Skip to content

Add task: football-match-scorers#49

Open
lichengliu03 wants to merge 2 commits into
cocoabench:mainfrom
lichengliu03:task/football-match-scorers-clean
Open

Add task: football-match-scorers#49
lichengliu03 wants to merge 2 commits into
cocoabench:mainfrom
lichengliu03:task/football-match-scorers-clean

Conversation

@lichengliu03
Copy link
Copy Markdown
Contributor

This task requires visual understanding, web browsing, and data extraction. It presents an image from a football match, and the agent must identify all players who scored goals and list them in chronological order.

The task tests whether agents can:

  • Access and analyze images from Google Drive links
  • Extract player names from match information
  • Identify chronological sequence of goals
  • Format output with full names
  • Maintain proper ordering

This task tests the agent's ability to analyze a football match image
and identify all goal scorers in chronological order. The task requires
visual understanding and attention to detail in extracting player names
and goal sequence from match information.

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
@cocoabench
Copy link
Copy Markdown
Owner

Hi @lichengliu03 Thanks for your contribution! We reviewed the task and spotted some issues:

  1. In evaluation.md, the example agent output is not provided.
  2. There are errors in one of the file: In instruction, "The order should be the sequence of goals scored, from earliest to latest.", whereas in solution, "List the players' full names in chronological order, separated by commas."
  3. The solution is pretty vague. It means that no concrete steps are described for a human to verify the problem-solving process. Could you include concrete details that lead to the final answer? For example, you should include the reasoning process of why this player is A (according to web search or other details?); you should also include a list of goals scored of each player and how a human could get that data.

We look forward to your response.

- Add detailed human solution with specific match identification
  (Arsenal 5-1 Man City, Feb 2 2025), goal minutes, and data sources
- Simplify evaluation.md format to match project conventions
- Add agent test result (ChatGPT 5.2 Thinking, Incorrect)
- Fix wording inconsistency between instruction.md and solution.md

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@lichengliu03 lichengliu03 force-pushed the task/football-match-scorers-clean branch from 41a8ae5 to 20b49e4 Compare March 6, 2026 05:42
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants