Hugging Face dataset collects Kimi K3 coding and debugging traces
greghavens · hf · 2026-07-23
greghavens/kimi-k3-coding-and-debugging-traces is trending on Hugging Face as a machine-generated dataset for text generation.
The dataset is stored as Parquet, has a small-to-mid size (1K<n<10K), and is licensed under CC BY 4.0. Its purpose is to provide coding and debugging traces that can support text generation and agent training workflows.
More from Research
- FrontierCS 2 launches on Kaggle: 48 tasks make agents beat published CS papers — MeganRisdal · 2026-10-27
- LightMTP: lightweight latent multi-token prediction boosts LLM planning in pretraining — gdm3000 · 2026-10-27
- DexPort: a handheld interface with a real robot hand for dexterous-hand data collection — pabbeel · 2026-10-27
- EMNLP study: AI literary translation reads fine but varies more, even within one book — tuvllms · 2026-10-27
- Lab monkeys at $30-80K each: a 4-year breeding cycle is the real preclinical drug bottleneck — zebird0 · 2026-10-27
- Custom VQA benchmark under 48GB VRAM: Qwen3.8-Flash-Next leads, Qwen3.8-27B regressed — EmilPi · 2026-10-27