Finance tasks for LLM evaluation
Evaluation sets for LLM domain adaptation in finance from the ICLR 2024 paper.
- Text classification
- Question answering
- Zero shot classification
- English
Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,788 in the catalog so far, with new ones added every hour.
Answered by Gemini 4 · the chat is kept only in this tab
Evaluation sets for LLM domain adaptation in finance from the ICLR 2024 paper.
NVIDIA code corpus metadata for pre-training Nemotron 3 family LLMs.
Cleaned logs of multi-step operation by the OpenAI Codex Desktop agent in a local Windows environment.
Public domain Japanese fiction and nonfiction literature corpus.
35 types of logic tasks with automated answer checking for RL with verifiable rewards.
Cleaned subset of biomedical texts and English questions for evaluating search and generation systems.
Question-answer pairs from real counseling sessions with psychologists for training empathetic dialogues.
A small set of multimodal feedback for aligning vision-language models.
Tabular and text corpus for training and evaluating language model safety using RLHF.
An English set of 177,597 financial instructions and Q&A pairs for LLM fine-tuning.
Curated multi-turn GPT-4 dialogues in Chinese, English, and other languages.
Hundreds of thousands of article-summary pairs in multiple languages for training and evaluating summarization systems.
Over 5 million sections of US laws, regulations, and prosecutor opinions for search and analysis.
A synthetic set of Python programming tasks in HumanEval format from ChatGPT 3.5 for distilling coding skills.
A massive 6.2 billion token text corpus for pre-training and SFT of models in Portuguese.
200k Chinese instruction-response pairs for aligning and training language models.
An open English corpus for training agents capable of information retrieval and question answering.
A large multi-task streaming audio dataset for training audio LLMs and voice agents.
77 real-world software engineering tasks for deep learning frameworks in English and Chinese.
English instruction data mixture for SFT of Tulu language assistants.
Hundreds of millions of cleaned Chinese texts for LLM pretraining and safety.
1.06M trajectories (state, action, next_state) from real websites for training browser world models.
Raw JSONL agent traces from Qwen3.7 Max with a ready-to-use tool schema.
Over 10k multi-turn examples for instruction fine-tuning in English.