datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
casual-conversation-DPOSohamGhadge/casual-conversation with rejected responses generated by Phi-3 q4 and poo
lm-eval-results-nbeerbower-Flammen-Bophades-7B-private
Dataset Card for Evaluation run of nbeerbower/Flammen-Bophades-7B
Dataset automatically created during the evaluation run of model nbeerbower/Flammen-Bophades-7B
The dataset is composed of 62 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest… See the full description on the dataset page: https://huggingface.co/datasets/nyu-dice-lab/lm-eval-results-nbeerbower-Flammen-Bophades-7B-private.Phoenix-SFT-v1
Phoenix v1 — Memory Extraction Dataset
Synthetic SFT training data for Phoenix, a small (Qwen2.5-class) model
that reads a window of chat messages between a user and an AI character
("Flame") and emits a structured JSON list of memorable facts the Flame
should remember about the user going forward.
This is the v1 dataset shipped by flammen.ai for
the memory layer that drives long-term continuity in async character
conversations. Released as part of flammen.ai's open-source data… See the full description on the dataset page: https://huggingface.co/datasets/flammenai/Phoenix-SFT-v1.flame-kindling-v1
flame-kindling-v1
A small, opinionated SFT dataset for finetuning a 3B-class instruct model into a character designer that emits a strict JSON schema from a free-text seed. Built to replace a general RP model (Mistral-Nemo-12B Mahou finetune) being shoehorned into JSON output for flammen.ai's Create-a-Flame pipeline.
400 (seed → DesignedFlame) pairs distilled from Claude Sonnet 4.5 with tool-forcing, validated against a strict pydantic schema, deduplicated by name and… See the full description on the dataset page: https://huggingface.co/datasets/flammenai/flame-kindling-v1.Grill-preprod-v1_chatML
Grill-v1
This dataset contains ChatML formatted conversation between a human and AI. The goal is simulating a preproduction version of flammen.ai in order to train the Mahou series of models.
Schema
chatID - index linking conversations
idx - index of response in a conversation
prompt - ChatML formatted conversation including flammen.ai system prompt
chosen - accepted response
rejected - rejected response (for DPO)
Method
SillyTavern was used as the frontend… See the full description on the dataset page: https://huggingface.co/datasets/flammenai/Grill-preprod-v1_chatML.character-roleplay-DPOThis is practical-dreamer/RPGPT_PublicDomain-alpaca with an added rejection column generated by Phi3-q4.
MahouMix-v1Date-DPO-v3
Date-DPO-v3
DPO dataset aiming to reduce output verbosity and "GPT-speak."
This volume adds roleplay elements and encourages chosen answers to respond as if the assistant were a human with personal preferences.
Method
100 questions were generated by flammen24-mistral-7B.
ChatGPT 3.5's one-shot answers were selected as the rejected responses.
flammen22X-mistral-7B AND flammen24-mistral-7B were used to generate chosen prompts. Many of these responses were manually edited.
Date-DPO-v1
Date-DPO-v1
DPO dataset aiming to reduce output verbosity and "GPT-speak."
Method
Categories and prompts spanning various topics were selected and generated.
ChatGPT 3.5's one-shot answers were selected as the rejected responses.
flammen19X-mistral-7B was used to generate chosen prompts. Many of these responses were prompted with "respond conversationally/succinctly" and modified to be shorter.
Grill-Mahou-v1_chatMLflammenai__Mahou-1.2a-mistral-7B-details
Dataset Card for Evaluation run of flammenai/Mahou-1.2a-mistral-7B
Dataset automatically created during the evaluation run of model flammenai/Mahou-1.2a-mistral-7B
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 3 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/flammenai__Mahou-1.2a-mistral-7B-details.FlameMix-DPO-v1A mix of DPO sets attempting to produce an uncensored LLM capable of fun, human-like conversation.
Contains:
flammenai/Date-DPO-v1, flammenai/Date-DPO-v2, and flammenai/Date-DPO-v3 - Q&A Flammen-Mistral vs one-shot ChatGPT 3.5
flammenai/Prude-Phi3-DPO - ResplendentAI/NSFW_RP_Format_NoQuote with rejected responses generated by microsoft/Phi-3-mini-4k-instruct-gguf
flammenai/casual-conversation-DPO - SohamGhadge/casual-conversation with rejected responses generated by… See the full description on the dataset page: https://huggingface.co/datasets/flammenai/FlameMix-DPO-v1.Mahou-ST-ChatML-Instructflammenai__Llama3.1-Flammades-70B-details
Dataset Card for Evaluation run of flammenai/Llama3.1-Flammades-70B
Dataset automatically created during the evaluation run of model flammenai/Llama3.1-Flammades-70B
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/flammenai__Llama3.1-Flammades-70B-details.Date-DPO-v2
Date-DPO-v2
DPO dataset aiming to reduce output verbosity and "GPT-speak."
Method
Categories and prompts spanning various topics were selected and generated.
ChatGPT 3.5's one-shot answers were selected as the rejected responses.
flammen20-mistral-7B was used to generate chosen prompts. Many of these responses were prompted with "respond conversationally, do not list anything" and modified to be shorter.
Grill-Flammen-v1_chatMLflammenai__flammen15-gutenberg-DPO-v1-7B-details
Dataset Card for Evaluation run of flammenai/flammen15-gutenberg-DPO-v1-7B
Dataset automatically created during the evaluation run of model flammenai/flammen15-gutenberg-DPO-v1-7B
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/flammenai__flammen15-gutenberg-DPO-v1-7B-details.flammenai__Mahou-1.2a-llama3-8B-details
Dataset Card for Evaluation run of flammenai/Mahou-1.2a-llama3-8B
Dataset automatically created during the evaluation run of model flammenai/Mahou-1.2a-llama3-8B
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest results.… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/flammenai__Mahou-1.2a-llama3-8B-details.flammenai__Mahou-1.5-mistral-nemo-12B-details
Dataset Card for Evaluation run of flammenai/Mahou-1.5-mistral-nemo-12B
Dataset automatically created during the evaluation run of model flammenai/Mahou-1.5-mistral-nemo-12B
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/flammenai__Mahou-1.5-mistral-nemo-12B-details.flammenai__Mahou-1.5-llama3.1-70B-details
Dataset Card for Evaluation run of flammenai/Mahou-1.5-llama3.1-70B
Dataset automatically created during the evaluation run of model flammenai/Mahou-1.5-llama3.1-70B
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/flammenai__Mahou-1.5-llama3.1-70B-details.Mahou-ST-Sampler-PresetGrill-preprod-v2_chatML
Grill-v2
This dataset contains ChatML formatted conversation between a human and AI. The goal is simulating a preproduction version of flammen.ai in order to train the Mahou series of models.
Schema
chatID - index linking conversations
idx - index of response in a conversation
prompt - ChatML formatted conversation including flammen.ai system prompt
chosen - accepted response
rejected - rejected response (for DPO)
Method
SillyTavern was used as the frontend… See the full description on the dataset page: https://huggingface.co/datasets/flammenai/Grill-preprod-v2_chatML.Prude-Phi3-DPOResplendentAI/NSFW_RP_Format_NoQuote inputs ran through Phi-3 Q4 to intentionally produce bad responses to be used for rejections.
Date-DPO-NoAsterisks