2D Grid Traversal Eval

[eugene-kim-pipe17]

Thank you for contributing an eval! ♥️

🚨 Please make sure your PR follows these guidelines, failure to follow the guidelines below will result in the PR being closed automatically. Note that even if the criteria are met, that does not guarantee the PR will be merged nor GPT-4 access granted. 🚨

PLEASE READ THIS:

In order for a PR to be merged, it must fail on GPT-4. We are aware that right now, users do not have access, so you will not be able to tell if the eval fails or not. Please run your eval with GPT-3.5-Turbo, but keep in mind as we run the eval, if GPT-4 gets higher than 90% on the eval, we will likely reject since GPT-4 is already capable of completing the task.

We plan to roll out a way for users submitting evals to see the eval performance on GPT-4 soon. Stay tuned! Until then, you will not be able to see the eval performance on GPT-4. Starting April 10, the minimum eval count is 15 samples, we hope this makes it easier to create and contribute evals.

Eval details 📑

Eval name

2D Grid Traversal

Eval description

Tests whether or not Chat GPT is able to keep track of an X,Y coordinate given an input scenario, rules, and instructions.

What makes this a useful eval?

This eval is an intersection of basic addition / subtraction, matrices, logic, and some elements that are common in games. I imagine that many people are interested in this kind of functionality.

Criteria for a good eval ✅

Below are some of the criteria we look for in a good eval. In general, we are seeking cases where the model does not do a good job despite being capable of generating a good response (note that there are some things large language models cannot do, so those would not make good evals).

Your eval should be:

• Thematically consistent: The eval should be thematically consistent. We'd like to see a number of prompts all demonstrating some particular failure mode. For example, we can create an eval on cases where the model fails to reason about the physical world.
• Contains failures where a human can do the task, but either GPT-4 or GPT-3.5-Turbo could not.
• Includes good signal around what is the right behavior. This means either a correct answer for Basic evals or the Fact Model-graded eval, or an exhaustive rubric for evaluating answers for the Criteria Model-graded eval.
• Include at least 15 high quality examples.

If there is anything else that makes your eval worth including, please document it below.

Unique eval value

I think what makes this eval unique is that it's something a young child can do if given a grid.

Eval structure 🏗️

Your eval should

• Check that your data is in evals/registry/data/{name}
• Check that your yaml is registered at evals/registry/evals/{name}.yaml
• Ensure you have the right to use the data you submit via this eval

(For now, we will only be approving evals that use one of the existing eval classes. You may still write custom eval classes for your own cases, and we may consider merging them in the future.)

Final checklist 👀

Submission agreement

By contributing to Evals, you are agreeing to make your evaluation logic and data under the same MIT license as this repository. You must have adequate rights to upload any data used in an Eval. OpenAI reserves the right to use this data in future service improvements to our product. Contributions to OpenAI Evals will be subject to our usual Usage Policies (https://platform.openai.com/docs/usage-policies).

• I agree that my submission will be made available under an MIT license and complies with OpenAI's usage policies.

Email address validation

If your submission is accepted, we will be granting GPT-4 access to a limited number of contributors. Access will be given to the email address associated with the merged pull request.

• I acknowledge that GPT-4 access will only be granted, if applicable, to the email address used for my merged pull request.

Limited availability acknowledgement

We know that you might be excited to contribute to OpenAI's mission, help improve our models, and gain access to GPT-4. However, due to the requirements mentioned above and high volume of submissions, we will not be able to accept all submissions and thus not grant everyone who opens a PR GPT-4 access. We know this is disappointing, but we hope to set the right expectation before you open this PR.

• I understand that opening a PR, even if it meets the requirements above, does not guarantee the PR will be merged nor GPT-4 access granted.

Submit eval

• I have filled out all required fields in the evals PR form
• (Ignore if not submitting code) I have run pip install pre-commit; pre-commit install and have verified that black, isort, and autoflake are running when I commit and push

Failure to fill out all required fields will result in the PR being closed.

Eval JSON data

Since we are using Git LFS, we are asking eval submitters to add in as many Eval Samples (at least 5) from their contribution here:

View evals in JSON

Eval

{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 3\nH: 3\nSP: 2,2\nD: right,up,up\nOOBB: wall\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"2,2"}
{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 3\nH: 3\nSP: 2,2\nD: right,up,up\nOOBB: pacman\t\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"0,1"}
{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 3\nH: 3\nSP: 1,1\nD: left,left,left,up,up,up\nOOBB: pacman\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"1,1"}
{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 3\nH: 3\nSP: 0,0\nD: left,left,left,left,up,up,up,up,down,right\nOOBB: wall\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"1,1"}
{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 3\nH: 3\nSP: 0,0\nD: left,left,left,left,up,up,up,up,down,right\nOOBB: pacman\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"0,0"}
{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 100\nH: 1\nSP: 0,0\nD: left,left,left,left,left,left,left,left\nOOBB: wall\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"0,0"}
{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 100\nH: 1\nSP: 0,0\nD: left,left,left,left,left,left,left,left,left\nOOBB: pacman\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"91,0"}
{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 100\nH: 1\nSP: 0,0\nD: left,left,left,left,left,right\nOOBB: wall\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"1,0"}
{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 10\nH: 10\nSP: 9,9\nD: right,up,down,down,left,left,down,down,down,left,left,left\nOOBB: wall\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"4,4"}
{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 10\nH: 10\nSP: 9,9\nD: right,up,down,down,left,left,down,down,down,left,left,left\nOOBB: pacman\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"5,5"}
{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 10\nH: 2\nSP: 8,1\nD: up,up,up,down,left,left,left,left,left,left,left,left,left,left,right\nOOBB: wall\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"1,0"}
{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 4\nH: 4\nSP: 2,2\nD: up,up,right,right,up,right,up,right,down,down,down,left,left,left\nOOBB: wall\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"0,0"}
{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 4\nH: 4\nSP: 2,2\nD: up,up,right,right,up,right,up,right,down,down,down,left,left,left\nOOBB: pacman\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"3,3"}
{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 5\nH: 3\nSP: 0,0\nD: down,up,left,right,down,down,up,up,left,left,right,right\nOOBB: pacman\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"0,0"}
{"input":[{"role":"system","content":"You are GridGPT, a program that can can track where a coordinate will be on a 2D plane after being given instructions.\n\nFor each prompt, you will be given the following input:\n* W: The number of tiles the grid spans across the X-axis. W >= 1\n* H: The number of tiles the grid spans across the Y-axis. H >= 1\n* SP: A zero-indexed X,Y coordinate that describes your starting position. It will be within the bounds of the grid. If your starting position was the at the left and bottom most tile, it would be 0,0\n* D: a comma separated list of directions you will take from the starting position. Example: left,right,right,up,down. Unless you go out of bounds (details below), left will decrement your X coordinate by 1, right will increment your X coordinate by 1, up will increment your Y coordinate by 1, and down will decrement your Y coordinate by 1.\n* OOBB: Out of bounds behavior. This parameter defines how the program should handle situations when you are asked to go out of bounds (e.g., current position is 0,0 and you go left or down). OOBB can be \"wall\" or \"pacman.\" If OOBB is set to \"wall,\" your current position does not change when you attempt to move out of bounds. Essentially, you will be stopped by an imaginary wall at the grid's boundary, preventing any movement in that direction. If OOBB is set to \"pacman,\" your position will wrap around the grid when you attempt to move out of bounds. In this mode, if you reach the top or right sides of the grid and continue moving in the same direction, your position along the direction's axis will be set to 0. Conversely, if you reach the left or bottom sides of the grid and continue moving, your position along the direction's axis will be set to the maximum value (W-1 for the X-axis and H-1 for the Y-axis). For example, if the grid is 10x10 and your current position is (9, 9), moving right with OOBB set to \"pacman\" will wrap you around to position (0, 9), as if you emerged from the opposite side of the grid. Similarly, moving down from position (0, 0) would wrap you around to position (0, 9). This behavior is reminiscent of the classic Pac-Man game, where the game character can move through one side of the screen and reappear on the opposite side."},{"role":"user","content":"Input\nW: 5\nH: 5\nSP: 4,4\nD: right,up,right,right,right,right,right\nOOBB: pacman\nWhat is your final position? Respond only with the X and Y coordinates as X,Y"}],"ideal":"0,0"}

[eugene-kim-pipe17]

Hello @andrew-openai , is there anything I should do to improve this PR? I'd love some feedback on this

[jorge-openai]

Thanks for opening this PR. I would like to suggest a couple of changes.

• Please upload samples.jsonl to Git LFS, all sample files for evals are hosted there.

• Add a description field in the .yaml file.

• Although Match can be used in this case, a better choice would be to use `Includes type. This will allow the model additional room for reasoning and still can be guided to answer in such a way that the response can be checked against the ideal. For example, " In your response, your final answer should be enclosed in brackets like so {answer here}".

We'd love to review the PR once the changes have been made.

[jorge-openai]

Closing for now, please feel free to reopen if you get a chance to address the comments.