datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
pokemon-llm-svg-bench-results
Pokémon LLM SVG Bench (V1.1)
This is a snapshot of scored SVG drawings from the Pokémon LLM SVG Bench — LLMs try to draw Pokémon in SVG, and we score the results. Prefer reading the live site for context; for scoring rules and methodology, see About.
Unofficial fan / research project. Pokémon images are copyrighted by The Pokémon Company and related companies involved in developing, operating, and managing the Pokémon series. Pokémon descriptions are sourced from… See the full description on the dataset page: https://huggingface.co/datasets/haxfenx/pokemon-llm-svg-bench-results.pokemon-red-sft
pokemon-kafka SFT corpus (v6u, 2026-09-07 late)
Deterministic conversion of pokemon-kafka telemetry (seed 42). 28003 examples: 25203 train, 2800 valid.
Seats
seat
domain
rows
what the row teaches
Wheelman
battle-outcome
3599
will this fight be won; fight or flee
Wheelman
move-choice
10993
damage bucket per move; best move per matchup
Wheelman
battle-action
719
next action from a won battle's turns
Extractor
puzzle-consult
605
menu choice at a wall… See the full description on the dataset page: https://huggingface.co/datasets/bdougie/pokemon-red-sft.adaption-pokemon-story-prompts
This dataset is a remastered version of this dataset prepared using Adaption's Adaptive Data platform.
adaption-pokemon_story_prompts
This dataset contains prompts instructing a model to write stories about specific Pokémon based on their detailed attributes, including stats, types, abilities, and lore. Each entry provides structured data such as height, weight, generation, and flavor text alongside an image URL. The primary focus is on generating creative narratives grounded… See the full description on the dataset page: https://huggingface.co/datasets/sarahooker/adaption-pokemon-story-prompts.pokemon-cards-image-and-annotationspokemon-showdown-grpo-tutorial
Pokémon Showdown GRPO tutorial dataset
Pre-built GRPO records for the ROCm AI Developer Hub tutorial.
Split
File
Records
demo
data/demo.jsonl
64
train
data/train.jsonl
2048
validate
data/validate.jsonl
32
Use via tutorial notebook Step 12 (load_grpo_tutorial_records) or regenerate with prepare_grpo_tutorial_data.py.
Companion scripts: https://github.com/GoldenGrapeGentleman/pokemon-showdown-agent-scripts
pokemon-red-telemetry-sft
Pokemon Red Telemetry SFT corpus (sft_v3)
590 chat-format SFT examples generated deterministically (seed 42) from ~38k pokemon.game.v1
telemetry events emitted by an autonomous Pokemon Red agent
(pokemon-kafka) and its training loop
(empirical-evidence). Labels for the battle
domains come from game RAM, not annotation.
This corpus was used to fine-tune a multi-task SmolLM3-3B LoRA in the empirical-evidence loop,
which beat the base model on held-out game-RAM labels… See the full description on the dataset page: https://huggingface.co/datasets/bdougie/pokemon-red-telemetry-sft.pokemon-training-datapokemon-lore-instructionspokemon_tcg_data_setsmistral_pokemonpokemon_transparentpokemon-alpaca_zh_130kpokemon-question-answer-tagalogSynthetic-Pokemondataset_pokemon_en_fr_gemma3Pokemon-yellowdataset_pokemon_en_frpokemonpokemon-R1-alpaca_zh_2kPOKEMONoverfit_small-aicrowd-pokemon-training-data
Dataset Description
The dataset was constructed through manual gameplay.
At each step, given the system and user prompts describing the current game state, a corresponding use_tool action was explicitly authored by a human player while directly interacting with the environment. This process ensured that every action strictly followed the defined interface and playbook constraints.
In some cases, data augmentation was applied by modifying object names to account for variations… See the full description on the dataset page: https://huggingface.co/datasets/small-lit/overfit_small-aicrowd-pokemon-training-data.overfit_small-aicrowd-pokemon_red-training-data-14
Dataset Description
The dataset was constructed through manual gameplay.
At each step, given the system and user prompts describing the current game state, a corresponding use_tool action was explicitly authored by a human player while directly interacting with the environment. This process ensured that every action strictly followed the defined interface and playbook constraints.
All data was generated firsthand through direct gameplay and prompt–action annotation, without relying… See the full description on the dataset page: https://huggingface.co/datasets/small-lit/overfit_small-aicrowd-pokemon_red-training-data-14.pokemonpokemon_llmpokemonpokemon2pokemon3Dive_Pokemon