
提案难度
2 / 5

建议验证方式

answer.txt city, building, and height match the lab-fixed answer (height within tolerance); citing URLs present. Graded by exact match + citations.

建议部分得分

40% birth city; 40% building+height; 20% URLs

R04.01.01.01-T20 Four-constraint cross-domain chain

原始指令

Find the person who: invented a device patented in a given decade, shared a surname with a named painter, was born in a city that later hosted a specified sporting event, and whose device is now displayed in a particular museum. Write the person's full name and the device to answer.txt with source URLs for each hop.

平台
web · 任意系统(网页)
应用
Chrome
交互表面
brow · 浏览器
GUI 需求 · 任务设定
AP · API 足够
GUI 需求 · 普通权限
G2 · GUI 必要
交互原语
nav · 平移 / 缩放 / 滚动 click · 指向与点击 type · 文字与表单 vis · 视觉判断
工作产物
WP12 · 答案 / 决策
条件
long · 长流程
提案难度
5 / 5

建议验证方式

answer.txt name+device equal the unique lab-fixed answer satisfying all four constraints; a URL is cited for each hop. Graded by exact match + per-hop citations.

建议部分得分

80% four constraints resolved with cited URLs, split evenly per hop; 20% exact name+device

已有研究映射

以下沿用研究目录中的基准与小类映射,不是逐题标注、执行记录或产品背书。

同工作流族的小类