neomatrix369/py-bug-trace-qwen3-6-35b-a3b-l2-rollouts
收藏资源简介:
该数据集是一个用于评估AI模型在复杂任务中表现的数据集,包含多个特征字段,如评估ID、模型名称、任务级别、示例ID、rollout编号、跟踪ID、提示信息(包括角色、内容、工具调用等)、完成内容(包括角色、内容、推理内容、工具调用等)、奖励值、以及详细信息(如任务ID、难度、类别、时间统计、令牌使用情况、完成状态、停止条件、性能指标如精确匹配奖励和回合数等)。此外,还包括精确匹配奖励、延迟时间和总时间。数据集仅包含训练集,有15个示例,总大小约84.9KB,适用于模型性能分析和基准测试。
This dataset is designed for evaluating the performance of AI models on complex tasks, featuring multiple fields such as eval_id, model, level, example_id, rollout_number, trace_id, prompt (including role, content, tool_calls, etc.), completion (including role, content, reasoning_content, tool_calls, etc.), reward, and info (e.g., id, difficulty, category, timing statistics, token usage, completion status, stop condition, metrics like exact_match_reward and num_turns). It also includes exact_match_reward, latency_ms, and total_time. The dataset contains only a train split with 15 examples and a total size of approximately 84.9KB, suitable for model performance analysis and benchmarking.




