datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
odexODEX is an Open-Domain EXecution-based NL-to-Code generation data benchmark.
It contains 945 samples with a total of 1,707 human-written test cases,
covering intents in four different natural languages -- 439 in English, 90 in Spanish, 164 in Japanese, and 252 in Russian.odexODEX dataset annotated with the ground-truth library documentation, to enable evaluations for retrieval and retrieval-augmented code generation.
Please refer to [code-rag-bench] for more details.
mistakes-analysis
Nanbeige-4-3B-Base: Semantic Blind Spots & Mistake Analysis
Overview
This dataset is a curated collection of 10+ high-fidelity semantic failures identified during the evaluation of the Nanbeige/Nanbeige4-3B-Base model.
While the Nanbeige model is highly capable for its size, my testing revealed specific "blind spots" in logical reasoning, strict constraint adherence, and mathematical precision. This dataset serves as a benchmark for where the model currently fails… See the full description on the dataset page: https://huggingface.co/datasets/Odelolasolomon/mistakes-analysis.TRObject-Dataset-Test
TRObject Code Generation Instruction Dataset
This dataset contains natural language instructions paired with TRObject code outputs.
It was created for fine-tuning and evaluating domain-specific LLMs that generate TRObject code for Clomosy-style mobile application development.
Dataset Description
TRObject is used in the Clomosy mobile application development platform. Since general-purpose LLMs do not reliably understand TRObject syntax or Clomosy-specific UI patterns… See the full description on the dataset page: https://huggingface.co/datasets/odenmehmet/TRObject-Dataset-Test.Oden-worldchess
Oden Chess Dataset
Dataset Summary
The Oden Chess Dataset is a comprehensive collection of over 4 million chess games compiled from top players, major tournaments, and categorized by opening systems. This dataset provides rich annotations including move sequences, board positions, player information, and game metadata, making it ideal for chess AI research, opening analysis, and statistical studies.
Dataset Details
Total Games: 4,047,908
Source Files:… See the full description on the dataset page: https://huggingface.co/datasets/BBSRguy/Oden-worldchess.