marin-community/openthoughts4-code-9168-prompts-qwen3-30b-a3b-thinking-2507-n16-flattened-logprobs-k16
收藏资源简介:
OpenThoughts-4 Code SDG: Qwen3-30B-A3B-Thinking-2507 (n=16, top-16 logprobs)数据集包含来自Qwen/Qwen3-30B-A3B-Thinking-2507模型在Marin OpenThoughts-4代码SDG提示集上的合成生成内容。每个提示被采样16次,并且对于每个生成的令牌,数据集存储了所选令牌的对数概率以及词汇表中前16个对数概率。该数据集适用于蒸馏、KL风格微调、重新排名和不确定性分析等任务。数据集详细说明了生成设置、模式、文件布局、配套数据集、许可证和引用信息。
The dataset OpenThoughts-4 Code SDG: Qwen3-30B-A3B-Thinking-2507 (n=16, top-16 logprobs) contains synthetic generations from the Qwen/Qwen3-30B-A3B-Thinking-2507 model on the Marin OpenThoughts-4 code SDG prompt set. Each prompt is sampled 16 times, and for every generated token, the dataset stores the chosen-token log probability plus the top-16 log probabilities over the vocabulary. The dataset is intended for tasks like distillation, KL-style fine-tuning, reranking, and uncertainty analysis. The README includes details about the generation setup, schema, file layout, companion datasets, license, and citation information.




