name stringlengths 3 18 | split stringclasses 1
value | game stringlengths 3 18 | source stringclasses 1
value | prompt stringlengths 4.65k 16.6k | system_prompt stringclasses 1
value | episodes int64 1k 1k | task_id int64 0 321 | family stringclasses 2
values | upstream_source_url stringlengths 120 135 | upstream_revision stringclasses 1
value | native_source_url stringlengths 111 138 ⌀ |
|---|---|---|---|---|---|---|---|---|---|---|---|
tongue_hunter | train | tongue_hunter | synth | Write from scratch the weights of a policy network that plays Tongue Hunter as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `ed... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 305 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/tongue_hunter.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/joyrider3774/crisp-game-lib-portable-sdl/blob/04bc2c4cfc2853628eda25a46190bb7ac6fe216a/src/games/gameTongueHunter.c |
forfour | train | forfour | synth | Write from scratch the weights of a policy network that plays Forfour as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edits` c... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 306 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/forfour.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/abagames/crisp-game-lib-games/blob/cfb39d2f988feb5918eb83697145b5a35adabf58/docs/forfour/main.js |
sky_raftsman | train | sky_raftsman | synth | Write from scratch the weights of a policy network that plays Sky Raftsman as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edi... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 307 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/sky_raftsman.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/joyrider3774/crisp-game-lib-portable-sdl/blob/04bc2c4cfc2853628eda25a46190bb7ac6fe216a/src/games/gameSkyRaftsman.c |
river_raid | train | river_raid | synth | Write from scratch the weights of a policy network that plays River Raid as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edits... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 308 | atari_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/river_raid.py | 4038830175c4201ca10e004ed4441985f2acddf2 | null |
t_laser | train | t_laser | synth | Write from scratch the weights of a policy network that plays T Laser as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edits` c... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 309 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/t_laser.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/abagames/crisp-game-lib-games/blob/cfb39d2f988feb5918eb83697145b5a35adabf58/docs/tlaser/main.js |
cue_crumb | train | cue_crumb | synth | Write from scratch the weights of a policy network that plays Cue Crumb as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edits`... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 310 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/cue_crumb.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/joyrider3774/crisp-game-lib-portable-sdl/blob/04bc2c4cfc2853628eda25a46190bb7ac6fe216a/src/games/gameCueCrumb.c |
marble_bounce | train | marble_bounce | synth | Write from scratch the weights of a policy network that plays Marble Bounce as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `ed... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 311 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/marble_bounce.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/joyrider3774/crisp-game-lib-portable-sdl/blob/04bc2c4cfc2853628eda25a46190bb7ac6fe216a/src/games/gameMarbleBounce.c |
sky_golf | train | sky_golf | synth | Write from scratch the weights of a policy network that plays Sky Golf as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edits` ... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 312 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/sky_golf.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/joyrider3774/crisp-game-lib-portable-sdl/blob/04bc2c4cfc2853628eda25a46190bb7ac6fe216a/src/games/gameSkyGolf.c |
elastic_hero | train | elastic_hero | synth | Write from scratch the weights of a policy network that plays Elastic Hero as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edi... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 313 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/elastic_hero.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/joyrider3774/crisp-game-lib-portable-sdl/blob/04bc2c4cfc2853628eda25a46190bb7ac6fe216a/src/games/gameElasticHero.c |
tarutobi | train | tarutobi | synth | Write from scratch the weights of a policy network that plays Tarutobi as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edits` ... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 314 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/tarutobi.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/abagames/crisp-game-lib-games/blob/cfb39d2f988feb5918eb83697145b5a35adabf58/docs/tarutobi/main.js |
qbert | train | qbert | synth | Write from scratch the weights of a policy network that plays Qbert as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edits` cod... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 315 | atari_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/qbert.py | 4038830175c4201ca10e004ed4441985f2acddf2 | null |
wiper | train | wiper | synth | Write from scratch the weights of a policy network that plays Wiper as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edits` cod... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 316 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/wiper.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/abagames/crisp-game-lib-games/blob/cfb39d2f988feb5918eb83697145b5a35adabf58/docs/wiper/main.js |
pump_press | train | pump_press | synth | Write from scratch the weights of a policy network that plays Pump Press as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edits... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 317 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/pump_press.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/abagames/crisp-game-lib-games/blob/cfb39d2f988feb5918eb83697145b5a35adabf58/docs/pumppress/main.js |
barge_hop | train | barge_hop | synth | Write from scratch the weights of a policy network that plays Barge Hop as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edits`... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 318 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/barge_hop.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/joyrider3774/crisp-game-lib-portable-sdl/blob/04bc2c4cfc2853628eda25a46190bb7ac6fe216a/src/games/gameBargeHop.c |
i_slash | train | i_slash | synth | Write from scratch the weights of a policy network that plays I Slash as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edits` c... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 319 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/i_slash.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/abagames/crisp-game-lib-games/blob/cfb39d2f988feb5918eb83697145b5a35adabf58/docs/islash/main.js |
smoke_g | train | smoke_g | synth | Write from scratch the weights of a policy network that plays Smoke G as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edits` c... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 320 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/smoke_g.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/abagames/crisp-game-lib-games/blob/cfb39d2f988feb5918eb83697145b5a35adabf58/docs/smokeg/main.js |
d_laser | train | d_laser | synth | Write from scratch the weights of a policy network that plays D Laser as well as possible.
There is no training, data, loss, gradients, fine-tuning, code execution, calculator, or environment feedback. This is one response. Briefly describe the strategy and hidden units, then provide the weights in the final `edits` c... | You directly program neural-network weights. Follow the task constraints exactly. | 1,000 | 321 | crisp_synthetic | https://github.com/Trajectorylabs/minatar-synth/blob/4038830175c4201ca10e004ed4441985f2acddf2/minatar_synth/games/d_laser.py | 4038830175c4201ca10e004ed4441985f2acddf2 | https://github.com/abagames/crisp-game-lib-games/blob/cfb39d2f988feb5918eb83697145b5a35adabf58/docs/dlaser/main.js |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.