{"oral": "如果咱们把这个 Loss Function 换成 Cross Entropy，啊不对，应该是 Focal Loss 的话，那我们可能在处理这种 class imbalance 样本不平衡问题时，能把那百分之十五，哦不对，百分之二十二点五 的误报率给压下去，这样在那个新，新苑，新苑园区测试得数据表现应该会更稳一些。", "clean": "如果咱们把 Loss Function 换成 Focal Loss 的话，那我们可能在处理这种 class imbalance 样本不平衡问题时，能把 22.5% 的误报率给压下去，这样在新苑园区测试的数据表现应该会更稳一些。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl10.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 1, "source_record_id": "recheck.jsonl:10"}
{"oral": "So I was thinking about emailing a publisher about a manuscript tomorrow morning... oh no, tomorrow evening, and I'm emailing you regarding some hashtag ideas for it, so I hope you can help me with that part, do you think that sounds okay?", "clean": "I was thinking about emailing a publisher about a manuscript tomorrow evening, and I'm emailing you regarding some hashtag ideas for it, so I hope you can help me with that part. Do you think that sounds okay?", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl301.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 172, "source_record_id": "recheck.jsonl:301"}
{"oral": "帮我把明早 8:30 的闹钟改成 7:15，顺便提醒我带上充电宝。", "clean": "帮我把明早 8:30 的闹钟改成 7:15，顺便提醒我带上充电宝。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl601.wav", "scene": "passthrough", "language": "zh", "voice_id": "ICL_uranus_zh_female_zhixinjiejie_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 330, "source_record_id": "recheck.jsonl:601"}
{"oral": "诶，如果咱们能把 Prisma 迁移到 Elixir 里面，嗯，也就是把原先 Astro 里面那两千三百个数据表全部导出来，我想想啊，如果能绕过 ConnectionRefused 这个报错的话，那我们就能直接在一次 epochs 运行完吧。", "clean": "如果咱们能把 Prisma 迁移到 Elixir 里面，也就是把原先 Astro 里面那2300个数据表全部导出来，如果能绕过 ConnectionRefused 这个报错的话，那我们就能直接在一次 epoch 运行完吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl701.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 406, "source_record_id": "recheck.jsonl:701"}
{"oral": "I need to dispute a charge. I need to dispute a charge for two thousand one hundred dollars. Look, the item I received is damaged, damaged and the box was open. The company is Comcast. Transaction number is B L L nine nine two eight four seven one. Order number... order number is hash B nine nine two zero one one. I cannot pay this.", "clean": "I need to dispute a charge for $2,100. The item I received is damaged and the box was open. The company is Comcast. Transaction number is BLL9928471. Order number is #B992011. I cannot pay this.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl101.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 54, "source_record_id": "recheck.jsonl:101"}
{"oral": "我感觉现在与其在建国门那边死磕，不如换个环境。最近跟韩嘉悦去逛万象城，发现那边比建国门更热闹。对了，我打算买那个极米投影仪，价格是两千八百块，不对，两千九百块，等等错了，是三千一百块！然后我那个…要是能把…怎么说呢…如果能把周子墨那份方案改完，我就能休息下。感觉现在的压力比之前大多了。", "clean": "我感觉现在与其在建国门那边死磕，不如换个环境。最近跟韩嘉悦去逛万象城，发现那边比建国门更热闹。对了，我打算买极米投影仪，价格是3100元。如果能把周子墨那份方案改完，我就能休息下。感觉现在的压力比之前大多了。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl201.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 119, "source_record_id": "recheck.jsonl:201"}
{"oral": "下午导航去皮都...不对，郫都，郫是郫县豆瓣的郫，都是都市的都，顺便帮我在药店买一盒霍香...不是，藿香正气液，藿是草字头下面一个霍元甲的霍。", "clean": "下午导航去郫都，顺便帮我在药店买一盒藿香正气液。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2001.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 911, "source_record_id": "explanation.jsonl:2001"}
{"oral": "So, uh, I just synced with Saanvi Patel regarding the dashboard, and, well, the MAU growth is looking... it's at twelve percent... no, thirteen percent... wait, that's wrong, it's actually fourteen percent! Anyway, I think we need to, I mean, we should probably check if Business Intelligence can... just, how should I put it, refine the data for Average Order Value. It was forty two dollars... no, forty five dollars... hold on, sorry, it's forty eight dollars! So, yeah, we're just waiting on DevOps to, uh, deploy the fix by Friday, or maybe Thursday, just to be safe, right?", "clean": "I just synced with Saanvi Patel regarding the dashboard, and the MAU growth is at 14%! I think we should probably check if Business Intelligence can refine the data for Average Order Value. It was $48! So, we're just waiting on DevOps to deploy the fix by Thursday, just to be safe, right?", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl407.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 230, "source_record_id": "recheck.jsonl:407"}
{"oral": "Hey, I want to benchmark two caching strategies. If we integrate Vault for secret management... then we could maybe... wait, let me rephrase... then we could compare the latency. I need to run aws eks update kubeconfig first, then run run istioctl analyze to check the Istio Service Mesh config. Once it looks good, I will trigger argocd app sync, trigger argocd app sync to deploy. Can you help me monitor the performance across ten thousand requests per second?", "clean": "Hey, I want to benchmark two caching strategies. If we integrate Vault for secret management, then we could compare the latency. I need to run aws eks update-kubeconfig first, then run istioctl analyze to check the Istio Service Mesh config. Once it looks good, I will trigger argocd app sync to deploy. Can you help me monitor the performance across 10,000 requests per second?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl18998.wav", "scene": "tech", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 810, "source_record_id": "assigned_records.jsonl:18998"}
{"oral": "这个Learning Rate设成零点零一，啊不对，零点零零一才行。", "clean": "Learning Rate设成0.001才行。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl18.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 8, "source_record_id": "recheck.jsonl:18"}
{"oral": "我怀疑怀疑这个 Learning Rate 设太高，或者说…重新看下，应该是 Gradient 梯度爆炸，梯度爆炸导致 Loss 没法收敛，得把 Batch Size 调成五百一十二试试。", "clean": "我怀疑这个 Learning Rate 设太高，应该是 Gradient 梯度爆炸，导致 Loss 没法收敛，得把 Batch Size 调成 512 试试。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl17.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 7, "source_record_id": "recheck.jsonl:17"}
{"oral": "张老师，我觉得这次实验把 Batch Size 设成五百一十二，哦不是，应该是一千零二十四，然后那个，怎么说呢，如果我们尝试在 Fine tuning 阶段引入 LORA，可能能把收敛速度提升百分之十五左右。", "clean": "张老师，我觉得这次实验把 Batch Size 设成1024，然后如果我们尝试在 Fine-tuning 阶段引入 LoRA，可能能把收敛速度提升15%左右。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl100.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 2, "source_record_id": "recheck.jsonl:100"}
{"oral": "其实这个Pipeline大概逻辑是，先是对原始数据集做一次Data Augmentation，然后把这些样本喂给那个，嗯，ResNet五十层，啊不对，应该是ResNet一百八层，用来提取特征，最后再通过一个Attention Mechanism把全局信息给聚合起来，整体的Accuracy大概提升了百分之四点二左右。", "clean": "其实这个Pipeline大概逻辑是，先是对原始数据集做一次Data Augmentation，然后把这些样本喂给ResNet-18，用来提取特征，最后再通过一个Attention Mechanism把全局信息给聚合起来，整体的Accuracy大概提升了4.2%左右。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl11.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 3, "source_record_id": "recheck.jsonl:11"}
{"oral": "我想想啊，这个实验结果其实得看这个learning rate，初始设成零点一…不对，零点零一…等等错了，是零点零零一！诶，如果你把batch size调到五百一十二，其实能缓解梯度爆炸问题吧，怎么说呢，主要是因为这个optimizer在收敛过程中，那个momentum参数给得太高，导致结果不稳定呀。", "clean": "实验结果其实得看learning rate，初始设成0.001！如果你把batch size调到512，其实能缓解梯度爆炸问题吧，主要是因为optimizer在收敛过程中，momentum参数给得太高，导致结果不稳定呀。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl12.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 4, "source_record_id": "recheck.jsonl:12"}
{"oral": "对了，我觉得这个 Attention 机制，Attention 机制得优化，之前调到零点五，不对，零点三，零点三才收敛，得重新跑跑实验。", "clean": "我觉得这个 Attention 机制得优化，之前调到 0.3 才收敛，得重新跑跑实验。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl29.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 11, "source_record_id": "recheck.jsonl:29"}
{"oral": "关于这个模型架构，我打算在Encoder部分引入，引入一种注意力机制，我想想，其实是想用那个，那个学习率设成零点零一，不对，零点零零一，等等错了，是一乘十的负四次方！然后我我我打算把这个数据集扩充，扩充到五千个样本，这样能提升Generalization能力，但那个，那个维度应该是两百五十六，不对，五百一十二，等等错了，是一千零二十四！这这这部分可能得重新推导一下。", "clean": "关于这个模型架构，我打算在Encoder部分引入一种注意力机制。学习率设成1e-4！然后我打算把这个数据集扩充到5000个样本，这样能提升Generalization能力，但维度应该是1024！这部分可能得重新推导一下。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl16.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 6, "source_record_id": "recheck.jsonl:16"}
{"oral": "哦对了，关于那个 ABLATION 实验的部分，我觉得你把那个 baseline 设得太高了，话说，如果咱们把 learning rate 调到零点零零一，那个 convergence 速度可能反而会慢下来，嗯，得重新审视一下，哎，再就是那个数据集在预处理阶段，好像有百分之五点三的数据点丢失了，这对 result 影响挺大，得赶紧给补上，不然审稿人肯定会质疑咱们那个 model 的 robustness。", "clean": "关于Ablation实验的部分，我觉得你把baseline设得太高了。如果咱们把learning rate调到0.001，convergence速度可能反而会慢下来，得重新审视一下。再就是数据集在预处理阶段，好像有5.3%的数据点丢失了，这对result影响挺大，得赶紧给补上，不然审稿人肯定会质疑咱们model的robustness。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl15.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 5, "source_record_id": "recheck.jsonl:15"}
{"oral": "嗯，我觉得这篇论文核心在于那个 Attention mechanism 的优化，然后它在数据集上跑出了百分之八十二的准确率，哦不是，是百分之八十八，另外它对 Hyper parameter 的敏感度分析做得挺细，还有就是关于 LORA 适配器在多模态场景下怎么说呢，其实推理延迟降低了约三十个毫秒吧，整体逻辑很强呀。", "clean": "我觉得这篇论文核心在于 Attention mechanism 的优化，它在数据集上跑出了 88% 的准确率，另外它对 Hyper-parameter 的敏感度分析做得挺细，还有关于 LoRA 适配器在多模态场景下，推理延迟降低了约 30ms，整体逻辑很强。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl2.wav", "scene": "academic", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 9, "source_record_id": "recheck.jsonl:2"}
{"oral": "说白了，咱们现在得调整调整这个 Learning Rate，Learning Rate 设得太高了会导致梯度爆炸，而且那个...那个 Batch Size 应该得...得调到五百一十二，现在这个设置在训练集上收敛太慢，简单来说，就是我们需要重新重新定义这个 Loss Function，可能得把 L two 正则项加进去，才能让模型...才能够泛化，得在二零二五年三月一号之前跑完这组对比实验。", "clean": "说白了，咱们现在得调整调整 Learning Rate，Learning Rate 设得太高了会导致梯度爆炸，而且 Batch Size 应该得调到 512，现在这个设置在训练集上收敛太慢。简单来说，就是我们需要重新定义这个 Loss Function，可能得把 L2 正则项加进去，才能让模型能够泛化，得在 2025年3月1日之前跑完这组对比实验。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl21.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 10, "source_record_id": "recheck.jsonl:21"}
{"oral": "王教授，嗯，关于这个 Transformer 架构，就是我想想啊，把参数量设成五百一十二万吧。", "clean": "王教授，关于Transformer架构，把参数量设成512万吧。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl31.wav", "scene": "academic", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 14, "source_record_id": "recheck.jsonl:31"}
{"oral": "哎呀太离谱了，嗯，这篇论文采样率设在二十赫兹…不对，五十赫兹…等等错了，是一百赫兹！但那那那个 SNR 提升太假，怎么说呢，居然能跑出百分之十五点二 提升，我觉得这这这这个 Baseline 根本没对齐吧，我想想啊，应该是把数据集分成了三份…不对，四份…等等错了，是五份！才出来的结果哈。", "clean": "太离谱了，这篇论文采样率设在100Hz！但SNR提升太假，居然能跑出15.2%提升，我觉得这个Baseline根本没对齐吧，应该是把数据集分成了5份才出来的结果哈。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl30.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 13, "source_record_id": "recheck.jsonl:30"}
{"oral": "说白了，我这套框架，嗯，其实是在解决 Cross Attention 机制里，怎么说呢，就是计算量太大的问题，我想想啊，如果把那个，把所有的 Token 映射到，诶，不对，应该是通过一个轻量化 LORA 模块，把参数量压到五百一十二个，这样在推理阶段吧，效率能提升百分之十五点三，也就是，简单来说，就是牺牲一点点精度，换取推理速度，哈。", "clean": "说白了，我这套框架其实是在解决 Cross-Attention 机制里计算量太大的问题。应该是通过一个轻量化 LoRA 模块，把参数量压到 512 个，这样在推理阶段效率能提升 15.3%，简单来说，就是牺牲一点点精度换取推理速度。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl3.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 12, "source_record_id": "recheck.jsonl:3"}
{"oral": "虽然王教授说这个 Baseline 效果好，效果好，但我认为这个模型在处理五百二十个样本时，这个样本量…怎么说呢…样本量太小了，太小了，没法证明 S-O-T-A 性能。", "clean": "虽然王教授说这个 Baseline 效果好，但我认为这个模型在处理520个样本时，样本量太小了，没法证明 SOTA 性能。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl35.wav", "scene": "academic", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 16, "source_record_id": "recheck.jsonl:35"}
{"oral": "那个，我也不太确定，但是感觉这里的 Loss function 应该是用 Cross Entropy，啊不对，是 Mean Squared Error，因为咱们处理的是回归问题，然后那个 Learning Rate 设置成零点零零一可能太小，导致收敛速度太慢，嗯，可能需要调高到零点零一，这样在 Validation set 上观察到的梯度下降才会更明显，我大概就是这么想", "clean": "我也不太确定，但感觉这里的 Loss function 应该是用 Mean Squared Error，因为咱们处理的是回归问题。Learning Rate 设置成 0.001 可能太小，导致收敛速度太慢，可能需要调高到 0.01，这样在 Validation set 上观察到的梯度下降才会更明显，我大概就是这么想。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl38.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 17, "source_record_id": "recheck.jsonl:38"}
{"oral": "嗯，我想想啊，与其用 ViT 这种纯 Transformer 架构，不如直接上 Stable Diffusion 3 吧。我觉得我觉得，用 Stable Diffusion 3 比 CLIP 更能捕捉细节哈。数据集那边，咱们用 Multi30K...不对，Multi300K...等等错了，是 Multi30K！怎么说呢，咱们得看 AUC 表现，看 AUC 表现能不能比之前提升百分之二点五呀。", "clean": "与其用 ViT 这种纯 Transformer 架构，不如直接上 Stable Diffusion 3 吧。我觉得用 Stable Diffusion 3 比 CLIP 更能捕捉细节。数据集那边，咱们用 Multi30K！咱们得看 AUC 表现，看 AUC 表现能不能比之前提升 2.5% 呀。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl34.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 15, "source_record_id": "recheck.jsonl:34"}
{"oral": "我也不太确定，但是感觉感觉这个 Top one Accuracy 掉这么多，可能是 C4 数据集预处理有问题。我想想，如果我们在 KV Cache 那块，KV Cache 优化得不到位，或者说，如果量化量化得太狠，比如 Quantization 选了四比特，可能会影响 RAG 检索效果。", "clean": "我也不太确定，但是感觉这个 Top-1 Accuracy 掉这么多，可能是 C4 数据集预处理有问题。我想想，如果我们在 KV Cache 那块，KV Cache 优化得不到位，或者说，如果量化得太狠，比如 Quantization 选了 4-bit，可能会影响 RAG 检索效果。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl41.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 19, "source_record_id": "recheck.jsonl:41"}
{"oral": "对了，你们做这道题的时候，注意把那个 Learning Rate 设成零点零零一，哦不是，应该是零点零零五，嗯，然后你们那个，如果把，怎么说呢，如果把 Batch Size 调到五百一十二的话，计算开销可能会太大，哈，建议先试下两百五十六看能不能跑通吧。", "clean": "你们做这道题的时候，注意把 Learning Rate 设成 0.005，然后如果把 Batch Size 调到 512 的话，计算开销可能会太大，建议先试下 256 看能不能跑通吧。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl42.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 20, "source_record_id": "recheck.jsonl:42"}
{"oral": "说白了，我觉得我觉得这个 Loss 曲线在震荡，简单来说，咱们学习率设零点一…不对，零点零一…等等错了，是一点零乘十的负四次方！再看这个 Batch Size，Batch Size 设五百一十二，其实其实没效果，那个...数据集...内存溢出...得重新跑，得重新跑，再优化下这个 LORA 层，它可能没收敛到", "clean": "说白了，我觉得 Loss 曲线在震荡。简单来说，咱们学习率设 1.0e-4！再看这个 Batch Size，Batch Size 设 512 其实没效果，数据集内存溢出，得重新跑，再优化下这个 LoRA 层，它可能没收敛。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl4.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 18, "source_record_id": "recheck.jsonl:4"}
{"oral": "嗯，现在这个模型的Zero shot能力，我想想啊，其实在处理那五百六十个样本时，表现得...其实应该是说，如果把Temperature调到零点七，结果会更稳吧，就是目前这种...由于分布偏移导致...这种case太多，导致整体Accuracy掉了一点，诶，你觉得这种方向还能走通吗？", "clean": "现在这个模型的Zero-shot能力，在处理560个样本时，如果把Temperature调到0.7，结果会更稳吧。目前由于分布偏移导致这种case太多，导致整体Accuracy掉了一点，你觉得这种方向还能走通吗？", "usage_text_words": null, "audio": "audio/academic__recheckjsonl43.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 21, "source_record_id": "recheck.jsonl:43"}
{"oral": "我觉得这个模型得换成Transformer...对了，核心是得提升泛化能力，泛化能力，因为现在的Overfitting太严重，那个数据集规模在三千个...不对，四千个...等等错了，是五千个！我们尝试...怎么说呢...如果把Dropout调高的话，能缓解这个问题，缓解这个问题，那个实验结果...哎呀对了，我们得重新跑一遍ABLATION study。", "clean": "我觉得这个模型得换成Transformer，核心是得提升泛化能力，因为现在的Overfitting太严重，数据集规模在5000个。我们尝试如果把Dropout调高的话，能缓解这个问题，实验结果我们得重新跑一遍Ablation study。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl45.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 23, "source_record_id": "recheck.jsonl:45"}
{"oral": "嗯，我发现这个 model 在 zero shot 场景下，它的泛化泛化能力其实挺强，但我们在测试集上跑出来的 accuracy 只有百分之六十二，啊不对，是百分之六十八，怎么说呢，我觉得我觉得这个 gap 主要是因为 prompt 没对齐，得再调调吧。", "clean": "我发现这个 model 在 zero shot 场景下，它的泛化能力其实挺强，但我们在测试集上跑出来的 accuracy 只有68%，我觉得这个 gap 主要是因为 prompt 没对齐，得再调调吧。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl44.wav", "scene": "academic", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 22, "source_record_id": "recheck.jsonl:44"}
{"oral": "我想请教一下，您在实验里先是用了那个Contrastive Learning，Contrastive Learning来做预训练，然后把模型在ImageNet数据集上跑了四十个epoch，不对，是五十个epoch，最后通过Fine tuning微调到下游任务，但我觉得这个Fine-tuning过程，这个Fine tuning过程可能导致过拟合，所以想问问您具体怎么平衡学习率？", "clean": "我想请教一下，您在实验里先是用Contrastive Learning来做预训练，然后把模型在ImageNet数据集上跑了50个epoch，最后通过Fine-tuning微调到下游任务，但我觉得这个Fine-tuning过程可能导致过拟合，所以想问问您具体怎么平衡学习率？", "usage_text_words": null, "audio": "audio/academic__recheckjsonl47.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 24, "source_record_id": "recheck.jsonl:47"}
{"oral": "我觉得这种方案，这种方案比直接用 Full Fine tuning 更好，与其追求参数量，追求参数量…怎么说呢…比提升 Zero shot 能力更关键。", "clean": "我觉得这种方案比直接用 Full Fine-tuning 更好，与其追求参数量，比提升 Zero-shot 能力更关键。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl6.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 29, "source_record_id": "recheck.jsonl:6"}
{"oral": "张教授，嗯，这篇文章在 SuperGLUE 上跑分挺高，但我看它对 log P y given x 的处理方式，怎么说呢，其实是……应该是想通过优化分布来降 WER，但逻辑上有点绕，吧。而且 ViT 这种结构，要是想把 KV Cache 规模压到五百一十二以下，可能很难实现。", "clean": "张教授，这篇文章在 SuperGLUE 上跑分挺高，但我看它对 $\\log P(y|x)$ 的处理方式，应该是想通过优化分布来降 WER，但逻辑上有点绕。而且 ViT 这种结构，要是想把 KV Cache 规模压到 512 以下，可能很难实现。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl53.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 27, "source_record_id": "recheck.jsonl:53"}
{"oral": "王教授，我之前主要在搞强化学习那边，尝试把PPO结合到Mixture of Experts架构里，具体参考了DeepSeek V3那种设计，嗯，在计算log P X的时候，我最初把IoU阈值设成零点五，啊不对，是零点七，然后发现收敛速度快了大概百分之十五左右。", "clean": "王教授，我之前主要在搞强化学习那边，尝试把PPO结合到Mixture of Experts架构里，具体参考了DeepSeek-V3那种设计，在计算log P X的时候，我最初把IoU阈值设成0.7，然后发现收敛速度快了大概15%左右。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl54.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 28, "source_record_id": "recheck.jsonl:54"}
{"oral": "关于这次模型在通用能力上表现，我把那个 ViT 结构稍微调了一下，我觉得我觉得这个方向能跑通，结果 MMLU 跑出来百分之五十二...不对，百分之五十五...等等错了，是百分之五十八！而且 Precision 提升提升了大概百分之三点五，Top one Accuracy 提升提升了百分之二点一，这就证明 E等于mc方 那个理论推导在实际场景里其实有效。", "clean": "关于这次模型在通用能力上的表现，我把ViT结构稍微调了一下，我觉得这个方向能跑通，结果MMLU跑出来58%！而且Precision提升了大概3.5%，Top-1 Accuracy提升了2.1%，这就证明E=mc^2那个理论推导在实际场景里其实有效。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl49.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 26, "source_record_id": "recheck.jsonl:49"}
{"oral": "我觉得比起单纯追求 CIDEr 分数，CIDEr 分数，其实去分析 Attention 权重比那个更关键。这篇论文用数据集是 Common Crawl...不对，LAION...等等错了，是 Common Crawl！而且与其把模型规模堆到...怎么说呢...与其通过增加参数量来提升，不如直接优化 CLIP 这种对齐方式，优化对齐方式。我记得那个采样率是百分之二...不对，百分之五...等等错了，是百分之十！这样对 Sora 这种生成质量提升更明显。", "clean": "我觉得比起单纯追求 CIDEr 分数，其实去分析 Attention 权重比那个更关键。这篇论文用数据集是 Common Crawl！而且与其通过增加参数量来提升，不如直接优化 CLIP 这种对齐方式。我记得那个采样率是 10%！这样对 Sora 这种生成质量提升更明显。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl48.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 25, "source_record_id": "recheck.jsonl:48"}
{"oral": "说白了，咱们这次对比，对比 DeepSeek V3 和 PaLM 2 主要是看推理能力，嗯，然后把 Sora 和 SAM 2 这种多模态也拉进来跑一遍，我想想啊，指标用 BLEU 四点零，不对，是 BLEU 四点五，看看能不能把性能性能拉上去吧。", "clean": "说白了，咱们这次对比 DeepSeek-V3 和 PaLM 2 主要是看推理能力，然后把 Sora 和 SAM 2 这种多模态也拉进来跑一遍，指标用 BLEU 4.5，看看能不能把性能拉上去吧。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl61.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 31, "source_record_id": "recheck.jsonl:61"}
{"oral": "诶，不是你想的那样，其实这不是个 PPO 问题。我想想啊，咱们得看 Recall，Recall 是… Recall 是在百分之八十五…不对，百分之七十五…等等错了，是百分之六十五！就是 DeepSeek V3 用 DPO 也就是 Direct Preference Optimization 解决 解决 这个问题 嘛。", "clean": "不是你想的那样，其实这不是个PPO问题。咱们得看Recall，Recall是在65%！就是DeepSeek-V3用DPO，也就是Direct Preference Optimization解决这个问题嘛。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl62.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 32, "source_record_id": "recheck.jsonl:62"}
{"oral": "你说那个 SAM 能不能直接套在 ViT 这种 backbone 上做 PEFT，嗯，我看他论文里写用 DALL E 2...哦不是，是 DALL E 3 做的数据增强，效果提升了百分之十二点五，我觉得可行。", "clean": "你说 SAM 能不能直接套在 ViT 这种 backbone 上做 PEFT，我看他论文里写用 DALL-E 3 做的数据增强，效果提升了 12.5%，我觉得可行。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl66.wav", "scene": "academic", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 34, "source_record_id": "recheck.jsonl:66"}
{"oral": "嗯，你们现在投简历，得关注关注多模态。比如你们在面试里聊 Sora 或者 DALL E 3，怎么说呢，不要只说效果好，得聊聊底层逻辑，比如 Stable Diffusion 3 这种架构，架构在什么地方起作用。而且我想想啊，如果你们讨论评估指标，别只盯着 BLEU，要把 METEOR 这种指标拿出来对比对比。其实关于优化器，你们可以用 Adafactor，用 Adafactor 这种省显存方案，其实…或者说，如果你们能证明在内存受限环境下，这个方案能跑通，竞争力就强吧。", "clean": "你们现在投简历，得关注多模态。比如你们在面试里聊 Sora 或者 DALL-E 3，不要只说效果好，得聊聊底层逻辑，比如 Stable Diffusion 3 这种架构，架构在什么地方起作用。而且如果你们讨论评估指标，别只盯着 BLEU，要把 METEOR 这种指标拿出来对比对比。其实关于优化器，你们可以用 Adafactor 这种省显存方案，或者说，如果你们能证明在内存受限环境下，这个方案能跑通，竞争力就强吧。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl60.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 30, "source_record_id": "recheck.jsonl:60"}
{"oral": "对了，我想想啊，咱们这次投 CVPR 吧，不对，是投 ICCV，反正得把 SAM 2 那个效果给体现出来。我觉得我觉得用 Qwen 2.5 做 prompt 增强应该行，就是 Common Crawl 的数据集 Common Crawl 规模实在太大，可能得先采样百分之十五吧。", "clean": "咱们这次投ICCV，反正得把SAM 2那个效果给体现出来。我觉得用Qwen-2.5做prompt增强应该行，Common Crawl的数据集规模实在太大，可能得先采样15%吧。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl68.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 35, "source_record_id": "recheck.jsonl:68"}
{"oral": "对了，咱们那个T5，T5模型在跑的时候，如果你用Adafactor，Adafactor优化器的话，内存压力会小很多。哦对了，关于那个指标，我觉得咱们得看Top one Accuracy，Top one Accuracy，不能只盯着METEOR，METEOR这个指标，怎么说呢，如果只看这个的话可能会误导，得综合评估。话说，那个FlashAttention，FlashAttention如果是部署在八块显卡上，那速度应该是，应该是能快三倍左右。", "clean": "咱们T5模型在跑的时候，如果你用Adafactor优化器的话，内存压力会小很多。关于那个指标，我觉得咱们得看Top-1 Accuracy，不能只盯着METEOR这个指标，如果只看这个的话可能会误导，得综合评估。FlashAttention如果是部署在8块显卡上，那速度应该是能快3倍左右。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl64.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 33, "source_record_id": "recheck.jsonl:64"}
{"oral": "我觉得与其继续在这个数据集上做Fine tuning，不如直接试试那个，那个，那那那个PEFT框架，毕竟比单纯调参更高效。我们那个，要是把，怎么说呢，如果把模型规模定在七十亿，不对，十三亿，等等错了，是七十亿参数！其实我我我觉得，比之前那个方案更稳妥，只要把学习率设成零点零零一就行。", "clean": "我觉得与其继续在这个数据集上做Fine-tuning，不如直接试试PEFT框架，毕竟比单纯调参更高效。如果把模型规模定在70亿参数，其实我觉得比之前那个方案更稳妥，只要把学习率设成0.001就行。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl7.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 37, "source_record_id": "recheck.jsonl:7"}
{"oral": "这次在WikiText上测完发现，CIDEr分数提升了百分之十二点五，不对，应该是百分之八点二。我觉得我觉得这个结果主要是因为引入FlashAttention之后，推理速度快了很多，快了很多。至于IoU指标，我尝试用DALL E 3生成样本去对齐，但DALL E 2，哦不是，DALL E 3生成图像边缘太模糊，导致指标没上去。", "clean": "这次在WikiText上测完发现，CIDEr分数提升了8.2%。我觉得这个结果主要是因为引入FlashAttention之后，推理速度快了很多。至于IoU指标，我尝试用DALL-E 3生成样本去对齐，但DALL-E 3生成图像边缘太模糊，导致指标没上去。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl69.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 36, "source_record_id": "recheck.jsonl:69"}
{"oral": "说白了，这篇论文在 GLUE 榜单上刷分太猛，嗯，但在 SQuAD 上表现一般，而且作者拿 Grok 1 做对比，不对，应该是拿 T5 做对比，其实在泛化能力这块还是差了一截吧。", "clean": "说白了，这篇论文在 GLUE 榜单上刷分太猛，但在 SQuAD 上表现一般，而且作者拿 T5 做对比，其实在泛化能力这块还是差了一截吧。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl75.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 41, "source_record_id": "recheck.jsonl:75"}
{"oral": "哎呀太离谱了，嗯，用 Direct Preference Optimization 刷出来的 MMLU 竟然涨了百分之十二点五，啊不，是百分之十四点二吧，就是那个 RoPE 结合 MoE 之后，在 T5 架构上表现太稳了呀。", "clean": "太离谱了，用 Direct Preference Optimization 刷出来的 MMLU 竟然涨了 14.2%，RoPE 结合 MoE 之后，在 T5 架构上表现太稳了呀。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl74.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 40, "source_record_id": "recheck.jsonl:74"}
{"oral": "对了，我想想啊，咱们在 COCO 上跑实验的时候，那个 NDCG 提升了百分之三点二，啊不对，是百分之四点八，这个指标指标其实挺关键吧。然后就是，关于 Mixture of Experts 那个架构，要是能把，怎么说呢，如果把 KV Cache 优化一下，能省不少显存哈。话说 BERT 预训练模型，模型规模现在是五百一十二万，诶，五百一十二万个参数，估计得重新调下学习率呀。", "clean": "咱们在 COCO 上跑实验的时候，NDCG 提升了 4.8%，这个指标其实挺关键吧。然后关于 Mixture of Experts 架构，如果把 KV Cache 优化一下，能省不少显存哈。话说 BERT 预训练模型规模现在是 512 万个参数，估计得重新调下学习率呀。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl71.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 38, "source_record_id": "recheck.jsonl:71"}
{"oral": "嗯，不是说 SAM 没对齐就不能跑，其实是 DeepSeek V3 这种规模模型在 ViT 结构上容易梯度爆炸，怎么说呢，我想想啊，我们在 MS COCO 上测 Top one Accuracy 只有百分之十二点五，吧，应该是学习率设太高了呀。", "clean": "不是说 SAM 没对齐就不能跑，其实是 DeepSeek-V3 这种规模模型在 ViT 结构上容易梯度爆炸，我们在 MS COCO 上测 Top-1 Accuracy 只有 12.5%，应该是学习率设太高了。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl77.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 43, "source_record_id": "recheck.jsonl:77"}
{"oral": "嗯，你得先关注一下 Qwen2 怎么处理 KV Cache，然后就是看 CLIP 的 alignment 效果，另外 Accuracy 这块，我想想啊，之前测的是百分之八十二…不对，百分之八十五…等等错了，是百分之八十八！诶，最后还得再把 SAM 2 部署到这个环境里去吧。", "clean": "你得先关注一下 Qwen2 怎么处理 KV Cache，然后看 CLIP 的 alignment 效果，另外 Accuracy 这块，之前测的是 88%！最后还得再把 SAM 2 部署到这个环境里去吧。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl76.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 42, "source_record_id": "recheck.jsonl:76"}
{"oral": "嗯, 我之前投递的时候发现现在的面试官特别在意数据规模, 比如你用的是 Pile 还是 SQuAD 一点零, 哦不是, 是 SQuAD 二点零, 这种数据集做微调, 就是, 现在的趋势是得把 FlashAttention 这种推理加速写进简历里, 怎么说呢, 哪怕只是跑过 DALL E 3 这种模型或者刷过 sacreBLEU 分数吧, 只要能证明你懂工程落地, 机会大很多, 你觉得呢?", "clean": "我之前投递的时候发现现在的面试官特别在意数据规模，比如你用的是 Pile 还是 SQuAD 2.0 这种数据集做微调，现在的趋势是得把 FlashAttention 这种推理加速写进简历里，哪怕只是跑过 DALL-E 3 这种模型或者刷过 sacreBLEU 分数吧，只要能证明你懂工程落地，机会大很多，你觉得呢？", "usage_text_words": null, "audio": "audio/academic__recheckjsonl86.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 45, "source_record_id": "recheck.jsonl:86"}
{"oral": "嗯，我之前的研究结论就得出了用 Mixture of Experts 能够显著提升模型在长文本下的泛化能力吧。怎么说呢，主要原因是在引入了 RAG 之后，我们发现传统的 arg max 采样容易导致输出崩塌，啊不对，是导致分布偏移，所以哈，我尝试用 DINOv2 提取的视觉特征来增强对齐。我想想啊，在 HumanEval 上的得分提升了百分之十二点五，也就是在推理能力上得到了明显增强，所以这个方案在实际部署时挺高效的呀。", "clean": "我之前的研究结论得出了用 Mixture of Experts 能够显著提升模型在长文本下的泛化能力。主要原因是在引入了 RAG 之后，我们发现传统的 argmax 采样容易导致分布偏移，所以我尝试用 DINOv2 提取的视觉特征来增强对齐。在 HumanEval 上的得分提升了 12.5%，也就是在推理能力上得到了明显增强，所以这个方案在实际部署时挺高效的。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl73.wav", "scene": "academic", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 39, "source_record_id": "recheck.jsonl:73"}
{"oral": "哎呀太夸张了，嗯，Phi 3 居然能通过 Chain of Thought 产生这种推理能力，就是，我们在做 arg max 采样的时候，采样的时候发现，我想想啊，结果结果完全超出了预期，诶，本来以为 DALL E 3 这种规模…其实是说，在视觉生成端才强，结果 Phi 3 竟然也这么猛吧。", "clean": "太夸张了，Phi-3居然能通过Chain of Thought产生这种推理能力，我们在做argmax采样的时候发现，结果完全超出了预期，本来以为DALL-E 3这种规模在视觉生成端才强，结果Phi-3竟然也这么猛吧。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl90.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 47, "source_record_id": "recheck.jsonl:90"}
{"oral": "王教授，嗯，我打算在 baseline 部分对比 BERT 和 DeepSeek V3，然后针对 Diffusion 模型引入 FlashAttention 优化，怎么说呢，我想把 ROUGE 指标提升到百分之二十，啊不对，是百分之二十五吧。", "clean": "王教授，我打算在 baseline 部分对比 BERT 和 DeepSeek-V3，然后针对 Diffusion 模型引入 FlashAttention 优化，我想把 ROUGE 指标提升到 25%。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl91.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 48, "source_record_id": "recheck.jsonl:91"}
{"oral": "你说这次投稿投 CVPR 能不能行？我看了下最近那个 Baseline 提升了百分之十二，哦不是，百分之十五点二，而且我们那个 SOTA 结果在两个数据集上都稳住了，虽然 Ablation Study 还缺一点，但我觉得整体逻辑能跑通，应该可以。", "clean": "你说这次投稿投 CVPR 能不能行？我看了下最近那个 Baseline 提升了15.2%，而且我们那个 SOTA 结果在两个数据集上都稳住了，虽然 Ablation Study 还缺一点，但我觉得整体逻辑能跑通，应该可以。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl97.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 51, "source_record_id": "recheck.jsonl:97"}
{"oral": "王教授, 这篇论文在 CLIP 对齐上做得还行, 但那个预训练语料库规模, 应该是十亿级...不对, 五百亿...等等错了, 是 Common Crawl 规模的！而且他用 Qwen 2 做 backbone 配合 Retrieval Augmented Generation, 在 Cityscapes 验证集上效果波动挺大, 感觉采样策略没对齐。", "clean": "王教授，这篇论文在 CLIP 对齐上做得还行，但预训练语料库规模是 Common Crawl 规模的！而且他用 Qwen-2 做 backbone 配合 Retrieval Augmented Generation，在 Cityscapes 验证集上效果波动挺大，感觉采样策略没对齐。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl93.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 49, "source_record_id": "recheck.jsonl:93"}
{"oral": "老师，我打算先是把 Recall at K 重新跑一下，重新跑一下，然后...我想想...如果能把 AUC ROC 曲线画得更漂亮点，再把 Swin Transformer 换成更轻量级模型，最后最后对比一下 Top one Accuracy 和 PPO 调优后结果，这样投顶会机会更大。", "clean": "老师，我打算先把 Recall@K 重新跑一下，然后如果能把 AUC-ROC 曲线画得更漂亮点，再把 Swin Transformer 换成更轻量级模型，最后对比一下 Top-1 Accuracy 和 PPO 调优后结果，这样投顶会机会更大。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl88.wav", "scene": "academic", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 46, "source_record_id": "recheck.jsonl:88"}
{"oral": "我想把那个，那那那个，Attention机制改成一种，呃，轻量化模型，大概能降三成计算量。", "clean": "我想把Attention机制改成一种轻量化模型，大概能降30%计算量。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl98.wav", "scene": "academic", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 52, "source_record_id": "recheck.jsonl:98"}
{"oral": "First I cannot log into my account password reset is not working and I just want to cancel my Disney plus subscription. Then I checked order four four nine two one zero and it says item shows as delivered but it's not here. Finally my email is Sally at i cloud dot com", "clean": "First, I cannot log into my account, password reset is not working, and I just want to cancel my Disney+ subscription.\nThen, I checked order 449210 and it says item shows as delivered but it's not here.\nFinally, my email is Sally@icloud.com", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl102.wav", "scene": "customer_service", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 55, "source_record_id": "recheck.jsonl:102"}
{"oral": "我觉得我觉得，与其死磕这个 DINOv2 预训练权重，不如直接在 DPO 阶段做对齐。你看这个 NDCG 指标，指标完全没起来，比之前那个版本更低。而且这个求和公式，sum i equals one to n x i，算出来结果是零点五，不对，是零点三二。咱们得思考思考，如果把 IoU 阈值调高，怎么说呢，如果把阈值设在零点七，可能效果会再好一点。现在的收敛速度太慢，慢到根本没法看，得重新检查检查学习率设定得对不对。", "clean": "我觉得与其死磕这个 DINOv2 预训练权重，不如直接在 DPO 阶段做对齐。你看这个 NDCG 指标，指标完全没起来，比之前那个版本更低。而且这个求和公式 $\\sum_{i=1}^{n} x_i$，算出来结果是 0.32。咱们得思考思考，如果把 IoU 阈值调高，如果把阈值设在 0.7，可能效果会再好一点。现在的收敛速度太慢，慢到根本没法看，得重新检查检查学习率设定得对不对。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl95.wav", "scene": "academic", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 50, "source_record_id": "recheck.jsonl:95"}
{"oral": "我觉得与其在这个Baseline上死磕，不如尝试引入Attention机制，毕竟对比实验显示，多头注意力比单一的MLP性能提升了百分之十二，不对，是百分之十五点八。而且如果把，怎么说呢，如果把这个维度从五百一十二调到一千零二十四，可能对收敛速度影响更道。其实在这种高维空间里，比起直接做映射，再考虑一个LORA的低秩分解可能更有效，这样能显著降低显存占用。", "clean": "我觉得与其在这个Baseline上死磕，不如尝试引入Attention机制，毕竟对比实验显示，多头注意力比单一的MLP性能提升了15.8%。而且如果把这个维度从512调到1024，可能对收敛速度影响更大。其实在这种高维空间里，比起直接做映射，再考虑一个LoRA的低秩分解可能更有效，这样能显著降低显存占用。", "usage_text_words": null, "audio": "audio/academic__recheckjsonl99.wav", "scene": "academic", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 53, "source_record_id": "recheck.jsonl:99"}
{"oral": "Oh my god! I am calling because I want to request a refund for a defective product, specifically for Disney plus. Uh, I mean, I already sent an email to billing dash dept at service dot net but nothing happened. The ticket number is TKT zero zero nine one eight two seven three. I said, TKT zero zero nine one eight two seven three. Just want my money back, ya. Anyway, I appreciate your help, thanks.", "clean": "Oh my god! I am calling because I want to request a refund for a defective product, specifically for Disney+, I already sent an email to billing-dept@service.net but nothing happened. The ticket number is TKT00918273. Just want my money back, ya. Anyway, I appreciate your help, thanks.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl104.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 56, "source_record_id": "recheck.jsonl:104"}
{"oral": "Hi, uh, I'm calling because my Wells Fargo statement looks wrong. I'm not totally sure but I think I was overcharged. I see a charge for sixty four fifteen, but it should be nineteen ninety five. I tried checking my inbox, Linda at yahoo dot co dot uk, but I can't find tracking number in my email, you know, just... I'm just confused why it's sixty four fifteen!", "clean": "Hi, I'm calling because my Wells Fargo statement looks wrong. I'm not totally sure but I think I was overcharged. I see a charge for $64.15, but it should be $19.95. I tried checking my inbox, Linda@yahoo.co.uk, but I can't find the tracking number in my email, I'm just confused why it's $64.15!", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl105.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 57, "source_record_id": "recheck.jsonl:105"}
{"oral": "Hi uh I have a question. I bought this thing and I already opened it but I want to return it. I'm not totally sure but I think... I think the box is open. Can I still return it? Also I have this other issue. I need to dispute a transaction I don't recognize. It's for Spotify and maybe Disney plus. Uh the order number is TRK zero zero nine two eight four seven five six. I'm not sure if I'm in right place but if you can't help just call me back at two one two five five five zero one four four. I really need to get this sorted out. I said call me back at two one two five five five zero one four four. Thanks.", "clean": "Hi, I have a question. I bought this thing and I already opened it but I want to return it. I'm not totally sure but I think the box is open. Can I still return it? Also I have this other issue. I need to dispute a transaction I don't recognize. It's for Spotify and maybe Disney+. The order number is TRK009284756. I'm not sure if I'm in the right place but if you can't help just call me back at 212-555-0144. I really need to get this sorted out. Call me back at 212-555-0144. Thanks.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl107.wav", "scene": "customer_service", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 59, "source_record_id": "recheck.jsonl:107"}
{"oral": "Actually... oh, wait. I just noticed something on my statement. Look, I am getting really frustrated here. I think I was charged twice for the same subscription. Like, the same exact charge twice! And then, get this, my credit card was declined but I have sufficient funds. I know I do. I have like five hundred dollars in there. I mean, I just used it for DoorDash like ten minutes ago and it worked fine. It worked fine! My email is Marcus at email dot com. Can you just check why it says I owe you two hundred forty nine ninety nine? I already paid that. I'm just... I'm just so tired of this. I keep getting the same answer. I need this fixed now. I mean, I really need this fixed now. Please just check the record. My email again is Marcus at email dot com. Let me know if you can see the double charge on the system.", "clean": "I just noticed something on my statement. Look, I am getting really frustrated here. I think I was charged twice for the same subscription. The same exact charge twice! And then, get this, my credit card was declined but I have sufficient funds. I know I do. I have $500 in there. I just used it for DoorDash 10 minutes ago and it worked fine. My email is Marcus@email.com. Can you just check why it says I owe you $249.99? I already paid that. I'm just so tired of this. I keep getting the same answer. I need this fixed now. Please just check the record. My email again is Marcus@email.com. Let me know if you can see the double charge on the system.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl106.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 58, "source_record_id": "recheck.jsonl:106"}
{"oral": "Hi yeah I need to cancel this subscription right now because honestly I received a damaged product. The order number is BKL nine nine two zero three XP. I mean... it was like... if I could just get a refund then we could maybe just call it even but it's one hundred forty seven fifty. One hundred forty seven fifty! And then there is this other thing... the order BLL twenty twenty four seven seven one two... uh... the promotional code isn't applying to my cart. It just... it's not working. I tried... let me see... if you can just fix the code then we could keep the account but honestly I just want out.", "clean": "Hi, I need to cancel this subscription right now because honestly I received a damaged product. The order number is BKL99203XP. If I could just get a refund then we could maybe just call it even, but it's $147.50. $147.50! And then there is this other thing... the order BLL20247712... the promotional code isn't applying to my cart. It's not working. I tried... let me see... if you can just fix the code then we could keep the account, but honestly I just want out.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl114.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 64, "source_record_id": "recheck.jsonl:114"}
{"oral": "Hi Sarah, look, I need to update the shipping address for my order, just... I need to update it because it's wrong. My claim number is BKL nine nine two zero three X. My email is sakura at yahoo dot com. The transaction code is TXN nine nine zero two one four four five... oh no, not four four five, it's four four zero. Uh, my phone number is... my phone number is five five five nine eight seven six five four three. I just... I want to make sure it gets to the right place, you know, so it doesn't get lost, ha.", "clean": "Hi Sarah, I need to update the shipping address for my order because it's wrong. My claim number is BKL99203X. My email is sakura@yahoo.com. The transaction code is TXN99021440. My phone number is 555-987-6543. I want to make sure it gets to the right place so it doesn't get lost.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl113.wav", "scene": "customer_service", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 63, "source_record_id": "recheck.jsonl:113"}
{"oral": "Hi, I'm calling because my package arrived totally smashed. I thought it was just the outer box, eh, but the item inside is broken too. The order number is ORD ninety nine twenty one C... no, it's ORD ninety nine twenty one B. The order order number is ORD ninety nine twenty one B. Just, I mean, I spent two hundred fifty four dollars and ten cents on this! The serial number is S four four nine one zero two eight three seven. I tried emailing help desk at service dot org, but no one answered. My phone number is five five five zero one two three four five six, just in case you need to reach me. How do we fix this, ya? I really need a replacement soon, soon. I hope you can help me sort this out, ha.", "clean": "Hi, I'm calling because my package arrived totally smashed. I thought it was just the outer box, but the item inside is broken too. The order number is ORD9921B. I spent $254.10 on this! The serial number is S449102837. I tried emailing help desk at service.org, but no one answered. My phone number is 555-012-3456, just in case you need to reach me. How do we fix this? I really need a replacement soon. I hope you can help me sort this out.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl108.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 60, "source_record_id": "recheck.jsonl:108"}
{"oral": "Honestly... I am just so frustrated right now. I just opened my package for order ninety nine two eight four seven one and it is completely smashed. Like, totally ruined. This thing cost twelve hundred dollars! Twelve hundred dollars! I have a ticket number, it is TICKET zero zero nine one eight two five... oh no, sorry, it is zero zero nine one eight two seven. Yeah, that one. I tried calling from my phone, plus one two zero two five five five zero one seven eight, but I just kept getting put on hold. I was even listening to Spotify just to keep my sanity while waiting! Can you just... I don't know, please just tell me how to get a refund or a replacement because this is ridiculous.", "clean": "Honestly, I am just so frustrated right now. I just opened my package for order 9928471 and it is completely smashed. Totally ruined. This thing cost $1,200! I have a ticket number, it is TICKET 0091827. I tried calling from my phone, +1 202-555-0178, but I just kept getting put on hold. I was even listening to Spotify just to keep my sanity while waiting! Can you please just tell me how to get a refund or a replacement because this is ridiculous.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl112.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 62, "source_record_id": "recheck.jsonl:112"}
{"oral": "Hi, yeah, I'm calling because my friend told me, quote, this place is the best for big groups, end quote, and honestly, I think I think it is. So, uh, I want to book a table for... wait... let me see... for twelve people. Oh, actually, thirteen people! Anyway, we are celebrating a work thing. I tried emailing your team at help desk at service dot org, but no reply. I mean, just, the email is help desk at service dot org. I'm worried about the budget, because my boss said, quote, we have three hundred fifty dollars, end quote, but then he changed it to two hundred fifteen twelve... no, wait, two hundred fifteen dollars and twelve cents. I think that's a bit tight, you know? I mean, I'm used to Spotify prices, not fancy dinners, ha! I just hope we can make it work, you know? I mean, I just... I want to make sure we get a spot, yeah.", "clean": "Hi, yeah, I'm calling because my friend told me, quote, \"this place is the best for big groups,\" end quote, and honestly, I think it is. So, I want to book a table for 13 people. Anyway, we are celebrating a work thing. I tried emailing your team at helpdesk@service.org, but no reply. The email is helpdesk@service.org. I'm worried about the budget, because my boss said, quote, \"we have $350,\" end quote, but then he changed it to $215.12. I think that's a bit tight. I'm used to Spotify prices, not fancy dinners, ha! I just hope we can make it work. I want to make sure we get a spot, yeah.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl110.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 61, "source_record_id": "recheck.jsonl:110"}
{"oral": "Hi, I am calling because I saw my bank statement and I was charged twice for the same subscription. It was ninety nine cents... ninety nine cents. And then there is this other charge for three hundred fifteen dollars and seventy five cents. I bought a high end blender, no, a coffee maker, and it arrived completely smashed. I am extremely disappointed, extremely disappointed with the quality. This product is totally broken, so I want to request a full refund for a defective product. I just want my money back to my account, to my account. I hope you can fix this in my bank a count.", "clean": "Hi, I am calling because I saw my bank statement and I was charged twice for the same subscription. It was $0.99. And then there is this other charge for $315.75. I bought a coffee maker, and it arrived completely smashed. I am extremely disappointed with the quality. This product is totally broken, so I want to request a full refund for a defective product. I just want my money back to my account. I hope you can fix this in my bank account.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl116.wav", "scene": "customer_service", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 66, "source_record_id": "recheck.jsonl:116"}
{"oral": "Hi I need help. My representative told me told me my card was sent but I never got it. I am calling Wells Fargo because I lost my credit card. The card number ends in CUST eight eight two nine PL. I really need a replacement replacement card. Also I tried to pay a fee of fifty four dollars and twenty cents but the promo code I entered is not applying to the total. The promo code is not applying. My shipping ID is SHIP nine nine two zero one X. Please check this for me.", "clean": "Hi, I need help. My representative told me my card was sent but I never got it. I am calling Wells Fargo because I lost my credit card. The card number ends in CUST8829PL. I really need a replacement card. Also, I tried to pay a fee of $54.20 but the promo code I entered is not applying to the total. The promo code is not applying. My shipping ID is SHIP99201X. Please check this for me.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl117.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 67, "source_record_id": "recheck.jsonl:117"}
{"oral": "Yeah hi, I am calling because I need to change my plan. I'm getting really frustrated here. I think I want to move to the basic package... uh... wait, not the basic, the silver plan. Does that sound right to you? Like, I just don't need all these extras. I mean, I have Hulu already so why am I paying for this? And since that... how should I put this... since that software update broke my device connectivity, I just can't even use the service. It's just not working. I saw an email from help dot desk at service dot org saying my bill is sixty five dollars and twenty cents. Wait, was it sixty five twenty... oh no, it was sixty eight twenty. Anyway, do you think we can just lower the monthly cost? I just need some help with this. I hope you can fix this issue for me, I really don't want to switch providers if I don't have too.", "clean": "Yeah hi, I am calling because I need to change my plan. I'm getting really frustrated here. I think I want to move to the silver plan. Does that sound right to you? I just don't need all these extras. I have Hulu already so why am I paying for this? And since that software update broke my device connectivity, I just can't even use the service. It's just not working. I saw an email from help.desk@service.org saying my bill is $68.20. Anyway, do you think we can just lower the monthly cost? I just need some help with this. I hope you can fix this issue for me, I really don't want to switch providers if I don't have to.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl115.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 65, "source_record_id": "recheck.jsonl:115"}
{"oral": "Oh my god! Look, I am calling about Wayfair. I have this pending order, and I need to update shipping address for it. The confirmation number is CONF seven seven three six two one zero. And the shipping ID is SHIP nine nine two eight three seven four one... oh wait, not four one, it is four two. Also, I can't even stay on line because my internet connection has been dropping every hour. Every single hour! I just need this fixed!", "clean": "Oh my god! Look, I am calling about Wayfair. I have this pending order, and I need to update the shipping address for it. The confirmation number is CONF7736210. And the shipping ID is SHIP99283742. Also, I can't even stay on line because my internet connection has been dropping every hour. Every single hour! I just need this fixed!", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl118.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 68, "source_record_id": "recheck.jsonl:118"}
{"oral": "Hi Sarah. My internet keeps dropping dropping. I cannot stay connected for more than ten minutes. My account ID number eight eight two seven one zero four four. The router says TRK zero zero nine two eight three seven four six five. Look, I paid three hundred ten twenty five last month and now my bill says one thousand fifty. I mean... it says... wait, it shows one thousand fifty dollars total. That is just wrong wrong. I only signed up for the five ninety nine add on. This is ridiculous!", "clean": "Hi Sarah. My internet keeps dropping. I cannot stay connected for more than 10 minutes. My account ID number 88271044. The router says TRK 0092837465. Look, I paid $310.25 last month and now my bill shows $1,050 total. That is just wrong. I only signed up for the $5.99 add on. This is ridiculous!", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl120.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 70, "source_record_id": "recheck.jsonl:120"}
{"oral": "Honestly, I am just so frustrated. I got my package but it is totally smashed. I used my Bank of America card for this, and I even tried calling DoorDash but no luck. Uh, my email is billie at yahoo dot co dot uk. The ID number is five five zero two nine three one one. Five five zero two nine three one one. And, just to make things worse, I need to update shipping address for an active order because I am moving. It is just a mess, really.", "clean": "Honestly, I am just so frustrated. I got my package but it is totally smashed. I used my Bank of America card for this, and I even tried calling DoorDash but no luck. My email is billie@yahoo.co.uk. The ID number is 55029311. And, just to make things worse, I need to update the shipping address for an active order because I am moving. It is just a mess, really.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl119.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 69, "source_record_id": "recheck.jsonl:119"}
{"oral": "Hi. I am calling because I see some extra charges on my bill. I think there is wrong billing address on invoice, it says something about Wells Fargo but it should be my personal account. And, uh, it is so weird because I tried to check it online but i can't log into my account after resetting the password. Now I am seeing a charge for forty nine ninety nine that looks like a DoorDash order? I never ordered that! Also the item I received is damaged, I mean the car had a huge scratch. Look, if you can just fix these charges then we could just settle this right now. I just want my money back, forty nine ninety nine, just give it back!", "clean": "Hi. I am calling because I see some extra charges on my bill. I think there is a wrong billing address on the invoice; it says something about Wells Fargo but it should be my personal account. And it is so weird because I tried to check it online but I can't log into my account after resetting the password. Now I am seeing a charge for $49.99 that looks like a DoorDash order? I never ordered that! Also the item I received is damaged, the car had a huge scratch. Look, if you can just fix these charges then we could just settle this right now. I just want my money back, $49.99, just give it back!", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl122.wav", "scene": "customer_service", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 72, "source_record_id": "recheck.jsonl:122"}
{"oral": "Hi, I need a refund for these extra charges on my bill. I mean, I just saw the statement and it's wrong. I rented a car, but somehow I'm seeing charges for... uh... Wayfair? And some Disney plus subscription? It makes no sense. My order number is order seven seven three one zero two. And there is this other code, B L L four four zero two ninety nine. I think... wait, let me see... I think there was a mistake with the account. I received a damaged item in the box... no, that's not right... I mean, the rental agreement was for a standard sedan, not these random things. Please just fix order seven seven three one zero two, okay?", "clean": "Hi, I need a refund for these extra charges on my bill. I just saw the statement and it's wrong. I rented a car, but somehow I'm seeing charges for Wayfair and some Disney+ subscription? It makes no sense. My order number is 773102. And there is this other code, BLL440299. I think there was a mistake with the account. The rental agreement was for a standard sedan, not these random things. Please just fix order 773102, okay?", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl121.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 71, "source_record_id": "recheck.jsonl:121"}
{"oral": "Hi I am calling about my claim. The agent told me the item shows as delivered but it is not here. It is not here! I mean look... look at the tracking. My claim number is SHIP four four eight two seven one zero... oh no sorry it is SHIP four four eight two seven one one. And my customer ID is CUST ninety nine two eight one X P. I also saw... I noticed... how should I put this... there is an unauthorized transaction on my statement. There is an unauthorized transaction on my statement from Xfinity and I just cannot believe this is happening.", "clean": "Hi, I am calling about my claim. The agent told me the item shows as delivered but it is not here. It is not here! Look at the tracking. My claim number is SHIP4482711. And my customer ID is CUST99281XP. I noticed there is an unauthorized transaction on my statement from Xfinity and I just cannot believe this is happening.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl123.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 73, "source_record_id": "recheck.jsonl:123"}
{"oral": "Hi, I need to change my appointment. I'm very disappointed with this... uh, just the whole scheduling thing. I have order nine nine two eight four seven one, and then order seven seven two one nine four zero, plus I got this T Mobile... wait, I mean, my ID number five five nine two one zero eight. I'm calling from the west garden... I mean, west yuan area, and I need to move it to next Tuesday, like, the twenty first, ya. I'm just really frustrated with how long it takes to get through to you guys, it's just... just too much.", "clean": "Hi, I need to change my appointment. I'm very disappointed with the whole scheduling thing. I have order 9928471, and then order 7721940, plus I got my ID number 5592108. I'm calling from the West Yuan area, and I need to move it to next Tuesday, the 21st. I'm just really frustrated with how long it takes to get through to you guys, it's just too much.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl124.wav", "scene": "customer_service", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 74, "source_record_id": "recheck.jsonl:124"}
{"oral": "Hi Sarah. Look, I am calling because my package never arrived. I am very disappointed with the service. It was supposed to be delivered by... uh... UPS... no, DHL... wait, that's wrong, it was FedEx! And it cost twelve forty nine. Twelve forty nine! Now I can't access streaming content on my smart TV and I think it's related to my T Mobile account. I just want to know where it's at, I've been waiting for days and I can't find the tracking info in my mail. It's just so frustrating, I mean, I really need this fixed right now because I paid for it and it's just not here. I just can't believe this happened in the middle of the week, I just can't.", "clean": "Hi Sarah. Look, I am calling because my package never arrived. I am very disappointed with the service. It was supposed to be delivered by FedEx! And it cost $12.49. Now I can't access streaming content on my smart TV and I think it's related to my T-Mobile account. I just want to know where it's at, I've been waiting for days and I can't find the tracking info in my mail. It's just so frustrating, I really need this fixed right now because I paid for it and it's just not here. I just can't believe this happened in the middle of the week, I just can't.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl125.wav", "scene": "customer_service", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 75, "source_record_id": "recheck.jsonl:125"}
{"oral": "Hi, uh, I think I lost my credit card and I need a replacement. I tried logging into Shopify to check my transactions but my password reset email is not arriving. My email is john at outlook dot com. I saw a charge from Hulu... oh no, not Hulu, it was Netflix, and that is why I got worried. Just, I am really stressed right now. I have a reference number here, it is SHIP ninety nine two zero one Y... wait, no, SHIP ninety nine two zero one X. Does that sound right to you? I mean, how long does it usually take to get a new card, right?", "clean": "Hi, I think I lost my credit card and I need a replacement. I tried logging into Shopify to check my transactions but my password reset email is not arriving. My email is john@outlook.com. I saw a charge from Netflix, and that is why I got worried. I am really stressed right now. I have a reference number here, it is SHIP99201X. Does that sound right to you? How long does it usually take to get a new card, right?", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl128.wav", "scene": "customer_service", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 77, "source_record_id": "recheck.jsonl:128"}
{"oral": "Hi. I just got my package and it's... it's totally smashed. This is completely unacceptable! I'm extremely disappointed with the service. I'm not totally sure but I think... I think the courier just threw it. The total was two hundred fifty dollars and forty cents. Two hundred fifty dollars and forty cents! I also have a Disney plus subscription linked to this account. My email is... my email is cici at outlook dot com.", "clean": "Hi. I just got my package and it's totally smashed. This is completely unacceptable! I'm extremely disappointed with the service. I'm not totally sure but I think the courier just threw it. The total was $250.40. I also have a Disney+ subscription linked to this account. My email is cici@outlook.com.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl130.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 78, "source_record_id": "recheck.jsonl:130"}
{"oral": "Hi, yeah, I need to request a refund... I mean, I'm calling because I need to request a refund for a duplicate payment. The order number is, uh, B Z seven seven one zero four two. B Z seven seven one zero four two. And, well, my order hasn't arrived after ten days. Ten days! Plus, I can't log into my account... I tried resetting... I can't log into my account after resetting password. Just to verify, my phone is four one five five five five zero one zero two, just let me know if you got that, ha.", "clean": "Hi, I'm calling because I need to request a refund for a duplicate payment. The order number is BZ771042. And my order hasn't arrived after 10 days. Plus, I can't log into my account after resetting password. Just to verify, my phone is 415-555-0102, just let me know if you got that.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl131.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 79, "source_record_id": "recheck.jsonl:131"}
{"oral": "Hi, I am calling because my subscription didn't cancel and I was billed, I think it was for Amazon, or wait... I mean I have a Wayfair account too, but this is for Amazon. The order number is seven seven three one zero two. I already tried to stop it but it just... it didn't go through. My customer ID is seven seven two zero four nine one one. I just really need to get this refund sorted out.", "clean": "Hi, I am calling because my subscription didn't cancel and I was billed for Amazon. The order number is 773102. I already tried to stop it but it didn't go through. My customer ID is 77204911. I just really need to get this refund sorted out.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl133.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 80, "source_record_id": "recheck.jsonl:133"}
{"oral": "Can someone tell me why there is a charge on my account? I mean, I just saw it on my Wells Fargo app and I don't recognize it. It says Hulu, it says Hulu for forty nine ninety nine. I'm just really confused about where my money went. Could you please help me with this? This is completely unacceptable. My account number is... account number is four zero two one eight five. I have no Hulu subscription, no Hulu subscription at all. This charge happened on October twelfth.", "clean": "Can someone tell me why there is a charge on my account? I just saw it on my Wells Fargo app and I don't recognize it. It says Hulu for $49.99. I'm just really confused about where my money went. Could you please help me with this? This is completely unacceptable. My account number is 402185. I have no Hulu subscription at all. This charge happened on October 12.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl134.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 81, "source_record_id": "recheck.jsonl:134"}
{"oral": "Hi. I'm calling because I'm not totally sure but I think my subscription was renewed without my permission. My subscription was renewed without my permission and I want a refund. I saw a charge from... Chase... no, Bank of America... wait that's wrong, it's Wells Fargo! It was for... five dollars... no, ten dollars... hold on, it's seven fifty! The order number is... S H I P twenty twenty three ninety nine... no, S H I P twenty twenty four ninety eight... sorry, it's S H I P twenty twenty four ninety nine one! Now I'm just confused if Hulu... if Hulu allows returns for opened items. I really need to know if Hulu allows returns because this was an accident.", "clean": "Hi. I'm calling because I'm not totally sure but I think my subscription was renewed without my permission and I want a refund. I saw a charge from Wells Fargo! It was for $7.50! The order number is SHIP2024991! Now I'm just confused if Hulu allows returns for opened items. I really need to know if Hulu allows returns because this was an accident.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl127.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 76, "source_record_id": "recheck.jsonl:127"}
{"oral": "Hi, uh, I'm calling because I see a charge I didn't make. My husband was like, what is this DoorDash charge for over a thousand dollars? And, well, I think it is a total scam! The amount is one thousand forty five seventy five... one thousand forty five seventy five. And the reference number... it is B L L four four nine two zero zero one... oh no, it is BLL four four nine two zero one one. Just, I don't get it, because I already pay for Spotify every month, so why is there a CONF XJ nine two L charge on here, you know?", "clean": "Hi, I'm calling because I see a charge I didn't make. My husband was, what is this DoorDash charge for over $1,000? And I think it is a total scam! The amount is $1,045.75. And the reference number is BLL4492011. I don't get it, because I already pay for Spotify every month, so why is there a CONF-XJ-92L charge on here?", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl138.wav", "scene": "customer_service", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 84, "source_record_id": "recheck.jsonl:138"}
{"oral": "Hi, yeah, I am calling because I see these extra charges on my bill. Look, I returned the car to the Wayfair... oh no, I mean the Walmart location. And the tracking number... it is not updating at all. The number is T R K zero zero nine two eight four seven five six. Again, T-R-K zero zero nine two eight four seven five six. I mean, if you can just verify it was dropped off on time, then we could just remove these fees, right? Because I received a shattered item in my shipment... wait, sorry, I am reading the wrong paper, I mean the car was clean when I left it. It is just so frustrating that the tracking is not updating. I need this fixed now because it is a lot of money.", "clean": "Hi, I am calling because I see these extra charges on my bill. I returned the car to the Walmart location. And the tracking number is not updating at all. The number is TRK009284756. If you can just verify it was dropped off on time, then we could just remove these fees, right? The car was clean when I left it. It is just so frustrating that the tracking is not updating. I need this fixed now because it is a lot of money.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl137.wav", "scene": "customer_service", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 83, "source_record_id": "recheck.jsonl:137"}
{"oral": "Oh my god! I just got my package and it is completely smashed. Look, the tracking number is T R K zero zero nine two eight three seven four six five. I mean, T R K zero zero nine two eight three seven four six five. It is totally ruined. And also, I tried to use a discount but the promo code isn't applying at checkout. I really need help. Also, I think I need a refund for a subscription I didn't authorize, and I need to update shipping address for a pending order. My order number is C N F eight eight three zero two P L. C N F eight eight three zero two P L.", "clean": "Oh my god! I just got my package and it is completely smashed. Look, the tracking number is TRK0092837465. It is totally ruined. And also, I tried to use a discount but the promo code isn't applying at checkout. I really need help. Also, I think I need a refund for a subscription I didn't authorize, and I need to update the shipping address for a pending order. My order number is CNF88302PL.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl140.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 85, "source_record_id": "recheck.jsonl:140"}
{"oral": "Sarah, look, I can't log into my account after the password reset, and just... wait... I mean, the whole thing is just a mess. I need to request a refund for a defective product, it was two hundred dollars... oh no, it was two hundred ten dollars. Like, my promo code isn't working at checkout, and I just want a simple answer, you know? I've been on hold for ten minutes and I'm just... how should I put it... I'm just really frustrated that this is happening again, ya. I just want my money back in my bank account, just... really quickly, ha.", "clean": "Sarah, I can't log into my account after the password reset, and the whole thing is just a mess. I need to request a refund for a defective product, it was $210. My promo code isn't working at checkout, and I just want a simple answer. I've been on hold for 10 minutes and I'm just really frustrated that this is happening again. I just want my money back in my bank account really quickly.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl142.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 86, "source_record_id": "recheck.jsonl:142"}
{"oral": "Hi, yeah, I am calling because my flight was just canceled and I am really disappointed with the service. My email is john dot doe underscore ninety nine at outlook dot com. I mean, if you can find another seat for me today, then we could avoid me spending night in hotel. Oh, wait, my confirmation number is alpha bravo seven... oh no, it is alpha bravo nine. I appreciate your help with this, John. But seriously, like, the item I received is damaged and the box was open... wait, sorry, I am reading the wrong note, I meant my luggage was damaged on the last leg. And also, the promotional code I entered is not applying to my cart, so I just want that sorted out, you know?", "clean": "Hi, I am calling because my flight was just canceled and I am really disappointed with the service. My email is john.doe_99@outlook.com. If you can find another seat for me today, then we could avoid me spending a night in a hotel. My confirmation number is alpha bravo 9. I appreciate your help with this, John. But seriously, the luggage was damaged on the last leg. And also, the promotional code I entered is not applying to my cart, so I just want that sorted out.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl135.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 82, "source_record_id": "recheck.jsonl:135"}
{"oral": "Honestly, I am just panicking right now. I lost my Wells Fargo card and I see a charge for one thousand four hundred fifty dollars and seventy five cents that I did not make! It is order seven seven two one nine four zero. Just... uh... I need this stopped. Could you please help me with this? I mean, if you cannot fix it, I want to cancel my membership immediately. I really do!", "clean": "Honestly, I am just panicking right now. I lost my Wells Fargo card and I see a charge for $1,450.75 that I did not make! It is order 7721940. I need this stopped. Could you please help me with this? If you cannot fix it, I want to cancel my membership immediately. I really do!", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl146.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 89, "source_record_id": "recheck.jsonl:146"}
{"oral": "I was just looking at my statement and I see two charges from AirBnB. One for sixty four twenty and another for seventy four twenty five. I mean I'm extremely disappointed in the service, extremely disappointed in the service because I never even made those bookings. My phone number... my phone number is plus one three ten five five five zero one one five... oh no, it's zero one one two. Anyway, I think my card got stolen. I need you to cancel my card and send me a replacement card.", "clean": "I was just looking at my statement and I see two charges from Airbnb. One for $64.20 and another for $74.25. I'm extremely disappointed in the service because I never even made those bookings. My phone number is +1 310-555-0112. Anyway, I think my card got stolen. I need you to cancel my card and send me a replacement card.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl145.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 88, "source_record_id": "recheck.jsonl:145"}
{"oral": "Hi, look, I'm extremely disappointed with this service. I'm not totally sure but I think... I think my promotional code isn't working. I tried it on Shopify, yeah, but it just won't go through. The order number... order number is Bravo eight eight two seven three one zero. I paid via PayPal, PayPal, and I just... how should I put it, just want this fixed now!", "clean": "Hi, I'm extremely disappointed with this service. I'm not totally sure but I think my promotional code isn't working. I tried it on Shopify, but it just won't go through. The order number is B8827310. I paid via PayPal, and I just want this fixed now!", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl148.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 90, "source_record_id": "recheck.jsonl:148"}
{"oral": "Actually, uh, I just noticed some weird fees on my statement. Like, there is a charge for... how should I put it... some kind of Spotify subscription? I never signed up for that! My booking number was B L L nine nine two eight three seven four... oh no, wait, it is B K L four four one zero nine two. Anyway, just check my email at my email at yahoo dot co dot u k, yeah. And the receipt number, just to be sure, it is S four four nine one zero two eight three seven. I mean, I just want this fixed, okay? It is just... so annoying, right?", "clean": "I just noticed some weird fees on my statement. There is a charge for some Spotify subscription? I never signed up for that! My booking number was BKL441092. Anyway, just check my email at yahoo.co.uk. And the receipt number, just to be sure, it is S449102837. I just want this fixed, okay? It is just so annoying, right?", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl150.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 91, "source_record_id": "recheck.jsonl:150"}
{"oral": "Hi uh I am calling about my bill. The billing cycle is inconsistent and... it is like... actually my billing is all messed up. I cannot access my cloud storage either. My number is plus one seven zero two five five five zero one three three. Wait... no... I mean plus one six one seven five five five zero one seven seven. I keep seeing Steam charges too!", "clean": "Hi, I am calling about my bill. The billing cycle is inconsistent and my billing is all messed up. I cannot access my cloud storage either. My number is +1 617 555 0177. I keep seeing Steam charges too!", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl161.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 95, "source_record_id": "recheck.jsonl:161"}
{"oral": "Hi Sarah, I am calling because I want to request a full refund for my last purchase. Look, I just saw some extra charges and it is not right. I want a full refund, seriously. The tracking number is T R K seven seven three two one zero nine eight five. That is T R K seven seven three two one zero nine eight five. I also need to update shipping address for open order, it is for Xfinity, order number S H I P eight eight two nine one zero zero four. Again, S H I P eight eight two nine one zero zero four.", "clean": "Hi Sarah, I am calling because I want to request a full refund for my last purchase. Look, I just saw some extra charges and it is not right. I want a full refund, seriously. The tracking number is TRK773210985. I also need to update the shipping address for the open order, it is for Xfinity, order number SHIP88291004.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl151.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 92, "source_record_id": "recheck.jsonl:151"}
{"oral": "I need a refund for these extra charges, I need a refund! I stayed at Hilton... the hotel, but my invoice... invoice I N V two zero two three zero zero two shows I received a used item as new. I mean, received a different item than ordered. I tried to fix it online but... unable to update credit card details, just unable to update. The charges... they are wrong. I mean, the bill is just... completely wrong. I cannot... I cannot get the system to let me change my card, and this bill... it is just not right!", "clean": "I need a refund for these extra charges. I stayed at Hilton, but my invoice INV2023002 shows I received a different item than ordered. I tried to fix it online but was unable to update credit card details. The bill is completely wrong, and I cannot get the system to let me change my card.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl157.wav", "scene": "customer_service", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 93, "source_record_id": "recheck.jsonl:157"}
{"oral": "Can I return it if it opened? I mean... I think I can but... wait, let me check. I want to change delivery address, change delivery address after shipping. My old address was one hundred one Maple Lane, Star City, Washington nine eight zero zero one. I... I need it sent to... let's say... two hundred two Birch Boulevard, Denver, Colorado eight zero zero zero one. My email street address late one two three at gmail dot com, but I forgot my security question answers, forgot my security question answers.", "clean": "Can I return it if it's opened? I think I can, but let me check. I want to change the delivery address after shipping. My old address was 101 Maple Lane, Star City, Washington 98001. I need it sent to 202 Birch Boulevard, Denver, Colorado 80001. My email address is late123@gmail.com, but I forgot my security question answers.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl158.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 94, "source_record_id": "recheck.jsonl:158"}
{"oral": "Hi, I just wanted to follow up on requesting a chargeback for my order. See, the... uh... delivery driver left package in rain and it's all ruined. I sent an email to info at company dot co dot u k, but... wait... I think I sent it to info at company dot com... oh no, it was dot co dot u k! Anyway, could you please look into this? I mean, the... how should I put it... the whole thing is just a mess, and I need my money back.", "clean": "Hi, I just wanted to follow up on requesting a chargeback for my order. The delivery driver left the package in the rain and it's all ruined. I sent an email to info@company.co.uk, but could you please look into this? The whole thing is just a mess, and I need my money back.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl162.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 96, "source_record_id": "recheck.jsonl:162"}
{"oral": "Broken? Yeah, it is broken. Two thousand five hundred dollars! Just... two thousand five hundred. Two fifteen seventy five too. Seat limit reached on my plan. Unable to export data. Case three three nine two eight one seven.", "clean": "Broken? Yeah, it is broken. $2,500! Just $2,500. 21575 too. Seat limit reached on my plan. Unable to export data. Case 3392817.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl167.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 98, "source_record_id": "recheck.jsonl:167"}
{"oral": "Hi uh I am calling because my voucher code not working and also like... I think payment processed twice for one order... wait... oh no it was processed three times... no wait, it was processed twice for order number seven four two one. I mean... if I can't get this sorted... how should I say... if you could just check the system then we could maybe... uh... just do a partial refund? I know it is opened but I appreciate your help with this... because the A P I key not working either.", "clean": "Hi, I am calling because my voucher code is not working and also I think payment was processed twice for order number 7421. If you could just check the system, then we could maybe just do a partial refund? I know it is open, but I appreciate your help with this because the API key is not working either.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl166.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 97, "source_record_id": "recheck.jsonl:166"}
{"oral": "Sarah, look, flight canceled. My ref is R E F one one zero two nine three eight... oh no, R E F one one zero two nine three four. I just cannot access cloud storage!", "clean": "Sarah, look, flight canceled. My ref is REF1102934. I just cannot access cloud storage!", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl168.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 99, "source_record_id": "recheck.jsonl:168"}
{"oral": "Hi, I'm not totally sure but I think I can return this opened Notion kit, I paid fifty dollars for it, or wait, it was UberEats... no, it was NordVPN, and I just wanted to check policy since my number is five five five, eight eight eight, one two three four. Is that possible?", "clean": "Hi, I'm not totally sure but I think I can return this opened Notion kit, I paid $50 for it, no, it was NordVPN, and I just wanted to check policy since my number is 555-888-1234. Is that possible?", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl171.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 102, "source_record_id": "recheck.jsonl:171"}
{"oral": "Oh my god! Eh, my monthly statement monthly statement is missing for October, for claim A C C six six seven seven eight eight three, and the DHL charge, the DHL charge is eighteen dollars, just... how to say... TXN four four one zero two nine three one is the reference, it's not showing up in my account ya, I just want to know why it's not there ba.", "clean": "Oh my god! My monthly statement is missing for October, for claim ACC6677883, and the DHL charge is $18. TXN44102931 is the reference; it's not showing up in my account. I just want to know why it's not there.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl169.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 100, "source_record_id": "recheck.jsonl:169"}
{"oral": "Actually... uh, I just noticed my Uber charge went through but then my credit card was declined despite funds being there. It is for order S H I P four four one zero two nine three, which is weird because Microsoft payment worked fine. Just... this is a total waste of time, honestly, you know?", "clean": "I just noticed my Uber charge went through but then my credit card was declined despite funds being there. It is for order SHIP4410293, which is weird because Microsoft payment worked fine. This is a total waste of time, honestly.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl176.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 104, "source_record_id": "recheck.jsonl:176"}
{"oral": "Hi look I just got my package and it is just a mess like the packaging is totally damaged and uh it is the completely incorrect item shipped to me. I paid three hundred ninety nine dollars for this thing and I am just so frustrated. It was sent to five five six six Front Street Liverpool UK L one one A A... oh wait no it is L one one B B. Anyway the transaction number is T X N six six one zero two nine three three. I just need you to fix this right now!", "clean": "Hi, I just got my package and it is just a mess. The packaging is totally damaged and it is the completely incorrect item shipped to me. I paid $399 for this thing and I am just so frustrated. It was sent to 5566 Front Street, Liverpool, UK L11BB. Anyway, the transaction number is TXN66102933. I just need you to fix this right now!", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl170.wav", "scene": "customer_service", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 101, "source_record_id": "recheck.jsonl:170"}
{"oral": "Ugh! I was charged seventy five twenty for requesting feature addition on Deezer. Wait, no, not feature addition, I meant account upgrade! I want close my account permanently. My reference number is T X three three four four five five one. Again, T X three three four four five five one.", "clean": "I was charged $75.20 for requesting an account upgrade on Deezer. I want to close my account permanently. My reference number is TX3344551.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl177.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 105, "source_record_id": "recheck.jsonl:177"}
{"oral": "Hi Sarah, look, I am calling cause I got this weird charge for thirteen ninety nine on my card. I tried using a discount code but it... it expired too early, and then I cannot disable auto renewal on their site. I cannot find invoice in my email either! My address is three twenty one Elm Street, Austin, Texas seven eight zero zero... oh no, seven eight seven zero one. Anyway, I just want this money back, seriously, thirteen ninety nine for something I cannot even access is just... it is ridiculous!", "clean": "Hi Sarah, I am calling cause I got this weird charge for $13.99 on my card. I tried using a discount code but it expired too early, and then I cannot disable auto-renewal on their site. I cannot find the invoice in my email either! My address is 321 Elm Street, Austin, Texas 78701. Anyway, I just want this money back, seriously, $13.99 for something I cannot even access is just ridiculous!", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl174.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 103, "source_record_id": "recheck.jsonl:174"}
{"oral": "I need you to rebook my flight right now because I'm not happy, I'm not happy with the way this was handled! Like, the flight was canceled and I just noticed a discrepancy, a discrepancy in the charges, there is thirteen seventy five added on. The documentation is outdated, documentation is outdated for my profile, and my address... it is six zero six Spruce Street, New York, New York, five one four. I think... how should I say... the system is just not updating. This whole situation is just... just a mess and I really need a new flight today!", "clean": "I need you to rebook my flight right now because I'm not happy with the way this was handled! The flight was canceled and I just noticed a discrepancy in the charges, there is $13.75 added on. The documentation is outdated for my profile, and my address is 606 Spruce Street, New York, New York, 514. I think the system is just not updating. This whole situation is just a mess and I really need a new flight today!", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl179.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 107, "source_record_id": "recheck.jsonl:179"}
{"oral": "Payment gateway timeout. Seventy nine dollars. SESS five five two eight one seven three... oh no, seven five... seven three seven. Fix it!", "clean": "Payment gateway timeout. $79. SESS 552817377. Fix it!", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl181.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 109, "source_record_id": "recheck.jsonl:181"}
{"oral": "Hi uh look I really need to cancel my Canva subscription right now. I mean it just keep charging me and... uh... how should I put it... it's just not working for me. My address is one one seven High Street London UK S W one A one A. Oh wait no... it is Calle Mayor ten Madrid twenty eight thousand one. I mean... no sorry... it is actually one one seven High Street London UK S W one A one A. My ticket is T R K three three four four five five six six. I demand credit for inconvenience... like a full refund. Does that sound fair to you?", "clean": "Hi, I really need to cancel my Canva subscription right now. It just keeps charging me and it's just not working for me. My address is 117 High Street, London, UK SW1A 1A. My ticket is TRK33445566. I demand credit for inconvenience, a full refund. Does that sound fair to you?", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl178.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 106, "source_record_id": "recheck.jsonl:178"}
{"oral": "Credit score affected, affected by mistake. One hundred twelve fifty, one hundred twelve fifty. I want lodge formal complaint.", "clean": "Credit score affected by mistake. 11250. I want to lodge a formal complaint.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl192.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 114, "source_record_id": "recheck.jsonl:192"}
{"oral": "First I filed claim REF two zero two three zero nine two, then I got bill BILL five five four eight two zero... it was American Airlines, American Airlines flight. Finally, order ORD five five two eight one seven three... wait... this order is wrong. Please escalate this, escalate this to a supervisor.", "clean": "1. I filed claim REF2023092\n2. I got bill BILL554820 from American Airlines flight\n3. Order ORD5528173\n\nPlease escalate this to a supervisor.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl187.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 112, "source_record_id": "recheck.jsonl:187"}
{"oral": "My customer ID CUST five five two eight one seven three, I'm unable to activate, unable to activate trial period. Also order ORD seven seven two eight one nine three, and then address eight zero eight Ash Street, Springfield, Massachusetts zero one one zero zero... oh no, zero one one zero one. Plus twelve fifty charge, twelve fifty charge is wrong!", "clean": "My customer ID CUST5528173, I'm unable to activate trial period. Also order ORD7728193, and then address 808 Ash Street, Springfield, Massachusetts 01101. Plus $12.50 charge, $12.50 charge is wrong!", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl182.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 110, "source_record_id": "recheck.jsonl:182"}
{"oral": "Can you help me? Yeah, I can. My Peloton... my Peloton arrived broken. I paid six ninety-nine... oh no, six ninety-five for shipping, and nineteen fifty for the tax. Just... just a mess. My email... wotson at mail dot ru... wait, Squarespace... eh, the order was from Squarespace. My email... email is wotson at mail dot ru.", "clean": "Can you help me? Yeah, I can. My Peloton arrived broken. I paid $6.95 for shipping, and $19.50 for the tax. Just a mess. The order was from Squarespace. My email is wotson@mail.ru.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl186.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 111, "source_record_id": "recheck.jsonl:186"}
{"oral": "Oh my god! I need help right now! I think I lost my credit card and uh there was a fraudulent transaction detected on my card. Like five hundred dollars just vanished. I tried to log in but I... I cannot activate two factor authentication. It just keeps failing. I sent an email to help dot me at zoho dot com and also contact at service dot net but no one replied. And get this... my replacement card... wait, no, the package sent to wrong state! It went to Ohio instead of New York! Can you please just... I mean, can you lock the account and send a new card to the right address? I really need this fixed fast, like, right now!", "clean": "Oh my god! I need help right now! I think I lost my credit card and there was a fraudulent transaction detected on my card. $500 just vanished. I tried to log in but I cannot activate two-factor authentication. It just keeps failing. I sent an email to help.me@zoho.com and also contact@service.net but no one replied. And get this... the package sent to the wrong state! It went to Ohio instead of New York! Can you please lock the account and send a new card to the right address? I really need this fixed fast, right now!", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl180.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 108, "source_record_id": "recheck.jsonl:180"}
{"oral": "Hi I'm calling because I received a generic response, I received a generic response to my ticket about my Qualcomm chip. I bought it from Newegg and the technician didn't show up, the technician didn't show up. I tried emailing samantha dot lee at zoho dot com but I... I mean I just need to reschedule. Please check samantha dot lee at zoho dot com.", "clean": "Hi, I'm calling because I received a generic response to my ticket about my Qualcomm chip. I bought it from Newegg and the technician didn't show up. I tried emailing samantha.lee@zoho.com, but I just need to reschedule. Please check samantha.lee@zoho.com.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl195.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 115, "source_record_id": "recheck.jsonl:195"}
{"oral": "Actually... I... I just remembered, my invoice... I mean, the bill... INV one one two eight three seven four is wrong. I requested a callback but never got one, and... how should I say... I expect a better level of service. My case is CAS eight eight two seven three six. Please call me back at plus one six one seven five five five zero one seven seven. I've been waitin for a while now, it's just... not right, yeah.", "clean": "I just remembered, the bill INV1128374 is wrong. I requested a callback but never got one, and I expect a better level of service. My case is CAS882736. Please call me back at +1 617 555 0177. I've been waiting for a while now, it's just not right.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl188.wav", "scene": "customer_service", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 113, "source_record_id": "recheck.jsonl:188"}
{"oral": "Hi yes I am calling because I just noticed a discrepancy in my billing. Now look I thought it was just a one time fee but oh no it is an unexpected monthly price hike! I am talking about policy B L L eight eight one seven two six three. My case number is C A S E five five two eight one seven three three... wait no sorry it is five five two eight one seven three four. Anyway I sent an email to customer help at i cloud dot com but nobody replied. I mean I really need this sorted out now because the cost is just too high. I mean it is just too high!", "clean": "Hi, yes, I am calling because I just noticed a discrepancy in my billing. Now look, I thought it was just a one-time fee but it is an unexpected monthly price hike! I am talking about policy BLL8817263. My case number is CASE55281734. Anyway, I sent an email to customerhelp@icloud.com but nobody replied. I really need this sorted out now because the cost is just too high.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl196.wav", "scene": "customer_service", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 116, "source_record_id": "recheck.jsonl:196"}
{"oral": "天呐今天热死我了，刚才跟林舒雅去万象城买露露乐蒙，感觉整个人都要融化掉，然后下午约了徐子涵去深夜食堂，哦不是，是去喝下午茶，还有那边空调好像坏了，真的太难受了", "clean": "天呐今天热死我了，刚才跟林舒雅去万象城买Lululemon，感觉整个人都要融化掉，然后下午约了徐子涵去喝下午茶，还有那边空调好像坏了，真的太难受了", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl208.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 121, "source_record_id": "recheck.jsonl:208"}
{"oral": "赵雨萌说周末去古北水镇,我觉得去古北水镇太远太远了,咱们还是去蓝色港湾,蓝色港湾逛逛,我想去买个佳洁时牙刷,不对,是买个电动牙刷,然后顺便去华山医院看下医生", "clean": "赵雨萌说周末去古北水镇，我觉得去古北水镇太远了，咱们还是去蓝色港湾逛逛，我想买个电动牙刷，然后顺便去华山医院看下医生", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl209.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 122, "source_record_id": "recheck.jsonl:209"}
{"oral": "This is a nightmare! Order CAS five five six four seven three eight... it's more like a prank than service. My address is twenty two Baker Street... oh no, I mean one one two two Side Street, Bristol, UK BS one one AA. And this account verification link expired... it's just too much, i can't get it to work.", "clean": "This is a nightmare! Order CAS5564738... it's more like a prank than service. My address is 1122 Side Street, Bristol, UK BS11AA. And this account verification link expired... it's just too much, I can't get it to work.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl200.wav", "scene": "customer_service", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 118, "source_record_id": "recheck.jsonl:200"}
{"oral": "Hi, I thought my appointment was for Tuesday... oh no, it's Wednesday. I need to reschedule because the software is incompatible with my OS, yeah. I've been waiting, waiting for a callback for three days now. My order number is ORD two two eight three seven four six, ORD two two eight three seven four six. The session is SESS one one zero two nine three. I'm still waiting for a reference, a reference... REF nine nine two eight three seven four, and I hope I can get a new time slot soon ha.", "clean": "Hi, I thought my appointment was for Wednesday. I need to reschedule because the software is incompatible with my OS. I've been waiting for a callback for 3 days now. My order number is ORD2283746. The session is SESS110293. I'm still waiting for a reference REF9928374, and I hope I can get a new time slot soon.", "usage_text_words": null, "audio": "audio/customer_service__recheckjsonl197.wav", "scene": "customer_service", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 117, "source_record_id": "recheck.jsonl:197"}
{"oral": "哎呀太心烦了，我家狗这个礼拜五就开始拉肚子，怎么说呢，佳洁士和林梦瑶说得用药，但梁思琪跟吴宇轩又说得先禁食，现在完全没头绪呀。", "clean": "太心烦了，我家狗这个礼拜五就开始拉肚子，佳洁士和林梦瑶说得用药，但梁思琪跟吴宇轩又说得先禁食，现在完全没头绪呀。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl214.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 124, "source_record_id": "recheck.jsonl:214"}
{"oral": "诶你听说了没，高子睿前天中午十二点在国贸中心，国贸中心被钱子涵给堵了，然后还有就是他们俩在南山路，南山路吵得挺凶，另外听说钱子涵再打算分手吧。", "clean": "你听说了没，高子睿前天12:00在国贸中心被钱子涵给堵了，然后他们俩在南山路吵得挺凶，另外听说钱子涵在打算分手吧。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl202.wav", "scene": "daily_chat", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 120, "source_record_id": "recheck.jsonl:202"}
{"oral": "诶，如果何天翊把特斯拉开过来接我，那我们就能去吃那家超赞的店！我想想啊，就在三号街...不对，四号街...等等错了，是五号街！之前赵嘉怡办理签证时穿那件lululemon衣服去吃过，说绝了呀。", "clean": "如果何天翊把特斯拉开过来接我，那我们就能去吃那家超赞的店！就在5号街！之前赵嘉怡办理签证时穿那件lululemon衣服去吃过，说绝了呀。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl212.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 123, "source_record_id": "recheck.jsonl:212"}
{"oral": "诶我们就去成都吧，怎么说呢，程逸凡说那边现在挺舒服的。我想想啊，下周二早晨在人民广场集合就行，到时候方圆苑那边杨睿博也一块儿过来哈。", "clean": "我们就去成都吧，程逸凡说那边现在挺舒服的。下周二早晨在人民广场集合就行，到时候方圆苑那边杨睿博也一块儿过来哈。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl221.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 126, "source_record_id": "recheck.jsonl:221"}
{"oral": "诶对了，今天地铁人真多，我都快挤成相片了，话说周子默之前说他在金茂大厦那边，嗯，帮周子涵搞什么智能家居配置，我想想啊，他是不是在花园路那站下车吧，怎么说呢，要是赶上早高峰，估计得在车里呆上三十分钟呀。", "clean": "今天地铁人真多，我都快挤成相片了。周子默之前说他在金茂大厦那边帮周子涵搞智能家居配置，他是不是在花园路那站下车吧，要是赶上早高峰，估计得在车里呆上30分钟呀。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl216.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 125, "source_record_id": "recheck.jsonl:216"}
{"oral": "诶不是你想那样，咱们不去人民广场，其实是去幸福花园那边，就是张峻熙说那边有个考研自习室，不对，是自习咖啡馆，还能奶茶续杯吧。", "clean": "不是你想那样，咱们不去人民广场，其实是去幸福花园那边，张峻熙说那边有个自习咖啡馆，还能奶茶续杯吧。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl226.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 130, "source_record_id": "recheck.jsonl:226"}
{"oral": "顾清然，诶，你想想前天中午十二点我见物业时，就是说那个鸟牌始祖鸟外套被刮到啦，怎么说呢，我跟他们反映过好几次了吧。", "clean": "顾清然，你想想前天12:00我见物业时，始祖鸟外套被刮到了，我跟他们反映过好几次了吧。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl223.wav", "scene": "daily_chat", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 128, "source_record_id": "recheck.jsonl:223"}
{"oral": "诶，如果郑嘉怡能开她那辆比亚迪的话，那我们就能在五道口…我想想啊…直接在五道口集合，然后去金水桥。不对，是去金水苑，这样方便点吧。郭佳怡也再来接我，就省得打车了。", "clean": "如果郑嘉怡能开她那辆比亚迪的话，那我们就能直接在五道口集合，然后去金水苑，这样方便点吧。郭嘉怡再来接我，就省得打车了。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl227.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 131, "source_record_id": "recheck.jsonl:227"}
{"oral": "诶，最近我跟孙嘉怡去国贸商城逛了逛，然后我还预约个除除螨服务，就是，家里那床垫得弄弄吧。另外我也跟苏曼青聊了，我想想啊，打算把家里那个…换成极米投影仪，怎么说呢，感觉看电影更舒服呀。", "clean": "最近我跟孙嘉怡去国贸商城逛了逛，然后我还预约个除螨服务，家里那床垫得弄弄吧。另外我也跟苏曼青聊了，打算把家里那个换成极米投影仪，感觉看电影更舒服呀。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl224.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 129, "source_record_id": "recheck.jsonl:224"}
{"oral": "不是你以为我逛二手市场买的，其实其实是梁思颖送我新手机！这手机快快得离谱，我刚才扫码扫码共享单车，几乎没延迟。我就想现在带它去西湖，去西湖…怎么说呢…去那边边喝奶茶加波霸边拍照，绝了！", "clean": "不是你以为我逛二手市场买的，其实是梁思颖送我新手机！这手机快得离谱，我刚才扫码共享单车，几乎没延迟。我就想现在带它去西湖，去那边边喝奶茶加波霸边拍照，绝了！", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl222.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 127, "source_record_id": "recheck.jsonl:222"}
{"oral": "我今天真累死了，先是开会开到晚上八点，开会开到晚上八点才结束，然后还得去东冬门接姜雨萌，接姜雨萌去第一人民医院看个门诊，最后我还得去蔚来逛逛折扣店，逛逛折扣店买个东西，简直在崩溃边缘", "clean": "我今天真累死了，先是开会开到20:00才结束，然后还得去东门接姜雨萌去第一人民医院看个门诊，最后我还得去蔚来逛逛折扣店买个东西，简直在崩溃边缘", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl229.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 132, "source_record_id": "recheck.jsonl:229"}
{"oral": "诶，如果周芷若能把人民广场金街那个项目搞定，那我们大概能拿三万五千块奖金吧，我想想啊，到时候我直接给韩雨薇买个极米投影仪。", "clean": "如果周芷若能把人民广场金街那个项目搞定，那我们大概能拿35000元奖金吧，到时候我直接给韩雨薇买个极米投影仪。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl234.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 135, "source_record_id": "recheck.jsonl:234"}
{"oral": "你说过那个剧没看过吧?我昨晚十一点才看完,简直绝了,剧情反转反转得太快,你绝对得看,得看这部!我记得咱俩约后天正午在金融街见,啊不对,在万达广场见,看完之后咱俩去人民广场逛逛,我觉得真的可以!", "clean": "你说过那个剧没看过吧？我昨晚23:00才看完，简直绝了，剧情反转得太快，你绝对得看这部！我记得咱俩约后天12:00在万达广场见，看完之后咱俩去人民广场逛逛，我觉得真的可以！", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl231.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 133, "source_record_id": "recheck.jsonl:231"}
{"oral": "刘子洋，绝了，今天地铁上人多到爆炸，我本来想在名创优品随便逛逛，结果人挤人，我记得下周三傍晚，哦不是，下周四傍晚蓝色港湾那边好像有活动，估计得比今天更挤，唐煜恒你到时候千万别凑热闹", "clean": "刘子洋，绝了，今天地铁上人多到爆炸，我本来想在名创优品随便逛逛，结果人挤人。我记得下周四傍晚蓝色港湾那边好像有活动，估计得比今天更挤，唐煜恒你到时候千万别凑热闹", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl232.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 134, "source_record_id": "recheck.jsonl:232"}
{"oral": "诶，你别以为这剧是那种霸总甜宠剧，其实其实是个悬疑剧，剧情反转反转得特别快。大后天咱们在世纪公园见，不对，南山路见，到时候边喝霸王茶姬边跟你聊，我想把赵宇轩这角色分析给你听。", "clean": "你别以为这剧是那种霸总甜宠剧，其实是个悬疑剧，剧情反转得特别快。大后天咱们在南山路见，到时候边喝霸王茶姬边跟你聊，我想把赵宇轩这角色分析给你听。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl239.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 138, "source_record_id": "recheck.jsonl:239"}
{"oral": "诶你听说了没，韩逸飞跟王梓豪在那滨江路吵起来了，我想想啊，应该是除夕年夜饭那天，在国贸中心那家餐厅，不对，是国贸中心对面那家，吵得特别凶呀。", "clean": "你听说了没，韩逸飞跟王梓豪在滨江路吵起来了，应该是除夕年夜饭那天，在国贸中心对面那家餐厅，吵得特别凶呀。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl241.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 139, "source_record_id": "recheck.jsonl:241"}
{"oral": "绝了今天这地铁人多得离谱，我本来打算下周一早晨去西单大悦城，啊不对，是北外滩那边，然后那个…怎么说呢…我想着顺便把衣服除除螨，还有那个陆子衿也得见一面，另外还得赶那个会，真的挤死我了", "clean": "绝了今天这地铁人多得离谱，我本来打算下周一早晨去北外滩那边，我想着顺便把衣服除除螨，还有陆子衿也得见一面，另外还得赶那个会，真的挤死我了", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl238.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 137, "source_record_id": "recheck.jsonl:238"}
{"oral": "今天地铁真是绝了，人多得要死，我挤在龙湖天街站那个口简直没法动，然后我还在旁边听见吴思琪，哦不是，是韩思琪在打电话聊养老金计算，另外那个吴子涵还在那刷短视频，耳朵贴我脸在，吵死我了！", "clean": "今天地铁真是绝了，人多得要死，我挤在龙湖天街站那个口简直没法动，然后我还在旁边听见韩思琪在打电话聊养老金计算，另外那个吴子涵还在那刷短视频，耳朵贴我脸上，吵死我了！", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl242.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 140, "source_record_id": "recheck.jsonl:242"}
{"oral": "诶，如果徐雅婷真在南山金贸中心上班，那我们就得去天坛公苑…不对，天坛公园…等等错了，是大学城找她！就是这事儿要是真的话，得太劲爆了吧。", "clean": "如果徐雅婷真在南山金贸中心上班，那我们就得去大学城找她！这事儿要是真的话，得太劲爆了吧。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl245.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 141, "source_record_id": "recheck.jsonl:245"}
{"oral": "小王，有人想一起去看演唱会吗，票是盲盒抽奖抽到盲盒抽奖抽到，地点在前海湾，啊不对，在金水桥，我想想，我想想，看完能住全季酒店，明天咱们国贸国贸碰头商量下", "clean": "小王，有人想一起去看演唱会吗，票是盲盒抽奖抽到的，地点在金水桥，看完能住全季酒店，明天咱们国贸碰头商量下", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl236.wav", "scene": "daily_chat", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 136, "source_record_id": "recheck.jsonl:236"}
{"oral": "说白了咱们就别纠结了，直接去万象城吃，那个...怎么说呢...要是孙雨薇和徐佳宜也来就更热闹，刚才周子涵说在东门...啊不对，是南门集合，咱们再那边见就行，估计得等个十分钟。", "clean": "说白了咱们就别纠结了，直接去万象城吃，要是孙雨薇和徐佳宜也来就更热闹，刚才周子涵说在南门集合，咱们在那边见就行，估计得等个10分钟。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl247.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 142, "source_record_id": "recheck.jsonl:247"}
{"oral": "去碧桂园吧，那个...今年夏天想去徒步登山，哎对了，潘悠悠说那边黄昏时分风景绝了，得赶紧订房", "clean": "去碧桂园吧，今年夏天想去徒步登山，潘悠悠说那边黄昏时分风景绝了，得赶紧订房", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl253.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 145, "source_record_id": "recheck.jsonl:253"}
{"oral": "我最近搬到华润海岸城那边住，嗯，睡眠质量好多了，现在每天早睡早起，前几天还跟梁娇去哈尔滨圣索菲亚，啊不对，是圣索菲亚教堂逛了逛。", "clean": "我最近搬到华润海岸城那边住，睡眠质量好多了，现在每天早睡早起，前几天还跟梁娇去哈尔滨圣索菲亚教堂逛了逛。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl255.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 146, "source_record_id": "recheck.jsonl:255"}
{"oral": "诶不是你觉得这种日子没法过吧，其实只要养成早起习惯就没那么难受，怎么说呢，别像旁氏那样想太多，我想想啊，哪怕傍晚六点去白云机场看飞机，或者买个Saint Laurent包包犒劳下自己，日子还是得再往前走呀。", "clean": "你觉得这种日子没法过吧，其实只要养成早起习惯就没那么难受，别像庞氏那样想太多，哪怕18:00去白云机场看飞机，或者买个Saint Laurent包包犒劳下自己，日子还是得再往前走呀。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl249.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 143, "source_record_id": "recheck.jsonl:249"}
{"oral": "诶我也不太确定周六能不能去，就是感觉穿衣指南说那边冷，我想想啊，得穿个Supreme外套吧，不然得感冒，而且我刚做完美甲款式还没干，要是像史甜甜那样在西安钟楼逛逛还好，爬山怕是会弄坏呀。", "clean": "我也不太确定周六能不能去，感觉穿衣指南说那边冷，得穿个Supreme外套吧，不然得感冒，而且我刚做完美甲款式还没干，要是像史甜甜那样在西安钟楼逛逛还好，爬山怕是会弄坏呀。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl256.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 147, "source_record_id": "recheck.jsonl:256"}
{"oral": "对了 那个...正午...得去 Zara 退货", "clean": "正午得去 Zara 退货", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl270.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 152, "source_record_id": "recheck.jsonl:270"}
{"oral": "诶我听说是午睡时间快递到货，他买了个SteelSeries…买了那个…过段时间打算给咱俩一人整一个哈", "clean": "我听说是午睡时间快递到货，他买了个SteelSeries，过段时间打算给咱俩一人整一个哈", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl267.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 151, "source_record_id": "recheck.jsonl:267"}
{"oral": "诶如果周六能穿通勤穿搭去万达，啊不对，华润橡树湾集合嘛，那咱们能省不少时间，就不用等余圆圆中泰那边的车哈。", "clean": "如果周六能穿通勤穿搭去华润橡树湾集合，那咱们能省不少时间，就不用等余圆圆中泰那边的车哈。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl265.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 149, "source_record_id": "recheck.jsonl:265"}
{"oral": "诶我想想啊，当时先是周睿非要拉我去超市打折抢衣服，然后咱们居然在打折区翻到件 Patagonia，原价得好几千块，当时只要三百块，不对，是两百块，最后咱们还约好下月十号一起去逛，结果现如今谁也没联系谁，就是挺逗的吧！", "clean": "当时先是周睿非要拉我去超市打折抢衣服，然后咱们居然在打折区翻到件Patagonia，原价得好几千元，当时只要200元，最后咱们还约好下月10号一起去逛，结果现如今谁也没联系谁，挺逗的吧！", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl263.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 148, "source_record_id": "recheck.jsonl:263"}
{"oral": "诶我家狗子拉肚子，我给他喂了东冬特饮，不对，绿城雅苑，等等错了，是Costco买的药！就是药药喂了两次两次，你觉得觉得这样行吧？", "clean": "我家狗子拉肚子，我给他喂了Costco买的药，喂了两次，你觉得这样行吧？", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl272.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 154, "source_record_id": "recheck.jsonl:272"}
{"oral": "顾凡，你快看便利店新品这个太绝了，碧桂园之源啊不对是农夫山泉出的，虽然换季感冒嗓子疼，但酒店预订完得赶紧冲一波！", "clean": "顾凡，你快看便利店新品这个太绝了，农夫山泉出的，虽然换季感冒嗓子疼，但酒店预订完得赶紧冲一波！", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl276.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 158, "source_record_id": "recheck.jsonl:276"}
{"oral": "诶，如果封梓涵在那边…不对，穆芊芊…等等错了，是邓承宇在那边等咱们，那咱们就能在融创之光融创之光吃，就是网络延迟网络延迟太严重太难受吧。", "clean": "如果邓承宇在那边等咱们，那咱们就能在融创之光吃，网络延迟太严重太难受了。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl275.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 157, "source_record_id": "recheck.jsonl:275"}
{"oral": "哎呀太绝了，开会开到晚上九点，我想想啊，本来打算去南京路买买南京路那个稻香村，结果结果…得改天了吧", "clean": "太绝了，开会开到21:00，本来打算去南京路买稻香村，结果得改天了", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl277.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 159, "source_record_id": "recheck.jsonl:277"}
{"oral": "诶，要是靳思琪能早点把绿城之新那个方案定下来，那我们现在估计早就在中山路逛街买Prada吧，怎么说呢，要是能提前两小时结束，我想想啊，我们就能直接快进到国庆长假模式，结果开到晚上九点多才散会，真的太累道了呀。", "clean": "要是靳思琪能早点把绿城之新那个方案定下来，那我们现在估计早就在中山路逛街买Prada吧。要是能提前2小时结束，我们就能直接快进到国庆长假模式，结果开到晚上21:00多才散会，真的太累了呀。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl273.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 155, "source_record_id": "recheck.jsonl:273"}
{"oral": "余卓，今天人民路地铁真绝了，人多得根本挤不上去，我刚才在那想...哎怎么说呢...感觉像在罐头里，本来打算去前海那个...不对，是简远那边开会，结果被挤得差点没把手机弄丢，还得赶紧找物业投诉下，哎呀这通勤真的太心累哈。", "clean": "余卓，今天人民路地铁真绝了，人多得根本挤不上去，我刚才在那感觉像在罐头里，本来打算去简远那边开会，结果被挤得差点没把手机弄丢，还得赶紧找物业投诉下，这通勤真的太心累哈。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl266.wav", "scene": "daily_chat", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 150, "source_record_id": "recheck.jsonl:266"}
{"oral": "诶，不是你想那样，其实这剧讲汰渍洗衣服，我想想啊，就是诸葛圆圆在天环广场...怎么说呢，在天环广场教春运抢票和恋爱技巧吧，挺绝呀！", "clean": "其实这剧讲的是汰渍洗衣服，诸葛圆圆在天环广场教春运抢票和恋爱技巧吧，挺绝的！", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl280.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 160, "source_record_id": "recheck.jsonl:280"}
{"oral": "诶我最近在搞电竞外设，电竞外设这边挺忙，然后之前跟褚芊芊去天河城逛了趟，另外卢依依和苏萌萌她们…她们说要一起约饭，我想想啊，其实就是想把那顿饭…想把聚餐时间定在周六晚上七点吧。", "clean": "我最近在搞电竞外设，电竞外设这边挺忙，之前跟褚芊芊去天河城逛了趟，另外卢依依和苏萌萌她们说要一起约饭，我想把聚餐时间定在周六19:00吧。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl271.wav", "scene": "daily_chat", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 153, "source_record_id": "recheck.jsonl:271"}
{"oral": "哎呀太心疼了，我家狗子突然拉肚子，我打算带它去无锡金碧辉煌广场那边看，啊不对，是绿城明珠那边，就是…我想想啊…那边有个宠物医院比较专业，我想着趁五一长假之前赶紧去，原本打算下周六，哦不是，下周日去，结果现在它状态太差了，得赶紧去就医，再拖下去怕不行的呀。", "clean": "太心疼了，我家狗子突然拉肚子，我打算带它去绿城明珠那边看，那边有个宠物医院比较专业，我想着趁5月1日长假之前赶紧去，原本打算下周日去，结果现在它状态太差了，得赶紧去就医，再拖下去怕不行的呀。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl274.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 156, "source_record_id": "recheck.jsonl:274"}
{"oral": "诶今天开会开到晚上八点多，就是任雨霏她们还得弄教师节贺卡，我想想啊，应该是卢圆圆和裴若芸去准备羽毛球赛吧。", "clean": "今天开会开到20:00多，任雨霏她们还得弄教师节贺卡，应该是卢圆圆和裴若芸去准备羽毛球赛吧。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl284.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 162, "source_record_id": "recheck.jsonl:284"}
{"oral": "翠湖那边人多，然后…那个…阵雨预警又发，还有招商那块堵死，哎", "clean": "翠湖那边人多，阵雨预警又发了，还有招商那块堵死了。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl292.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 167, "source_record_id": "recheck.jsonl:292"}
{"oral": "诶咱们周末去爬山吧，柏岚说最近风景好，风景好呀。我想想啊，晚上八点一刻钱萌跟金铭道，下周二下周二再一起聚，哈。", "clean": "咱们周末去爬山吧，柏岚说最近风景好。我想想，晚上20:15钱萌跟金铭说，下周二再一起聚。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl285.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 163, "source_record_id": "recheck.jsonl:285"}
{"oral": "诶，我刚才弄加湿器开关弄半天，就是本周五下午去绿城之辉财商教育，啊不对，是绿城之辉商场，给咱提供点情绪价值吧哈。", "clean": "我刚才弄加湿器开关弄半天，本周五下午去绿城之辉商场，给咱提供点情绪价值吧。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl288.wav", "scene": "daily_chat", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 165, "source_record_id": "recheck.jsonl:288"}
{"oral": "韩梦琪说长乐路长乐路有家店，不对，诚品书店那边，你觉得呢", "clean": "韩梦琪说诚品书店那边有家店，你觉得呢", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl294.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 168, "source_record_id": "recheck.jsonl:294"}
{"oral": "诶对了，咱们明天早上跑步哈，我想想啊，之前邢雨桐说那条公交线路那个...公交线路好像得绕路，咱们还是走着去吧。话说去年年底葛萌推荐那家便利店零食，便利店零食真的绝了，跑完咱俩去买点，怎么说呢，就是感觉得奖励下自己嘛。", "clean": "咱们明天早上跑步，之前邢雨桐说那条公交线路好像得绕路，咱们还是走着去吧。去年年底葛萌推荐那家便利店零食真的绝了，跑完咱俩去买点，感觉得奖励下自己嘛。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl291.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 166, "source_record_id": "recheck.jsonl:291"}
{"oral": "司徒远在IFC商城，不对，国金，等等错了，是IFC商城发型设计跟顾芊顾芊在一起，严逸严逸知道了，你觉得呢", "clean": "司徒远在IFC商城发型设计跟顾芊在一起，严逸知道了，你觉得呢", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl296.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 169, "source_record_id": "recheck.jsonl:296"}
{"oral": "周铭周末去爬山吧,我想想啊,最近三里屯太古里太挤太挤了,咱们去山里透透气,下山再吃路边摊小吃,吃路边摊小吃,再买杯古茗和水果拼盘", "clean": "周铭周末去爬山吧，最近三里屯太古里太挤了，咱们去山里透透气，下山再吃路边摊小吃，再买杯古茗和水果拼盘", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl300.wav", "scene": "daily_chat", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 171, "source_record_id": "recheck.jsonl:300"}
{"oral": "Oh by the way, I am emailing a realtor about a property viewing... wait, no, texting a friend to confirm a hike by the middle of next week. I would be gratefull if Abigail could join us in the garden yuan.", "clean": "Oh by the way, I am texting a friend to confirm a hike by the middle of next week. I would be grateful if Abigail could join us in the Garden Yuan.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl305.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 174, "source_record_id": "recheck.jsonl:305"}
{"oral": "Uh, for the... wait, the guy at six hundred sixty six Mountain Pass Salt Lake City Utah eight four one zero one. Need... need to email client for deposit. I... I am contacting you about... uh... voice note for lecture. Due by end of day tomorrow.", "clean": "The guy at 666 Mountain Pass, Salt Lake City, Utah 84101. Need to email client for deposit. I am contacting you about the voice note for lecture, due by end of day tomorrow.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl304.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 173, "source_record_id": "recheck.jsonl:304"}
{"oral": "诶我想着与其咱们周末在那保利之辉逛街不如一起去爬山，反正我半年前在九江那边去过一次感觉特别好。我想想啊，咱们那个…要是能早点出发，怎么说呢，比每月第一天去挤人山人海要强多了。对了我得先去趟邮储银行取点钱，等我办完到家就给你发消息哈，咱们到时候在那个山苑门口见。", "clean": "我想着与其咱们周末在保利之辉逛街，不如一起去爬山，反正我半年前在九江那边去过一次感觉特别好。咱们要是能早点出发，比每月1日去挤人山人海要强多了。我得先去趟邮储银行取点钱，等我办完到家就给你发消息，咱们到时候在山苑门口见。", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl286.wav", "scene": "daily_chat", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 164, "source_record_id": "recheck.jsonl:286"}
{"oral": "Need to finish texting a vet texting a vet about pet appointment by next Friday. If Anthony confirms, then we could meet at one hundred North Tryon Street, Charlotte, North Carolina twenty eight thousand two hundred two. Wait, address is address is one two three four Highland Avenue, Los Angeles, California nine thousand one. Hope he can make it to the meeting for the new project.", "clean": "Need to finish texting a vet about pet appointment by next Friday. If Anthony confirms, then we could meet at 1234 Highland Avenue, Los Angeles, California 9001. Hope he can make it to the meeting for the new project.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl306.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 175, "source_record_id": "recheck.jsonl:306"}
{"oral": "诶今天地铁真的挤爆了，陆峰说月底之前要搞定动漫展那个事儿，结果他早起习惯太差，得十点，啊不对，是十点半才起，我想想，反正他这种…就是说…得赶紧跟任悠悠对一下进度，你觉得呢？", "clean": "今天地铁真的挤爆了，陆峰说月底之前要搞定动漫展那个事儿，结果他早起习惯太差，10:30才起，他得赶紧跟任悠悠对一下进度，你觉得呢？", "usage_text_words": null, "audio": "audio/daily_chat__recheckjsonl298.wav", "scene": "daily_chat", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 170, "source_record_id": "recheck.jsonl:298"}
{"oral": "Actually, uh, the address is eight zero eight Synergy Lane, Atlanta, Georgia three zero three zero eight. I'm looking forward to... let me see... writing a list of packing essentials for skiing, and texting sibling about family gift, just, yeah, by tomorrow noon, ha.", "clean": "The address is 808 Synergy Lane, Atlanta, Georgia 30308. I'm looking forward to writing a list of packing essentials for skiing, and texting sibling about family gift by tomorrow noon.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl312.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 177, "source_record_id": "recheck.jsonl:312"}
{"oral": "So Isabella was telling me she was texting a friend about a gym class and said just wanted to touch base. Now why use that? It sounds like I am writing to thank you for a business meeting. Oh wait, she was sending a text to her partner about grocery items... no, not partner, her roommate! Anyway, it just feels way too formal for a casual chat, right?", "clean": "Isabella was telling me she was texting a friend about a gym class and said just wanted to touch base. Now why use that? It sounds like I am writing to thank you for a business meeting. She was sending a text to her roommate about grocery items. Anyway, it just feels way too formal for a casual chat, right?", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl309.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 176, "source_record_id": "recheck.jsonl:309"}
{"oral": "Need to set follow up on the twelfth of October. I'll text coworker about broken printer and... wait, send email to vendor requesting quote first. Also need note for workout goal. Oh, and tell Sarah I'm sorry to hear that. Does that sound okay to you or should I change something?", "clean": "Need to set follow up on October 12. I'll text coworker about broken printer and send email to vendor requesting quote first. Also need note for workout goal. Oh, and tell Sarah I'm sorry to hear that. Does that sound okay to you or should I change something?", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl313.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 178, "source_record_id": "recheck.jsonl:313"}
{"oral": "Meeting tomorrow. Tomorrow at twelve thirty PM. More like... more like chat than pitch. Address three hundred thirty three Market Street. Market Street, Philadelphia, PA one nine zero seven. Email at sign. Texting teammate. Teammate about game strategy. In about twenty minutes.", "clean": "Meeting tomorrow at 12:30pm. More like a chat than a pitch. Address 333 Market Street, Philadelphia, PA 1907. Email @. Texting teammate about game strategy in about 20 minutes.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl315.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 180, "source_record_id": "recheck.jsonl:315"}
{"oral": "I need reminder to schedule oil change and Zara... Zara needs a voice memo for project brainstorming this coming Wednesday. Also email to vendor about warranty. Wait... email the vendor... the warranty email. Anyway, main thing is just feed the cat twice day.", "clean": "I need a reminder to schedule an oil change, and Zara needs a voice memo for project brainstorming this coming Wednesday. Also, email the vendor about the warranty. Anyway, the main thing is just feed the cat twice a day.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl318.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 181, "source_record_id": "recheck.jsonl:318"}
{"oral": "Oh! Right, need to send text to sister about flight details and remember to pay electric bill. Also, insert link for that medical portal. The address for the specialist is five hundred fifty five Gold Mine Road, Reno, Nevada eight nine five zero one. I have to be there in five hours, wait, no, actually in five hours from now.", "clean": "Right, need to send text to sister about flight details and remember to pay electric bill. Also, insert link for that medical portal. The address for the specialist is 555 Gold Mine Road, Reno, Nevada 89501. I have to be there in 5 hours from now.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl314.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 179, "source_record_id": "recheck.jsonl:314"}
{"oral": "Honestly, send to hiring manager, send to hiring manager, just... a email to bank regarding dispute, for the dispute, and maybe... texting friend apologize late. Oh, drafting newsletter for subscribers, for subscribers... writing note for grocery store budget, two hundred dollars budget, in the budget, write it in the budjet... wait, just write it in the budjet.", "clean": "Honestly, send an email to the hiring manager, an email to the bank regarding the dispute, and maybe text a friend to apologize for being late. Also, draft a newsletter for subscribers and write a note for the grocery store budget, a $200 budget.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl321.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 184, "source_record_id": "recheck.jsonl:321"}
{"oral": "I need to finish my to do list. Just wanted to touch base with Grayson... Grayson about the refund email... oh no, email to customer about refund. Put an asterisk asterisk next to it. Put that note for a book reading list in the new... new folder... wait, not new folder, current folder. Send it by five o clock today.", "clean": "I need to finish my to do list:\n1. Touch base with Grayson about the email to customer about refund. Put ** next to it.\n2. Put that note for a book reading list in the current folder.\n3. Send it by 5:00 today.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl322.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 185, "source_record_id": "recheck.jsonl:322"}
{"oral": "Um, so for the blog, I want to start with what Sarah told me, she was like, we need to meet next Saturday at four PM at four zero four Elm Street, Boston, Massachusetts zero two one zero eight. Now, thinking about that, it feels way too rushed. I mean, I'll probably be drafting a formal apology to a client in about twenty minutes, so my head is already spinning. I should probably push our brainstorm to next Wednesday morning instead. Actually, let's just say next Wednesday... yeah, morning works better. I just need some space to breathe before we dive into the strategy.", "clean": "For the blog, I want to start with what Sarah told me: we need to meet next Saturday at 4pm at 404 Elm Street, Boston, Massachusetts 02108. Now, thinking about that, it feels way too rushed. I'll probably be drafting a formal apology to a client in about 20 minutes, so my head is already spinning. I should probably push our brainstorm to next Wednesday morning instead. Let's just say next Wednesday morning works better. I just need some space to breathe before we dive into the strategy.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl320.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 183, "source_record_id": "recheck.jsonl:320"}
{"oral": "First, three hundred thirty three Rose Garden. Pasadena, California nine one one zero one. Then, tomorrow at ten AM... oh no, nine AM. Finally, eight PM this Friday. Just heads up. Drafting request for refund.", "clean": "1. 333 Rose Garden, Pasadena, California 91101\n2. Tomorrow at 9am\n3. 8pm this Friday\n\nJust heads up. Drafting request for refund.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl327.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 189, "source_record_id": "recheck.jsonl:327"}
{"oral": "Okay, so for the blog post, if I can just... let me think... if I can finish the first three sections, then we could probably insert link to the portfolio by the end of the week. I just need to stop texting a coach about practice and actually focus. Wait, no, I should first finish drafting apology to neighbor for loud music because that is way more urgent. Oh, and I need to check in with Mila about the... the layout... actually, maybe just ask her to review the draft on Friday. I'll try to get it all done by Friday night, though I'm not sure if that is possible.", "clean": "For the blog post, if I can finish the first 3 sections, then we could probably insert a link to the portfolio by the end of the week. I just need to stop texting a coach about practice and actually focus. I should first finish drafting an apology to my neighbor for loud music because that is way more urgent. Oh, and I need to check in with Mila and maybe just ask her to review the draft on Friday. I'll try to get it all done by Friday night, though I'm not sure if that is possible.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl319.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 182, "source_record_id": "recheck.jsonl:319"}
{"oral": "First, note the Phoenix office at five zero five Ash Street, Phoenix, Arizona eight five zero zero one. Then, uh, we need to... how should I put this... update the Napa site to five five five Vineyard Road, Napa, California nine four five five eight. Oh no, wait, eight four five five eight. Finally, emailing vendor about shipment delay for eight zero eight Liberty Avenue, Philadelphia, Pennsylvania one nine one zero one by next payday.", "clean": "1. Note the Phoenix office at 505 Ash Street, Phoenix, Arizona 85001\n2. Update the Napa site to 555 Vineyard Road, Napa, California 84558\n3. Emailing vendor about shipment delay for 808 Liberty Avenue, Philadelphia, Pennsylvania 19101 by next payday", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl323.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 186, "source_record_id": "recheck.jsonl:323"}
{"oral": "Gift for Jackson. Maybe something for drafting a formal letter of intent... drafting a formal letter of intent. Address is seven eighty... no, seven ninety... wait wrong, seven eighty nine Ocean Drive Miami Florida three three one three nine. He was texting a teammate... texting a teammate about the play this coming... this coming Monday... no, Tuesday evening.", "clean": "Gift for Jackson. Maybe something for drafting a formal letter of intent. Address is 789 Ocean Drive, Miami, Florida 33139. He was texting a teammate about the play this coming Tuesday evening.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl325.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 188, "source_record_id": "recheck.jsonl:325"}
{"oral": "Honestly, I need to text friend about baby shower, text friend about baby shower, and maybe do it at dawn. I've been waking up every Monday morning, every Monday morning, to write. Let's see, Charles is at one Infinite Loop, Cupertino, California nine five zero one four.", "clean": "Honestly, I need to text friend about baby shower, and maybe do it at dawn. I've been waking up every Monday morning to write. Let's see, Charles is at 1 Infinite Loop, Cupertino, California 95014.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl330.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 191, "source_record_id": "recheck.jsonl:330"}
{"oral": "Need to send voice memo for project timeline, and uh, email recruiter after interview. Also, emailing bookstore about pre order. I should... actually, just note down four zero four zero Ocean Boulevard, Long Beach, California nine zero eight zero two, at midnight on New Year's.", "clean": "Need to send voice memo for project timeline, and email recruiter after interview. Also, emailing bookstore about pre order. I should just note down 4040 Ocean Boulevard, Long Beach, California 90802, at midnight on New Year's.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl329.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 190, "source_record_id": "recheck.jsonl:329"}
{"oral": "Actually, writing a list... writing a list for Tokyo. Hey, in thirty... no, half an hour, touch base touch base. Note parking spot garage.", "clean": "Writing a list for Tokyo. In half an hour, touch base. Note parking spot garage.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl333.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 194, "source_record_id": "recheck.jsonl:333"}
{"oral": "Hmm, today was just a blur. I spent way too long texting a friend to ask if they're awake, just to see if we could grab coffee. Then I had to go to one one one one Pennsylvania Avenue, Washington, DC two zero zero zero four... wait, no, it was one one zero one Pennsylvania Avenue, Washington, DC two zero zero zero four. I was there to meet Jack. I...I told him I'm following up on my application, but he was busy. If Luca... if Luca comes back tomorrow, then we could maybe try to reschedule. I mean, I think think the timing was just off, ya know? My head is just... just spinning from all the travel. I really need to call Luca... call Luca again tomorrow morning.", "clean": "Today was just a blur. I spent way too long texting a friend to ask if they're awake, just to see if we could grab coffee. Then I had to go to 1101 Pennsylvania Avenue, Washington, D.C. 20004. I was there to meet Jack. I told him I'm following up on my application, but he was busy. If Luca comes back tomorrow, then we could maybe try to reschedule. I think the timing was just off. My head is just spinning from all the travel. I really need to call Luca again tomorrow morning.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl324.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 187, "source_record_id": "recheck.jsonl:324"}
{"oral": "Sarah, hope you're having great week. Lunch... no, email vendor bulk pricing.", "clean": "Sarah, hope you're having a great week. Email vendor bulk pricing.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl338.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 198, "source_record_id": "recheck.jsonl:338"}
{"oral": "What else do I need to pack? Oh right, just... uh... gotta remember to send that email to teacher about sick child before leaving for four hundred fifty six Corporate Plaza, Dallas, Texas seven five two zero one. And, let me see, I need to be up by six A M to handle texting coach about practice time. Then I gotta mail the package to six hundred six Birch Boulevard, Boston, Massachusetts zero two one zero eight. I think I got everything now, right?", "clean": "What else do I need to pack? Oh right, just gotta remember to send that email to teacher about sick child before leaving for 456 Corporate Plaza, Dallas, Texas 75201. And, let me see, I need to be up by 6am to handle texting coach about practice time. Then I gotta mail the package to 606 Birch Boulevard, Boston, Massachusetts 02108. I think I got everything now, right?", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl331.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 192, "source_record_id": "recheck.jsonl:331"}
{"oral": "Who is it? Oh, Aditi. Thanks for the update update. I am emailing emailing vendor about shipment delay. Next Wednesday at ten... no, eleven. Address nine zero nine Fusion Drive, Chicago, Illinois six zero six zero five. Sent it to her e mail.", "clean": "Who is it? Oh, Aditi. Thanks for the update. I am emailing the vendor about the shipment delay. Next Wednesday at 11:00. Address 909 Fusion Drive, Chicago, Illinois 60605. Sent it to her email.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl334.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 195, "source_record_id": "recheck.jsonl:334"}
{"oral": "Why call Jayden? Uh, in reference to John. No, texting client to confirm meeting. Note for luggage list. Need two bags. Wait, three bags. Hope it's a new star. Just go to the garden. Wait, garden yard.", "clean": "Why call Jayden in reference to John? Texting client to confirm meeting. Note for luggage list: need 3 bags. Hope it's a new star. Just go to the garden yard.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl337.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 197, "source_record_id": "recheck.jsonl:337"}
{"oral": "Why should we hire Omar? Well, he is great at... uh... managing a movie watch list, oh no, drafting a LinkedIn outreach message. He can... let me see... finish a voice note for daily reflection in six hours... no, in four hours. He has a real knack for these tasks and is a great addition to the team. Hope this helps him get the job in the new city.", "clean": "Why should we hire Omar? Well, he is great at drafting a LinkedIn outreach message. He can finish a voice note for daily reflection in 4 hours. He has a real knack for these tasks and is a great addition to the team. Hope this helps him get the job in the new city.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl332.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 193, "source_record_id": "recheck.jsonl:332"}
{"oral": "Actually, I am drafting a welcome email, drafting a welcome email for new hire Yara. Tell Daniel meeting is tomorrow at ten AM... no, nine AM. Address is five zero five Bayou Boulevard, Baton Rouge, Louisiana seven zero eight zero one. Seven zero eight zero one.", "clean": "I am drafting a welcome email for new hire Yara. Tell Daniel the meeting is tomorrow at 9am. Address is 505 Bayou Boulevard, Baton Rouge, Louisiana 70801.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl335.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 196, "source_record_id": "recheck.jsonl:335"}
{"oral": "Honestly, I am drafting a press release for a product, uh, wait, make that email to client about project scope. I would like to invite you to review it. Close quote. Oh, and reminder to buy stamps.", "clean": "Honestly, I am drafting an email to client about project scope. I would like to invite you to review it. Close quote. Oh, and reminder to buy stamps.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl340.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 199, "source_record_id": "recheck.jsonl:340"}
{"oral": "Address five hundred South Main Street. Oh no, North Main Street. Salt Lake City Utah eight four one zero one. Like texting friend for recommendation. Not meeting minutes note. This Sunday eleven thirty AM. Thank you for help.", "clean": "Address 500 North Main Street, Salt Lake City, Utah 84101. Texting a friend for a recommendation, not meeting minutes notes. This Sunday 11:30am. Thank you for the help.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl341.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 200, "source_record_id": "recheck.jsonl:341"}
{"oral": "Perfect! Let us follow up with Thierry and Xavier. We will set biweekly meetings biweekly meetings on Monday mornings. I am taking notes taking notes after doctor appointment. Oh wait. Also need to cc Aarav on this email.", "clean": "Perfect! Let us follow up with Thierry and Xavier. We will set biweekly meetings on Monday mornings. I am taking notes after a doctor appointment. Also need to cc Aarav on this email.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl352.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 206, "source_record_id": "recheck.jsonl:352"}
{"oral": "Honestly, I need to finish dictating project status update for manager first thing tomorrow. I'm reaching out to see if Yael... Yael has the latest figures. Send it to the marketing department... oh no, sales department. I should check the budget... check the budget for phase two. Put exclamation point after the deadline reminder. Just make sure Yael confirms by ten a m.", "clean": "Honestly, I need to finish dictating the project status update for the manager first thing tomorrow. I'm reaching out to see if Yael has the latest figures. Send it to the sales department. I should check the budget for phase 2. Put an exclamation point after the deadline reminder. Just make sure Yael confirms by 10am.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl351.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 205, "source_record_id": "recheck.jsonl:351"}
{"oral": "Wait, why did I wait so long to start this? I mean, I should probably just start by recording voice memo for a dream description, then maybe transition into drafting a memo on office policy... wait, no, let me do the note for a book summary first. Oh, Layla mentioned something about a meeting on May fifth, so maybe that goes here? Just... how should I put it, I need to organize these thoughts before I forget everything. It is basically just a giant brain dump, I guess.", "clean": "Why did I wait so long to start this? I should probably just start by recording a voice memo for a dream description, then maybe transition into drafting a memo on office policy... no, let me do the note for a book summary first. Layla mentioned something about a meeting on May 5, so maybe that goes here? I need to organize these thoughts before I forget everything. It is just a giant brain dump, I guess.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl343.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 202, "source_record_id": "recheck.jsonl:343"}
{"oral": "Actually... I need to email emailing a landlord about broken heater at nine zero nine Juniper Lane Nashville Tennessee three seven two zero one. Also need... need to send email to team lead for approval on twelfth of April for four three two one North Shore Drive Chicago Illinois six zero six zero one. Oh wait... I meant send to lead for approval on the twelfth of April for four three two one North Shore Drive Chicago Illinois six zero six zero one. Make sure to CC the lead for approval.", "clean": "I need to email a landlord about a broken heater at 909 Juniper Lane, Nashville, Tennessee 37201. Also need to send an email to the team lead for approval on April 12 for 4321 North Shore Drive, Chicago, Illinois 60601. Make sure to CC the lead for approval.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl342.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 201, "source_record_id": "recheck.jsonl:342"}
{"oral": "Memo to self. I need to write this recommendation for Yuki. She worked under me for three years... three years in the marketing department. She was actually the one who handled that big project... wait, not the big project, the mid sized campaign for the European market. She handled the campaign with such precision. I remember when she managed the budget of twelve thousand dollars... oh no, it was fifteen thousand dollars. She managed the budget perfectly. Her ability to lead... how to put it... if she takes charge of a team, the results are always consistent. Consistent results every time. She is definitely the best candidate for this senior role.", "clean": "Memo to self. I need to write this recommendation for Yuki. She worked under me for 3 years in the marketing department. She was the one who handled the mid-sized campaign for the European market. She handled the campaign with such precision. I remember when she managed the budget of $15,000. She managed the budget perfectly. Her ability to lead... if she takes charge of a team, the results are always consistent. Consistent results every time. She is definitely the best candidate for this senior role.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl350.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 204, "source_record_id": "recheck.jsonl:350"}
{"oral": "Voice journal entry. October twenty fourth. So I finally heard back from Aadhya. She sent this weirdly formal message, like, I hope this email finds you well, which is just so not her. She said she is following up on our previous conversation regarding the new gallery space at eighty eight Orchard Road, Singapore two three eight eight two three. Then she mentioned Yael is handling the lease. Actually, she said Yael is in the marketing department... oh no, she is in the sales department. Anyway, my take is that they are just trying to sound professional to hide the fact that they are totally disorganized. It feels like a corporate mask.", "clean": "Voice journal entry. October 24. I finally heard back from Aadhya. She sent this weirdly formal message, \"I hope this email finds you well,\" which is just so not her. She said she is following up on our previous conversation regarding the new gallery space at 88 Orchard Road, Singapore 238823. Then she mentioned Yael is handling the lease. She said Yael is in the sales department. Anyway, my take is that they are just trying to sound professional to hide the fact that they are totally disorganized. It feels like a corporate mask.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl346.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 203, "source_record_id": "recheck.jsonl:346"}
{"oral": "Hey, wonder if you got a second to check this? Actually, I'm just summarizing a quarterly budget meeting for you. Hope you're having a great week. So, I spoke with Aarav and... uh... Guillaume regarding the attached file. We're looking at a total spend of twelve thousand dollars... oh no, it's actually fifteen thousand dollars for the first phase. And we need to... let me see... I mean, we should probably reallocate those funds to the marketing wing by Friday the twenty second. Let me know if that works.", "clean": "Hey, wonder if you got a second to check this? I'm just summarizing a quarterly budget meeting for you. Hope you're having a great week. So, I spoke with Aarav and Guillaume regarding the attached file. We're looking at a total spend of $15,000 for the first phase. And we should probably reallocate those funds to the marketing wing by Friday the 22nd. Let me know if that works.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl353.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 207, "source_record_id": "recheck.jsonl:353"}
{"oral": "Okay, just to be clear, the main thing is keep the plants alive while I am gone. Just because they are really finicky, you know. Following up on our last call, address is Kurfürstendamm twenty one, Berlin ten seven one nine. Just a heads up, I am also writing a reminder list for a child's birthday party, and I need you to mail it to Björn. Oh, wait, not Björn, it is Ben. Anyway, the package is on the table. I wrote the address with a semicolon... no, I mean a comma, just to separate the city. Just leave the keys under the mat, yeah.", "clean": "Okay, just to be clear, the main thing is keep the plants alive while I am gone. Just because they are really finicky. Following up on our last call, address is Kurfürstendamm 21, Berlin 10719. Just a heads up, I am also writing a reminder list for a child's birthday party, and I need you to mail it to Ben. Anyway, the package is on the table. I wrote the address with a comma, just to separate the city. Just leave the keys under the mat, yeah.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl354.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 208, "source_record_id": "recheck.jsonl:354"}
{"oral": "So how did today actually go? I mean, pretty hectic. Started off at five hundred Terry Francana Plaza, San Jose, nine five one one three. Then drove over to four hundred fifty Serra Mall, Stanford, California. Uh, it was for that meeting... wait, no, it was a site visit. Anyway, just a heads up that I still need to send that email to the team. Apologies for the late response, just... how should I put it, just got bogged down in traffic. I gotta be up at the crack of dawn tomorrow, so I should probably crash now, ya.", "clean": "So how did today actually go? Pretty hectic. Started off at 500 Terry Francana Plaza, San Jose, 95113. Then drove over to 450 Serra Mall, Stanford, California. It was for that site visit. Anyway, just a heads up that I still need to send that email to the team. Apologies for the late response, just got bogged down in traffic. I gotta be up at the crack of dawn tomorrow, so I should probably crash now, ya.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl359.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 209, "source_record_id": "recheck.jsonl:359"}
{"oral": "Hi team. Um, just following up on our conversation with Siobhan. I want to... uh... let's see... I want to mention that I'm not totally sure but I think we should pivot the strategy. Wait, actually... maybe we can just add a section about... uh... recording a sudden business idea for a startup in the appendix. Also, I'm still dictating a complaint email to a shipping company for that missing package from January twelfth, so I'll send that separately. Just make sure the total is five hundred twenty dollars.", "clean": "Hi team. Just following up on our conversation with Siobhan. I want to mention that I'm not totally sure but I think we should pivot the strategy. Maybe we can just add a section about recording a sudden business idea for a startup in the appendix. Also, I'm still dictating a complaint email to a shipping company for that missing package from January 12, so I'll send that separately. Just make sure the total is $520.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl364.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 211, "source_record_id": "recheck.jsonl:364"}
{"oral": "Actually... oh, right. Need to email Aoife. Tell her apologies for late response. I was... uh... taking notes during virtual team stand up... no, wait, it was summarizing quarterly budget meeting. Anyway, remind her about that semicolon issue in the report. Just need to send that by five o clock.", "clean": "Need to email Aoife. Tell her apologies for the late response. I was summarizing the quarterly budget meeting. Anyway, remind her about that semicolon issue in the report. Just need to send that by 5:00.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl368.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 213, "source_record_id": "recheck.jsonl:368"}
{"oral": "Hey Sarah, following up on our previous conversation about dinner. I'm creating shopping list for specific recipe, uh, just capturing quick reminder for medication dosage too so I don't forget. I need two tablespoons of honey, two tablespoons of honey and three hundred grams of flour. Wait, I'm writing to follow up on the spice list, it should be four teaspoons of salt. I'm creating detailed to do list for home renovation project later but first I need to get to the store. Actually the store is... the store is closed on Sundays, right? I wonder if I can go to the new park yuan instead.", "clean": "Hey Sarah, following up on our previous conversation about dinner. I'm creating a shopping list for a specific recipe, just capturing a quick reminder for medication dosage too so I don't forget. I need 2 tablespoons of honey and 300 grams of flour. I'm writing to follow up on the spice list, it should be 4 teaspoons of salt. I'm creating a detailed to-do list for a home renovation project later but first I need to get to the store. Actually the store is closed on Sundays, right? I wonder if I can go to the new Park Yuan instead.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl366.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 212, "source_record_id": "recheck.jsonl:366"}
{"oral": "Wait, taking quick note of parking spot location... it was near the south gate... no, north gate... wait wrong, it was east gate! Okay, for the blog post, if we focus on productivity tools then we could highlight Yuki and Aoife. They had that great debate. I remember Aadhya... Aadhya mentioned... how to put this... if we analyze taking voice memo during university lecture, we can see the contrast. The lecture was on... March tenth... no, March twelfth... wait wrong, it was March fifteenth! We need to emphasize emphasize the efficiency of async notes. Let me check check the recording again.", "clean": "Taking quick note of parking spot location... it was east gate! Okay, for the blog post, if we focus on productivity tools then we could highlight Yuki and Aoife. They had that great debate. I remember Aadhya mentioned if we analyze taking voice memos during university lecture, we can see the contrast. The lecture was on March 15! We need to emphasize the efficiency of async notes. Let me check the recording again.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl360.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 210, "source_record_id": "recheck.jsonl:360"}
{"oral": "Does this version sound professional enough? Yeah I think so. Um, following up on our conversation, I am reviewing the part about drafting a text to a landlord about a leaking faucet. The phrase I'm writing to inform you that... maybe too formal for a text message. Oh and sorry for the late response, that part is good. The address is Kurfürstendamm two hundred twelve... oh no, it is two hundred thirteen, Berlin. Let's just keep that as is.", "clean": "Does this version sound professional enough? Yeah I think so. Following up on our conversation, I am reviewing the part about drafting a text to a landlord about a leaking faucet. The phrase \"I'm writing to inform you that...\" maybe too formal for a text message. Oh and sorry for the late response, that part is good. The address is Kurfürstendamm 213, Berlin. Let's just keep that as is.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl381.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 219, "source_record_id": "recheck.jsonl:381"}
{"oral": "Okay let's start with the intro. I want to talk about those weirdly specific habits we have when multitasking. Like recording a recipe, recording a recipe while cooking. Or maybe drafting a Slack message to a remote teammate. I'm not totally sure but I think most people struggle with focus. Oh and I need to remember the address for that Tokyo office. It's two one one Nihonbashi, Chuo ku, Tokyo. Wait, actually I think I have to finish the draft by close of business tomorrow. Close of business tomorrow. Then there's that funny example from the movie, forty two Wallaby Way, Sydney. Forty two Wallaby Way, Sydney. I'll add that in later.", "clean": "Okay let's start with the intro. I want to talk about those weirdly specific habits we have when multitasking. Recording a recipe while cooking. Or maybe drafting a Slack message to a remote teammate. I'm not totally sure but I think most people struggle with focus. Oh and I need to remember the address for that Tokyo office. It's 211 Nihonbashi, Chuo-ku, Tokyo. I think I have to finish the draft by close of business tomorrow. Then there's that funny example from the movie, 42 Wallaby Way, Sydney. I'll add that in later.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl372.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 215, "source_record_id": "recheck.jsonl:372"}
{"oral": "Right, leave a note for Guillaume. Mention the utility bill. He told me, I quote, this is a total rip off, end quote. Which, yeah, I agree. So, I'm dictating a complaint letter to utility company now. Start it... Dear Mr. Henderson, I'm writing to request a full audit of my account. Could you please clarify if the surcharge is correct? Wait, change that to... clarify if the fifty two dollar fee is actually valid. No, make it fifty two point five zero dollars. Just leave it as fifty two dollars.", "clean": "Right, leave a note for Guillaume. Mention the utility bill. He told me, I quote, \"this is a total rip off,\" end quote. Which I agree. So, I'm dictating a complaint letter to the utility company now. Start it... Dear Mr. Henderson, I'm writing to request a full audit of my account. Could you please clarify if the $52 fee is actually valid.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl384.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 221, "source_record_id": "recheck.jsonl:384"}
{"oral": "Got it! Hope all is well. Texting texting sibling about bill. Tomorrow at three... no, four PM. Wanted clarify clarify Wyatt gift.", "clean": "Got it! Hope all is well. Texting sibling about bill. Tomorrow at 4pm. Wanted to clarify Wyatt's gift.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl394.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 227, "source_record_id": "recheck.jsonl:394"}
{"oral": "Need packing list. It's more like... than emailing a customer to resolve a dispute. No, wait. Like voice memo for brainstorming session on branding. Address is two twenty two Music Row, Nashville, Tennessee three seven two zero one. I'm emailing you regarding... no, scratch that. Just pack the laptop.", "clean": "Need packing list. Voice memo for brainstorming session on branding. Address is 222 Music Row, Nashville, Tennessee 37201. Just pack the laptop.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl392.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 225, "source_record_id": "recheck.jsonl:392"}
{"oral": "Voice journal entry for today. I was at Champ de Mars... address is... address is five Avenue Anatole France, Paris, around noon on the fifteenth. I spent most of time creating a daily to do list for home renovation. Wait, I think I should schedule the final walk through for the third Tuesday of next month... oh no, the third Thursday of next month. Then I spent afternoon taking meeting minutes during virtual sync with remote team. Hmm, it feels like I'm just... just reacting to things instead of planning. I wonder if you think you think this level of detail in a to do list is overkill, or is it actually helpful?", "clean": "Voice journal entry for today. I was at Champ de Mars... address is 5 Avenue Anatole France, Paris, around noon on the 15th. I spent most of the time creating a daily to-do list for home renovation. I think I should schedule the final walk through for the third Thursday of next month. Then I spent the afternoon taking meeting minutes during a virtual sync with the remote team. It feels like I'm just reacting to things instead of planning. I wonder if you think this level of detail in a to-do list is overkill, or is it actually helpful?", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl386.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 222, "source_record_id": "recheck.jsonl:386"}
{"oral": "Um... gift ideas... let me see... oh... the thing about emailing a colleague for a peer review... wait... texting a friend to ask if they are awake... no that's not it... I mean... for mom... maybe a planner? Or... oh... that email to customer about refund... uh... next Thursday at three... wait... no... next Wednesday at three... oh no, not Wednesday... oh not... it's next Thursday at three... ah no... it is next Friday at three! Uh... anyway... the main thing... I need to send that email to colleague for a peer review this coming Saturday at nine A M.", "clean": "Gift ideas... the thing about emailing a colleague for a peer review... texting a friend to ask if they are awake... for mom, maybe a planner? Or that email to customer about refund next Friday at 3:00. Anyway, the main thing is I need to send that email to colleague for a peer review this coming Saturday at 9am.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl390.wav", "scene": "dictation_memo", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 224, "source_record_id": "recheck.jsonl:390"}
{"oral": "Oh! Need to text dentist to move appointment. Then... wait... email board members about agenda, no, board of directors. Oh not, board members. I hope this email finds you well... let me see... emailing a teacher about a student. Text to coworker about meeting room five... oh not, room six. Hope they get it in time for the new year. Just make sure you check the mail at the garden court... oh, garden yuan.", "clean": "Need to text dentist to move appointment. Then email board members about agenda. Emailing a teacher about a student. Text to coworker about meeting room 6. Hope they get it in time for the new year. Just make sure you check the mail at the Garden Yuan.", "usage_text_words": null, "audio": "audio/dictation_memo__recheckjsonl398.wav", "scene": "dictation_memo", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 228, "source_record_id": "recheck.jsonl:398"}
{"oral": "Honestly, the numbers are... well, they are mixed. Conversion rate conversion rate is up to two point four percent, and MAU looks solid, but we got this churn spike in mid market tier. I think... I think it's because of... how should I put this... it's basically an OKR misalignment between product and sales. Siobhan O'Neill mentioned she mentioned it in the sync yesterday. We need to fix this misalignment misalignment before the end of quarter.", "clean": "Honestly, the numbers are mixed. Conversion rate is up to 2.4%, and MAU looks solid, but we got this churn spike in mid market tier. I think it's because of an OKR misalignment between product and sales. Siobhan O'Neill mentioned it in the sync yesterday. We need to fix this misalignment before the end of quarter.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl409.wav", "scene": "meeting", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 232, "source_record_id": "recheck.jsonl:409"}
{"oral": "Actually... oh, I just remembered, uh, we need to look at the latest feedback from... uh... Ji won Park in the Sales team... oh wait, not Sales, Marketing team. Anyway, she mentioned that our CAC is like, way higher than expected. And uh, it looks like our burn rate exceeding monthly budget by, uh, maybe twelve percent? So we really need to focus on cloud spend optimization for AWS. But like, the main issue is these... uh... cross functional dependency bottlenecks, right? It is just making everything move so slow.", "clean": "I just remembered we need to look at the latest feedback from Ji-won Park in the Marketing team. Anyway, she mentioned that our CAC is way higher than expected. And it looks like our burn rate is exceeding the monthly budget by maybe 12%? So we really need to focus on cloud spend optimization for AWS. But the main issue is these cross-functional dependency bottlenecks, right? It is just making everything move so slow.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl410.wav", "scene": "meeting", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 233, "source_record_id": "recheck.jsonl:410"}
{"oral": "So uh looking at the data we definitely saw some onboarding friction in the user flow which... well it hit our conversion rate. It went up to four point two percent... oh no sorry, it was three point eight percent. But the good thing is the average order value actually climbed. Now if we look at the ARPU and the monthly recurring revenue... it seems like the high tier is sticking. I mean... does that align with what you guys are seeing on the campaign side?", "clean": "Looking at the data, we definitely saw some onboarding friction in the user flow which hit our conversion rate. It was 3.8%. But the good thing is the average order value actually climbed. Now if we look at the ARPU and the monthly recurring revenue, it seems like the high tier is sticking. Does that align with what you guys are seeing on the campaign side?", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl413.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 236, "source_record_id": "recheck.jsonl:413"}
{"oral": "So uh looking at the data from the test we saw a bump in LTV by about twelve percent... oh wait no it was actually fourteen point five percent. And then we noticed a significant lift in pipeline velocity... I mean... basically the leads are moving through the funnel way faster now. Plus we should see these gains stabilize within the next two quarters... also we need to track how that pipeline velocity affects the long term LTV across different segments.", "clean": "Looking at the data from the test, we saw a bump in LTV by about 14.5%. And then we noticed a significant lift in pipeline velocity; the leads are moving through the funnel way faster now. Plus we should see these gains stabilize within the next 2 quarters... also we need to track how that pipeline velocity affects the long term LTV across different segments.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl423.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 241, "source_record_id": "recheck.jsonl:423"}
{"oral": "Oh! So, uh, we saw user retention dip after version two point one release. I mean, Sasha Ivanova and the team think it's just... how should I put it... user onboarding friction. It's hitting LTV to CAC ratio, and honestly, it might affect Monthly Recurring Revenue, right?", "clean": "Oh! So, we saw user retention dip after version 2.1 release. Sasha Ivanova and the team think it's user onboarding friction. It's hitting LTV to CAC ratio, and honestly, it might affect Monthly Recurring Revenue, right?", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl432.wav", "scene": "meeting", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 245, "source_record_id": "recheck.jsonl:432"}
{"oral": "Look, we got serious latency issues in US East region, latency issues that are killing our MAU. If we can't stabilize this, then we could see... how should I put it... if we don't fix this now we could lose the product market fit for new AI feature. We're seeing churn increase in EMEA market, churn increase that is hitting like three point five percent. It is... it is worrying because our active user growth... our active user growth is slowing down. We need to check the server logs in the eastern region, I mean the US East region, and see if the load balancer is failing. We need to act fast or we lose the traction we built in the new system.", "clean": "Look, we got serious latency issues in US East region, latency issues that are killing our MAU. If we can't stabilize this, then we could lose the product market fit for the new AI feature. We're seeing churn increase in EMEA market, churn increase that is hitting 3.5%. It is worrying because our active user growth is slowing down. We need to check the server logs in the US East region and see if the load balancer is failing. We need to act fast or we lose the traction we built in the new system.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl424.wav", "scene": "meeting", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 242, "source_record_id": "recheck.jsonl:424"}
{"oral": "So, uh, do we actually think this timeline is... I mean, is it realistic given where we are? I don't know, honestly, I'm kind of worried. Like, I was talking to Liam O'Connor and, uh, Alistair Cook, and it feels like we're hitting some major cross functional dependency bottlenecks. Specifically with the P ninety nine latency targets. I think Sanjay Gupta mentioned we're looking at, uh, four hundred milliseconds... oh no, sorry, five hundred milliseconds, which is way over budget. So, yeah, I just... I'm not sure we can hit the date without, you know, shifting some resources.", "clean": "Do we actually think this timeline is realistic given where we are? I don't know, honestly, I'm worried. I was talking to Liam O'Connor and Alistair Cook, and it feels like we're hitting some major cross-functional dependency bottlenecks, specifically with the P99 latency targets. I think Sanjay Gupta mentioned we're looking at 500 milliseconds, which is way over budget. So, I'm not sure we can hit the date without shifting some resources.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl425.wav", "scene": "meeting", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 243, "source_record_id": "recheck.jsonl:425"}
{"oral": "Basically we're sticking with React for frontend, sticking with React because it just handles state better, and we use a PostgreSQL database, oh no, MongoDB for the backend, so we... I mean... if we look at the architecture, it's just more flexible, more flexible for our current scale of about five hundred users.", "clean": "We're sticking with React for frontend because it just handles state better, and we use MongoDB for the backend. If we look at the architecture, it's just more flexible for our current scale of about 500 users.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl440.wav", "scene": "meeting", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 248, "source_record_id": "recheck.jsonl:440"}
{"oral": "So looking at the data, we saw a churn spike among enterprise tier users, a churn spike among enterprise tier users, right around mid March. It looks like we had some latency issues in the US East region, which probably hit the experience. Now, our Customer Acquisition Cost was around twelve dollars... oh no, it was fifteen dollars. If we can stabilize the platform performance, then we could probably get buy in from leadership to scale the spend. But we gotta be careful because the burn rate, the burn rate is currently way higher than we projected for this quarter.", "clean": "So looking at the data, we saw a churn spike among enterprise tier users right around mid-March. It looks like we had some latency issues in the US-East region, which probably hit the experience. Now, our Customer Acquisition Cost was around $15. If we can stabilize the platform performance, then we could probably get buy-in from leadership to scale the spend. But we gotta be careful because the burn rate is currently way higher than we projected for this quarter.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl436.wav", "scene": "meeting", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 246, "source_record_id": "recheck.jsonl:436"}
{"oral": "Honestly, I know the team had some reservations, but if we look at the API uptime requirements, this is the only way to guarantee those nines. I mean, I've already started to socialize the proposal with Arjun Mehta and he's mostly on board, though we still need to tighten up the docs by the next steering committee. It's not like we're trying to invent something as radical as Satoshi Nakamoto did with Bitcoin or whatever, it's just... it's just about scalability. I think, uh, if we pivot now, we'll just be pushing the technical debt further down the road.", "clean": "Honestly, I know the team had some reservations, but if we look at the API uptime requirements, this is the only way to guarantee those nines. I've already started to socialize the proposal with Arjun Mehta and he's mostly on board, though we still need to tighten up the docs by the next steering committee. It's not like we're trying to invent something as radical as Satoshi Nakamoto did with Bitcoin or whatever, it's just about scalability. I think if we pivot now, we'll just be pushing the technical debt further down the road.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl431.wav", "scene": "meeting", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 244, "source_record_id": "recheck.jsonl:431"}
{"oral": "Uh, just to be clear, it is not a database lock. Like, people think it is, but it's actually... uh... the issue is... wait, let me rephrase... we have a memory leak in the API gateway. It hit ninety eight percent usage at ten fifteen AM, and it's crashing the pods.", "clean": "Just to be clear, it is not a database lock. People think it is, but we have a memory leak in the API gateway. It hit 98% usage at 10:15am, and it's crashing the pods.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl442.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 249, "source_record_id": "recheck.jsonl:442"}
{"oral": "Sarah, timeline... timeline looks... no, looks tight. Two weeks... oh no, three weeks, three weeks is not enough.", "clean": "Sarah, timeline looks tight. 3 weeks is not enough.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl444.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 250, "source_record_id": "recheck.jsonl:444"}
{"oral": "I think it's more like... it's more like a quick fix than a long term solution. Like, the logic logic here is a bit messy. It's more like a patch than a real refactor. We should probably look at line eighty four eighty four again.", "clean": "I think it's more like a quick fix than a long term solution. The logic here is a bit messy. It's more like a patch than a real refactor. We should probably look at line 84 again.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl449.wav", "scene": "meeting", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 252, "source_record_id": "recheck.jsonl:449"}
{"oral": "We add add the AI filter... no, AI search... search search for ten users.", "clean": "We add the AI search for 10 users.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl456.wav", "scene": "meeting", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 257, "source_record_id": "recheck.jsonl:456"}
{"oral": "Look, I think we should go with a microservices approach, just... because the latency is like ten milliseconds... oh no, it is twenty five milliseconds, and we really... if we just... I mean, if we decouple the data layer, we can scale it to five thousand concurrent users, right?", "clean": "I think we should go with a microservices approach because the latency is 25 milliseconds, and if we decouple the data layer, we can scale it to 5,000 concurrent users, right?", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl447.wav", "scene": "meeting", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 251, "source_record_id": "recheck.jsonl:447"}
{"oral": "So Sarah mentioned Sarah mentioned we should cut marketing by ten percent... wait, fifteen percent... no sorry, it is twenty percent! Anyway, the quarterly report... oh right, I think we need to double double the budget for RD because it's the only way we hit targets.", "clean": "Sarah mentioned we should cut marketing by 20%! Anyway, the quarterly report... I think we need to double the budget for R&D because it's the only way we hit targets.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl451.wav", "scene": "meeting", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 253, "source_record_id": "recheck.jsonl:451"}
{"oral": "So uh look, we saw that conversion rate drop after v two point four release and I think... I mean... if we look at the API rate limiting thresholds, it's clear we need a change. Now, Amara Okafor suggested we could... uh... wait, let me rephrase that, we should actually scale the limits dynamically. I'm not talking about some crazy Satoshi Nakamoto level decentralization here, just basic load balancing. I want to get this fixed before Friday EOD so it doesn't affect the next sprint. Does that sound reasonable to you or do you think it'll take too long to implement in the current system? I mean we need to make sure the team can actually get it done by the end of the week so we don't have to delay the launch. But yeah, what's your take on the approach?", "clean": "Look, we saw that conversion rate drop after v2.4 release and I think if we look at the API rate limiting thresholds, it's clear we need a change. Now, Amara Okafor suggested we should scale the limits dynamically. I'm not talking about some crazy Satoshi Nakamoto level decentralization here, just basic load balancing. I want to get this fixed before Friday EOD so it doesn't affect the next sprint. Does that sound reasonable to you or do you think it'll take too long to implement in the current system? We need to make sure the team can actually get it done by the end of the week so we don't have to delay the launch. But yeah, what's your take on the approach?", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl437.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 247, "source_record_id": "recheck.jsonl:437"}
{"oral": "So, if we shift, shift about ten percent of the marketing budget, just... just ten percent, then we could, we could potentially scale our reach, reach more leads. I mean, it would really help us get to the next level, how should I say, it's definitely a win-win for the team, right?", "clean": "If we shift about 10% of the marketing budget, then we could potentially scale our reach, reach more leads. It would really help us get to the next level; it's definitely a win win for the team, right?", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl455.wav", "scene": "meeting", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 256, "source_record_id": "recheck.jsonl:455"}
{"oral": "He said two weeks. No, three weeks. Still too short, right?", "clean": "He said 3 weeks. Still too short, right?", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl459.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 260, "source_record_id": "recheck.jsonl:459"}
{"oral": "Change this, change this loop. It's too slow, too slow.", "clean": "Change this loop. It's too slow.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl465.wav", "scene": "meeting", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 262, "source_record_id": "recheck.jsonl:465"}
{"oral": "Uh first we launch in May... oh no, June. Then we we target ten thousand users, just to get the data data in. Finally, I think, it'll drive ten percent growth, right? Actually, it'll be a new star in the market.", "clean": "First we launch in June.\nThen we target 10,000 users, just to get the data in.\nFinally, I think it'll drive 10% growth, right? Actually, it'll be a new star in the market.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl457.wav", "scene": "meeting", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 258, "source_record_id": "recheck.jsonl:457"}
{"oral": "So we saw user drop off at ten percent... no, twelve percent during onboarding. We really need to simplify simplify the flow. I think we should add add a one click sign in to boost conversion.", "clean": "So we saw user drop off at 12% during onboarding. We really need to simplify the flow. I think we should add a one click sign in to boost conversion.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl458.wav", "scene": "meeting", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 259, "source_record_id": "recheck.jsonl:458"}
{"oral": "So, I'm not totally sure but I think, how should I put it, we might have missed the mark on some deliverables. I mean, I feel I feel like the friction between product and engineering was... it was just too high. We had like twelve... oh no, fourteen different blockers in the last sprint. And I think we need to need to rethink how we handle the hand off, you know, maybe set up some clearer guidelines for the next quarter, though I'm not completely certain if that solves the root cause, right?", "clean": "I'm not totally sure but I think we might have missed the mark on some deliverables. I feel the friction between product and engineering was just too high. We had 14 different blockers in the last sprint. And I think we need to rethink how we handle the hand-off, maybe set up some clearer guidelines for the next quarter, though I'm not completely certain if that solves the root cause, right?", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl452.wav", "scene": "meeting", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 254, "source_record_id": "recheck.jsonl:452"}
{"oral": "So, uh, for the backend, we are mostly using Node.js with a PostgreSQL database, and, well, we have integrated Redis for caching to keep things fast. I mean, it is basically the standard setup we have had since twenty twenty two. Oh, and we are currently migrating some of the legacy services over to Go, specifically the ones handling high concurrency, which should, you know, probably reduce our latency by about fifteen percent. I think it makes sense given our current scale, but, uh, I wanted to see, like, what your take is on that? Do you think we should stick with that roadmap or maybe look into something like Rust for those specific modules?", "clean": "For the backend, we are mostly using Node.js with a PostgreSQL database, and we have integrated Redis for caching to keep things fast. It is the standard setup we have had since 2022. And we are currently migrating some of the legacy services over to Go, specifically the ones handling high concurrency, which should probably reduce our latency by about 15%. I think it makes sense given our current scale, but I wanted to see what your take is on that? Do you think we should stick with that roadmap or maybe look into something like Rust for those specific modules?", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl454.wav", "scene": "meeting", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 255, "source_record_id": "recheck.jsonl:454"}
{"oral": "Bottom line is we need to cut cut the marketing spend by twelve percent. Twelve percent because... wait, fifteen percent. Oh no, sorry, it's actually ten percent. Since ROI is... ROI is falling, we should move fund to the new new product line, make sure we arrive at goal by December thirtieth. It's the only way to stay in budget for this year's fiscal year.", "clean": "Bottom line is we need to cut the marketing spend by 10%. Since ROI is falling, we should move funds to the new product line, make sure we arrive at goal by December 30. It's the only way to stay in budget for this year's fiscal year.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl460.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 261, "source_record_id": "recheck.jsonl:460"}
{"oral": "So, if we shift ten percent... no, twelve percent of the marketing spend, then we could, uh, we could actually scale the growth, scale the growth in the APAC region. I mean, if we... if we reallocate those funds... wait, we could probably hit target by October twelfth, oh no, October fourteenth.", "clean": "If we shift 12% of the marketing spend, then we could actually scale the growth in the APAC region. If we reallocate those funds, we could probably hit target by October 14.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl466.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 263, "source_record_id": "recheck.jsonl:466"}
{"oral": "Eh, Sarah said the churn rate hit twelve percent... no, it was fourteen percent. I think we need to look look at the feedback, just... how to put it... re examine the feedback patterns, you know?", "clean": "Sarah said the churn rate hit 14%. I think we need to look at the feedback, re-examine the feedback patterns.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl475.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 267, "source_record_id": "recheck.jsonl:475"}
{"oral": "So uh looking at the conversion rates for variant B it seems like we actually saw a jump of about three point five percent which is great but then again when you look at the bounce rate it kind of... well it suggests that the new landing page might be a bit too cluttered for the mobile users so maybe we should try a third version for the launch on June first twenty twenty five to see if we can hit that five hundred twelve lead target", "clean": "Looking at the conversion rates for variant B, it seems we saw a jump of about 3.5%, which is great, but then again when you look at the bounce rate, it suggests that the new landing page might be a bit too cluttered for the mobile users, so maybe we should try a third version for the launch on June 1, 2025 to see if we can hit that 512 lead target", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl468.wav", "scene": "meeting", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 265, "source_record_id": "recheck.jsonl:468"}
{"oral": "So um, looking at the pipeline, I think we really need to focus on the senior dev roles for the marketing team... oh no, I mean the product team, since we got like, uh, four open headcounts right now and we're sort of... we're trailing behind on the... actually, let's just prioritize the backend stuff first because the current velocity is like, um, twelve percent... wait, no, fifteen percent lower than we planned for this quarter.", "clean": "Looking at the pipeline, I think we really need to focus on the senior dev roles for the product team, since we got 4 open headcounts right now and we're trailing behind on the... actually, let's just prioritize the backend stuff first because the current velocity is 15% lower than we planned for this quarter.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl467.wav", "scene": "meeting", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 264, "source_record_id": "recheck.jsonl:467"}
{"oral": "I think some think we can finish by Friday... oh no, by next Tuesday. But we really need, we really need to scale back scope. We should... I mean... if we cut two features, we can actually hit that date, hit that date.", "clean": "I think we can finish by next Tuesday. But we really need to scale back scope. If we cut 2 features, we can actually hit that date.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl476.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 268, "source_record_id": "recheck.jsonl:476"}
{"oral": "So, um, looking back at the last three months, if we just... if we actually shift our focus toward the Q four targets, then we could probably avoid the same bottlenecks we hit in August... oh no, I mean September, and maybe just streamline how the dev team and product sync up every Tuesday morning to keep things moving faster, you know?", "clean": "Looking back at the last 3 months, if we shift our focus toward the Q4 targets, then we could probably avoid the same bottlenecks we hit in September, and maybe just streamline how the dev team and product sync up every Tuesday morning to keep things moving faster.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl479.wav", "scene": "meeting", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 269, "source_record_id": "recheck.jsonl:479"}
{"oral": "So, look, I think we all feel the weight of this quarter, and I know, I know the numbers were tough, like, we hit maybe forty two percent of target, but honestly, the main issue was... the main issue was the misalignment between product and engineering. We need to find a way to find a way to sync better, maybe start by reviewing the roadmap on October twelfth, or we could just... actually, let's just focus on the blockers first because the blockers are what's really killing our velocity right now.", "clean": "I think we all feel the weight of this quarter, and I know the numbers were tough, we hit maybe 42% of target, but honestly, the main issue was the misalignment between product and engineering. We need to find a way to sync better, maybe start by reviewing the roadmap on October 12, or actually, let's just focus on the blockers first because the blockers are what's really killing our velocity right now.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl469.wav", "scene": "meeting", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 266, "source_record_id": "recheck.jsonl:469"}
{"oral": "Sarah, conversion rate rose ten percent... no, twelve percent. Twelve percent is a big win for the new page, the new page.", "clean": "Sarah, conversion rate rose 12%. 12% is a big win for the new page.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl482.wav", "scene": "meeting", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 272, "source_record_id": "recheck.jsonl:482"}
{"oral": "Okay, bottom line is we gotta roll back the latest deploy. Uh, look, we seeing a spike in five hundred errors, like, maybe ten percent... oh no, it's actually twenty two percent of all traffic. So, we need to get this fixed fast before it hit the east... uh, winter region. I mean, the East region. Let's move.", "clean": "Bottom line is we gotta roll back the latest deploy. We're seeing a spike in 500 errors, maybe 22% of all traffic. So, we need to get this fixed fast before it hits the East region. Let's move.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl480.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 270, "source_record_id": "recheck.jsonl:480"}
{"oral": "Look, we need to roll back now. Uh, the error rate is hitting five percent, and... I mean, if we... we should probably just revert to version two point zero four immediately.", "clean": "Look, we need to roll back now. The error rate is hitting 5%, and we should probably just revert to v2.04 immediately.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl487.wav", "scene": "meeting", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 275, "source_record_id": "recheck.jsonl:487"}
{"oral": "Look, we've been looking at the current backlog and the team team is feeling a bit stretched, so we might need to push the launch to October twelfth... oh wait, October fourteenth. I think we think we should probably trim some features just to be safe, so let's aim for a soft launch by October twentieth instead, yeah?", "clean": "Look, we've been looking at the current backlog and the team is feeling a bit stretched, so we might need to push the launch to October 14. I think we should probably trim some features just to be safe, so let's aim for a soft launch by October 20 instead, yeah?", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl481.wav", "scene": "meeting", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 271, "source_record_id": "recheck.jsonl:481"}
{"oral": "Q3 loss was ten percent... no, twelve percent.", "clean": "Q3 loss was 12%.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl495.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 279, "source_record_id": "recheck.jsonl:495"}
{"oral": "I'm not totally sure but I think... uh... if we push the date to October twelfth, we might... wait, let me rephrase... if we shift the deadline to October twelfth, we can probably hit those three main KPIs.", "clean": "I'm not totally sure, but I think if we shift the deadline to October 12, we can probably hit those three main KPIs.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl486.wav", "scene": "meeting", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 274, "source_record_id": "recheck.jsonl:486"}
{"oral": "Honestly, the current sprint cycle, sprint cycle is just too slow. We need to move to two week cycles, two week cycles, to hit that ninety percent stability target by December twelfth.", "clean": "Honestly, the current sprint cycle is just too slow. We need to move to 2-week cycles to hit that 90% stability target by December 12.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl488.wav", "scene": "meeting", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 276, "source_record_id": "recheck.jsonl:488"}
{"oral": "Actually, I think some people think this loop is causing the memory leak, but it is not, it is not the loop. The issue is actually the cache size... oh no, the cache limit, it is set to five hundred megabytes, which is way too high, so we should probably reduce it to two hundred fifty six megabytes, you know, just to be safe, right?", "clean": "I think some people think this loop is causing the memory leak, but it is not the loop. The issue is the cache limit; it is set to 500MB, which is way too high, so we should probably reduce it to 256MB just to be safe, right?", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl485.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 273, "source_record_id": "recheck.jsonl:485"}
{"oral": "Um, so the conversion rate hit twelve percent... oh no, it was fourteen point two percent, and we... just... the landing page... like, the bounce rate was... wait, let me just say, it actually performed better, right?", "clean": "The conversion rate hit 14.2%, and the landing page bounce rate actually performed better, right?", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl494.wav", "scene": "meeting", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 278, "source_record_id": "recheck.jsonl:494"}
{"oral": "预算能加点吗？十万，不对，十五万，我觉得行。", "clean": "预算能加点吗？15万，我觉得行。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl513.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 285, "source_record_id": "recheck.jsonl:513"}
{"oral": "我我我今年KPI达成率是百分之一百一十二，嗯，其实应该是百分之一百一十五，就是我想想啊，整体超额完成了一点。", "clean": "我今年KPI达成率是115%，整体超额完成了一点。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl506.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 282, "source_record_id": "recheck.jsonl:506"}
{"oral": "So uh look, I know last quarter was... it was rough, really rough for everyone. We had those server outages and, like, the team burnout... um, wait, the budget cut was... ten percent... no, sorry, twelve percent... wait, that's wrong, it was fifteen percent! And uh, the timeline... that was a mess, I think. I mean, the communication... uh, actually, the real issue we gotta tackle today is how we handle the handoff between product and engineering.", "clean": "Look, I know last quarter was rough, really rough for everyone. We had those server outages and the team burnout. The budget cut was 15%! And the timeline was a mess, I think. The real issue we gotta tackle today is how we handle the handoff between product and engineering.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl489.wav", "scene": "meeting", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 277, "source_record_id": "recheck.jsonl:489"}
{"oral": "Uh, I think we should, well, maybe add one click checkout, and then, how should I put it, just, like, a loyalty system, plus, oh, we could get ten percent more users, ya, in the new market, right, ba.", "clean": "I think we should add one-click checkout, a loyalty system, plus we could get 10% more users in the new market.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl497.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 280, "source_record_id": "recheck.jsonl:497"}
{"oral": "Look, for the phase one deadline, it's more like... we are polishing existing features than building new ones, just to hit October twelfth. I think we... let's say, it's better to trim scope now than risk a delay, you know?", "clean": "For the phase 1 deadline, we are polishing existing features than building new ones, just to hit October 12. I think it's better to trim scope now than risk a delay.", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl500.wav", "scene": "meeting", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 281, "source_record_id": "recheck.jsonl:500"}
{"oral": "简单来说，昨天那个接口跑了五百次，呃，结果…怎么说呢，其实整体成功率在百分之九十八左右，得在下周二前把这块新逻辑给对齐。", "clean": "简单来说，昨天那个接口跑了500次，结果整体成功率在98%左右，得在下周二前把这块新逻辑给对齐。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl509.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 284, "source_record_id": "recheck.jsonl:509"}
{"oral": "老大说流程得简化，嗯，我觉得……其实应该把……怎么说呢，直接在星周报里同步，吧，这样能省掉两三个环节呀。", "clean": "老大说流程得简化，我觉得其实应该直接在星周报里同步，这样能省掉2-3个环节呀。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl507.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 283, "source_record_id": "recheck.jsonl:507"}
{"oral": "嗯，咱们之前聊过上半年投放转化率才百分之三点五，就是得再增加五万块预算，怎么说呢，不然新项目很难跑通吧。", "clean": "咱们之前聊过上半年投放转化率才3.5%，得再增加50000元预算，不然新项目很难跑通吧。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl517.wav", "scene": "meeting", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 286, "source_record_id": "recheck.jsonl:517"}
{"oral": "嗯，与其死磕功能迭代，不如先把转化率提高到百分之十二，啊不对，百分之十五吧，这比再死等新用户更有意义哈。", "clean": "与其死磕功能迭代，不如先把转化率提高到15%，这比再死等新用户更有意义哈。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl520.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 288, "source_record_id": "recheck.jsonl:520"}
{"oral": "先是看那个，那那那个DAU，增长了百分之五，不对，百分之八，然后是留存率稍微掉了点，最后整体转化大概百分之十二。", "clean": "先是看DAU增长了8%，然后是留存率稍微掉了点，最后整体转化大概12%。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl524.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 290, "source_record_id": "recheck.jsonl:524"}
{"oral": "刚才小李说预算得控制在五万，啊不对，是六万。我觉得预算预算得再加一点，怎么说呢，增加十个百分点吧，方案整体方案整体才稳妥哈。", "clean": "刚才小李说预算得控制在6万。我觉得预算得再加一点，增加10%吧，整体方案才稳妥哈。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl522.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 289, "source_record_id": "recheck.jsonl:522"}
{"oral": "嗯，老大之前说咱们组现在只有五个人，哦不是，应该是六个。怎么说呢，我我我觉得这规模其实挺挺合适的，就是能保证效率吧。", "clean": "老大之前说咱们组现在应该是6个人。我觉得这规模其实挺合适的，能保证效率吧。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl530.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 292, "source_record_id": "recheck.jsonl:530"}
{"oral": "下个月市场波动挺大，咱们那个目标…嗯…怎么说呢…得定在五十万，不对，六十万，等等错了，是八十万！", "clean": "下个月市场波动挺大，咱们目标得定在80万！", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl532.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 294, "source_record_id": "recheck.jsonl:532"}
{"oral": "那个，你可能觉得，得先写完文档才，嗯，不是那样，其实是先出个 Demo，大概，得，三五个版本，迭代出来再写。", "clean": "你可能觉得得先写完文档，其实是先出个Demo，大概得3-5个版本迭代出来再写。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl531.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 293, "source_record_id": "recheck.jsonl:531"}
{"oral": "嗯，我看方案里预算写了三万块，啊不对，是五万块，就是成本有点高吧，你觉得呢？", "clean": "我看方案里预算写了5万元，成本有点高吧，你觉得呢？", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl539.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 297, "source_record_id": "recheck.jsonl:539"}
{"oral": "如果咱们能把转化率转化率提高到百分之五，提高到百分之五，那我们在下个季度下个季度得能多赚三百万，到那个时候再考虑扩充团队。", "clean": "如果咱们能把转化率提高到5%，那我们在下个季度得能多赚300万元，到那个时候再考虑扩充团队。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl540.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 298, "source_record_id": "recheck.jsonl:540"}
{"oral": "嗯，大家可能觉得这次转化率掉是因为投放渠道问题，其实不是，怎么说呢，主要还是产品端某个 Bug 导致，我看数据是百分之十二点五，啊不对，是百分之八点二，这个波动导致我们整体 KPI 没达标吧。", "clean": "大家可能觉得这次转化率掉是因为投放渠道问题，其实不是，主要还是产品端某个 Bug 导致，我看数据是 8.2%，这个波动导致我们整体 KPI 没达标吧。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl529.wav", "scene": "meeting", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 291, "source_record_id": "recheck.jsonl:529"}
{"oral": "说白了，嗯，我们组现在就是主攻主攻那个底层架构，怎么说呢，目前大概有十二个人，十二个人在做，整体节奏挺快吧，再之后我们会往新方向走哈。", "clean": "说白了，我们组现在主攻底层架构，目前大概有12个人在做，整体节奏挺快吧，之后我们会往新方向走哈。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl542.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 299, "source_record_id": "recheck.jsonl:542"}
{"oral": "其实这事不是单纯 链路问题，嗯，其实是，怎么说呢，核心在于那个接口权限，大概有百分之十二…不对，百分之十八 接口没对齐，导致整体风控延迟了。", "clean": "其实这事不是单纯链路问题，核心在于接口权限，大概有18%接口没对齐，导致整体风控延迟了。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl533.wav", "scene": "meeting", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 295, "source_record_id": "recheck.jsonl:533"}
{"oral": "我想想啊，如果今年我们能把那个转化率提升到百分之十二，啊不对，是百分之十五吧，那我们明年大概能，怎么说呢，争取把整体GMV拉高到五千万左右哈。", "clean": "如果今年我们能把转化率提升到15%，那我们明年大概能争取把整体GMV拉高到5000万左右。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl543.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 300, "source_record_id": "recheck.jsonl:543"}
{"oral": "结论百分之十，啊不对，百分之十二，主要因为咱们那个新渠道转化率涨快了。", "clean": "结论12%，主要因为咱们新渠道转化率涨快了。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl550.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 303, "source_record_id": "recheck.jsonl:550"}
{"oral": "哦对了，那个，截止日期...应该是，呃，下周三，不对，二零二四年十月十八号，大家记得，那个，及时把PPT，上传到Drive，千万别...忘了。", "clean": "截止日期应该是2024年10月18日，大家记得及时把PPT上传到Drive，千万别忘了。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl546.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 301, "source_record_id": "recheck.jsonl:546"}
{"oral": "刚才王总提到提到市场饱和，我觉得我觉得这个观点不对，目前份额只有百分之十二，啊不对，百分之十五，还得再深挖深挖，才能到目标点。", "clean": "刚才王总提到市场饱和，我觉得这个观点不对，目前份额只有15%，还得再深挖，才能到目标点。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl549.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 302, "source_record_id": "recheck.jsonl:549"}
{"oral": "哦对了，嗯，这次转化率转化率得达到百分之十五，怎么说呢，这个指标指标得在二零二五年三月前完成吧。", "clean": "这次转化率得达到15%，这个指标得在2025年3月前完成吧。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl554.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 305, "source_record_id": "recheck.jsonl:554"}
{"oral": "咱们这次转化率转化率到了百分之十二，啊不对，百分之十四，这个增长增长势头挺好。", "clean": "咱们这次转化率到了14%，增长势头挺好。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl553.wav", "scene": "meeting", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 304, "source_record_id": "recheck.jsonl:553"}
{"oral": "哦对了，嗯，交付进度目前在百分之八十，啊不，百分之八十五，就是，我们那个...怎么说呢...整体跑通了，在验收阶段哈。", "clean": "交付进度目前在85%，整体跑通了，在验收阶段。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl557.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 306, "source_record_id": "recheck.jsonl:557"}
{"oral": "嗯，先是去年六月，我们团队才三个人，然后，怎么说呢，规模在，扩到了十来个，最后，诶，现在的架构应该是分了三个小组吧。", "clean": "去年6月，我们团队才3个人，然后规模扩到了10来个，最后现在的架构应该是分了3个小组吧。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl561.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 308, "source_record_id": "recheck.jsonl:561"}
{"oral": "嗯，不是说项目延期，其实其实是进度调整，怎么说呢，原定百分之八十，我想想，其实是达到了百分之九十五，吧。", "clean": "不是说项目延期，其实是进度调整，原定80%，其实是达到了95%。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl563.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 310, "source_record_id": "recheck.jsonl:563"}
{"oral": "王总，人力资源人力资源得调一下，大概得要三个人三个人，不对，两个就行，下周二下午两点半开会讨论", "clean": "王总，人力资源得调一下，大概要2个人就行，下周二14:30开会讨论", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl564.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 311, "source_record_id": "recheck.jsonl:564"}
{"oral": "你说咱们这个PR流程能不能再快点，嗯，我想想啊，之前一个需求跑了三四天，其实只要把代码提交到Git以后，等一个小时左右审批完就行，就是这样吧。", "clean": "你说咱们这个PR流程能不能再快点，之前一个需求跑了3-4天，其实只要把代码提交到Git以后，等1小时左右审批完就行。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl562.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 309, "source_record_id": "recheck.jsonl:562"}
{"oral": "大家注意一下进度，先是需要在6月1号前完成初稿，然后是6月15号前把审核流程走完，最后在6月30号前正式提交，希望能严格把控时间节点。", "clean": "大家注意一下进度：\n1. 6月1日前完成初稿\n2. 6月15日前把审核流程走完\n3. 6月30日前正式提交\n希望能严格把控时间节点。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl558.wav", "scene": "meeting", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 307, "source_record_id": "recheck.jsonl:558"}
{"oral": "嗯，咱们组平时同步进度就是，每周一上午十点，啊不对，是十点半开周会，然后...那个流程应该是...怎么说呢，大家把文档填好传到共享盘吧，你觉得这样操作行吗呀？", "clean": "咱们组平时同步进度，每周一10:30开周会，流程应该是大家把文档填好传到共享盘吧，你觉得这样操作行吗？", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl566.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 313, "source_record_id": "recheck.jsonl:566"}
{"oral": "嗯，目前ROI大概…就是…百分之十五吧，诶，你觉得呢哈？", "clean": "目前ROI大概15%，你觉得呢？", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl580.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 317, "source_record_id": "recheck.jsonl:580"}
{"oral": "我想想啊，与其盯着百分之三点五这个转化率，不如看活跃用户比上个月增长了五百一十二个，这更关键吧。", "clean": "与其盯着3.5%这个转化率，不如看活跃用户比上个月增长了512个，这更关键吧。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl577.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 315, "source_record_id": "recheck.jsonl:577"}
{"oral": "这个季度增长率大概百分之十二，额，其实说得更准点，也就是百分之十二点三，基本达标。", "clean": "这个季度增长率大概12%，其实说得更准点，也就是12.3%，基本达标。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl582.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 318, "source_record_id": "recheck.jsonl:582"}
{"oral": "先是把那个API接通，然后…其实应该是先调接口，最后再把这三个页面上线，估计得花五天，得赶紧弄完。", "clean": "先调接口，最后把这三个页面上线，估计得花5天，得赶紧弄完。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl573.wav", "scene": "meeting", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 314, "source_record_id": "recheck.jsonl:573"}
{"oral": "与其投广告投广告，不如投人投人，预算得三万，不对，五万。", "clean": "与其投广告，不如投人，预算得5万。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl591.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 322, "source_record_id": "recheck.jsonl:591"}
{"oral": "嗯，之前看市场规模，市场规模大概有三千万，啊不对，是五千万，怎么说呢，我想想啊，其实核心核心在于这个 ROI 能够快速跑通吧。", "clean": "之前看市场规模大概有5000万，其实核心在于这个ROI能够快速跑通吧。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl578.wav", "scene": "meeting", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 316, "source_record_id": "recheck.jsonl:578"}
{"oral": "老王，我想想啊，咱们这次用三台服务器，哦不是，五台，得保证在高峰期能扛住。", "clean": "老王，咱们这次用5台服务器，得保证在高峰期能扛住。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl585.wav", "scene": "meeting", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 319, "source_record_id": "recheck.jsonl:585"}
{"oral": "诶，去年KPI达成率百分之八十，哦不是，百分之九十二，嗯，这个进度…我想想啊…其实进度比预期快，就是挺稳吧？", "clean": "去年KPI达成率92%，进度比预期快，挺稳吧？", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl586.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 320, "source_record_id": "recheck.jsonl:586"}
{"oral": "其实不是说进度慢，嗯，是那个…风险点在…其实是逻辑漏洞，可能导致百分之五得损耗，得再关注下。", "clean": "其实不是说进度慢，是风险点在逻辑漏洞，可能导致5%的损耗，得再关注下。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl592.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 323, "source_record_id": "recheck.jsonl:592"}
{"oral": "诶，其实留存率不是在下降，怎么说呢，整体趋势是在回升。我想想啊，上周大概百分之十二，不对，百分之十五，回升回升了百分之十五吧。", "clean": "其实留存率不是在下降，整体趋势是在回升。上周大概回升了15%吧。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl588.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 321, "source_record_id": "recheck.jsonl:588"}
{"oral": "哦对了，那个合规风险点，大概有百分之三点五，得在二零二四年十二月一号前搞定，这个新规定...呃，得赶紧再确认一下，对，是对的。", "clean": "合规风险点大概有3.5%，得在2024年12月1日前搞定，这个新规定得赶紧再确认一下。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl594.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 324, "source_record_id": "recheck.jsonl:594"}
{"oral": "嗯，刚才老王说用 Redis 缓存，就是我觉得响应时间得压到两百毫秒，哦不是，应该是五十毫秒，怎么说呢，这样用户体验才稳吧。", "clean": "刚才老王说用 Redis 缓存，我觉得响应时间得压到 50 毫秒，这样用户体验才稳吧。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl595.wav", "scene": "meeting", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 325, "source_record_id": "recheck.jsonl:595"}
{"oral": "嗯，与其死磕这百分之五，就是，比追求量，得更看重客单价吧。", "clean": "与其死磕这5%，比追求量，得更看重客单价吧。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl597.wav", "scene": "meeting", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 327, "source_record_id": "recheck.jsonl:597"}
{"oral": "老王，那个需求文档需求文档得赶紧出，周五周五下午三点半之前得交，不然进度进度会卡住。", "clean": "老王，需求文档得赶紧出，周五15:30之前得交，不然进度会卡住。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl598.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 328, "source_record_id": "recheck.jsonl:598"}
{"oral": "咱们人力成本预计增加百分之十，啊不对，百分之十五，人力成本得覆盖。那个招聘渠道，对了，能不能给咱们额外拨两万块", "clean": "咱们人力成本预计增加15%，人力成本得覆盖。招聘渠道能不能给咱们额外拨20000元", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl596.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 326, "source_record_id": "recheck.jsonl:596"}
{"oral": "嗯，如果这波流量能维持在两千五百左右，两千五百的话，那咱们转化率哈，估计能到百分之五，百分之五吧。", "clean": "如果这波流量能维持在2500左右，那咱们转化率估计能到5%。", "usage_text_words": null, "audio": "audio/meeting__recheckjsonl600.wav", "scene": "meeting", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 329, "source_record_id": "recheck.jsonl:600"}
{"oral": "关于这篇论文的第三章，我认为在讨论 Transformer 模型时，应该更详细地分析自注意力机制的计算开销。", "clean": "关于这篇论文的第三章，我认为在讨论 Transformer 模型时，应该更详细地分析自注意力机制的计算开销。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl604.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 332, "source_record_id": "recheck.jsonl:604"}
{"oral": "关于第三季度市场分析报告，我认为我们需要在用户留存率这一块增加更多的数据对比图表。", "clean": "关于第三季度市场分析报告，我认为我们需要在用户留存率这一块增加更多的数据对比图表。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl605.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 333, "source_record_id": "recheck.jsonl:605"}
{"oral": "关于这篇论文的第三章，我认为在分析对比实验结果时，需要引入更多的消融实验来验证该模块的有效性。", "clean": "关于这篇论文的第三章，我认为在分析对比实验结果时，需要引入更多的消融实验来验证该模块的有效性。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl603.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 331, "source_record_id": "recheck.jsonl:603"}
{"oral": "最近压力有点大，感觉需要找个周末去山里徒步放松一下，你有兴趣一起去吗？", "clean": "最近压力有点大，感觉需要找个周末去山里徒步放松一下，你有兴趣一起去吗？", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl607.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 334, "source_record_id": "recheck.jsonl:607"}
{"oral": "备忘：明天上午 9 点开周会，记得带上上季度的销售数据分析报告。", "clean": "备忘：明天上午 9 点开周会，记得带上上季度的销售数据分析报告。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl612.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 337, "source_record_id": "recheck.jsonl:612"}
{"oral": "今天下午 3 点的周会重点讨论 Q3 的预算分配以及新产品的上线计划，请大家提前准备好各自的汇报 PPT。", "clean": "今天下午 3 点的周会重点讨论 Q3 的预算分配以及新产品的上线计划，请大家提前准备好各自的汇报 PPT。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl608.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 335, "source_record_id": "recheck.jsonl:608"}
{"oral": "帮我把下午3点在会议室B的周会推迟到4点，并通知所有参会人员。", "clean": "帮我把下午3点在会议室B的周会推迟到4点，并通知所有参会人员。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl615.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 340, "source_record_id": "recheck.jsonl:615"}
{"oral": "记得明天早上8点提醒我给客户发邮件，附件要加上最新的报价单和产品规格书。", "clean": "记得明天早上8点提醒我给客户发邮件，附件要加上最新的报价单和产品规格书。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl613.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 338, "source_record_id": "recheck.jsonl:613"}
{"oral": "最近压力有点大，感觉每天都在重复同样的事情，好想找个周末去山里徒步放松一下。", "clean": "最近压力有点大，感觉每天都在重复同样的事情，好想找个周末去山里徒步放松一下。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl614.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 339, "source_record_id": "recheck.jsonl:614"}
{"oral": "我想搜一下2024年最受欢迎的开源深度学习框架对比分析。", "clean": "我想搜一下2024年最受欢迎的开源深度学习框架对比分析。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl616.wav", "scene": "passthrough", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 341, "source_record_id": "recheck.jsonl:616"}
{"oral": "我想在 2025 年的第三季度完成这个项目的初步调研，重点分析用户留存率和转化率。", "clean": "我想在 2025 年的第三季度完成这个项目的初步调研，重点分析用户留存率和转化率。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl609.wav", "scene": "passthrough", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 336, "source_record_id": "recheck.jsonl:609"}
{"oral": "请帮我导航到上海市浦东新区世纪大道 1 号，避开拥堵路段。", "clean": "请帮我导航到上海市浦东新区世纪大道 1 号，避开拥堵路段。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl618.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 342, "source_record_id": "recheck.jsonl:618"}
{"oral": "帮我把明天下午 3点 的会议提醒改成 3点半，顺便在日程表里加上准备 PPT 的事项。", "clean": "帮我把明天下午 3:00 的会议提醒改成 3:30，顺便在日程表里加上准备 PPT 的事项。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl619.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 343, "source_record_id": "recheck.jsonl:619"}
{"oral": "帮我记一下：明天下午 2 点记得给客户发邮件，确认最终的合同条款。", "clean": "帮我记一下：明天下午 2 点记得给客户发邮件，确认最终的合同条款。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl622.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 346, "source_record_id": "recheck.jsonl:622"}
{"oral": "我想搜索一下 2024 年最受欢迎的开源大语言模型对比分析报告。", "clean": "我想搜索一下 2024 年最受欢迎的开源大语言模型对比分析报告。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl620.wav", "scene": "passthrough", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 344, "source_record_id": "recheck.jsonl:620"}
{"oral": "提醒我明天早上 8 点记得带上笔记本电脑和充电线。", "clean": "提醒我明天早上 8 点记得带上笔记本电脑和充电线。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl625.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 349, "source_record_id": "recheck.jsonl:625"}
{"oral": "请问从现在的位置到上海图书馆最快的路线是什么？预计需要多久？", "clean": "请问从现在的位置到上海图书馆最快的路线是什么？预计需要多久？", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl624.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 348, "source_record_id": "recheck.jsonl:624"}
{"oral": "请帮我把这段 Python 代码中的循环逻辑优化一下，现在的运行时间太长了。", "clean": "请帮我把这段 Python 代码中的循环逻辑优化一下，现在的运行时间太长了。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl621.wav", "scene": "passthrough", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 345, "source_record_id": "recheck.jsonl:621"}
{"oral": "这次季度汇报的重点应该放在用户增长率和获客成本的优化上，尤其是 20-30 岁这个年龄段。", "clean": "这次季度汇报的重点应该放在用户增长率和获客成本的优化上，尤其是 20-30 岁这个年龄段。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl623.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 347, "source_record_id": "recheck.jsonl:623"}
{"oral": "请导航到上海市静安区南京西路 1618 号，避开拥堵路段。", "clean": "请导航到上海市静安区南京西路 1618 号，避开拥堵路段。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl626.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 350, "source_record_id": "recheck.jsonl:626"}
{"oral": "周六晚上 7 点在三里屯见，我们去吃那家新开的泰国餐厅吧。", "clean": "周六晚上 7 点在三里屯见，我们去吃那家新开的泰国餐厅吧。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl631.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 355, "source_record_id": "recheck.jsonl:631"}
{"oral": "最近读的那本书非常有启发，尤其是关于时间管理的那一部分，让我重新思考了工作优先级。", "clean": "最近读的那本书非常有启发，尤其是关于时间管理的那一部分，让我重新思考了工作优先级。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl628.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 352, "source_record_id": "recheck.jsonl:628"}
{"oral": "最近在阅读关于 Transformer 架构的论文，特别是关于注意力机制在长文本处理上的效率优化问题。", "clean": "最近在阅读关于 Transformer 架构的论文，特别是关于注意力机制在长文本处理上的效率优化问题。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl627.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 351, "source_record_id": "recheck.jsonl:627"}
{"oral": "请将这段 Python 代码中的循环逻辑优化一下，尽量降低时间复杂度，避免使用嵌套循环。", "clean": "请将这段 Python 代码中的循环逻辑优化一下，尽量降低时间复杂度，避免使用嵌套循环。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl630.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 354, "source_record_id": "recheck.jsonl:630"}
{"oral": "关于深度学习在医学影像分析中的应用，我认为目前的注意力机制在处理长距离依赖时仍有提升空间。", "clean": "关于深度学习在医学影像分析中的应用，我认为目前的注意力机制在处理长距离依赖时仍有提升空间。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl629.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 353, "source_record_id": "recheck.jsonl:629"}
{"oral": "你好，我的订单号是 88273615，我想咨询一下包裹目前到哪个转运中心了？", "clean": "你好，我的订单号是 88273615，我想咨询一下包裹目前到哪个转运中心了？", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl632.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 356, "source_record_id": "recheck.jsonl:632"}
{"oral": "请帮我搜索一下 Python 中如何使用异步 IO 提高大规模网络请求的并发性能。", "clean": "请帮我搜索一下 Python 中如何使用异步 IO 提高大规模网络请求的并发性能。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl633.wav", "scene": "passthrough", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 357, "source_record_id": "recheck.jsonl:633"}
{"oral": "帮我订一张明天下午 2 点从上海到北京的高铁二等座机票。", "clean": "帮我订一张明天下午 2 点从上海到北京的高铁二等座机票。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl635.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 358, "source_record_id": "recheck.jsonl:635"}
{"oral": "帮我记录一条备忘录：记得在周三下午 3 点预约牙医，并带上之前的 X 光片。", "clean": "帮我记录一条备忘录：记得在周三下午 3 点预约牙医，并带上之前的 X 光片。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl638.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 361, "source_record_id": "recheck.jsonl:638"}
{"oral": "今天心情不错，打算周末约几个好朋友一起去郊区露营，顺便拍一些照片。", "clean": "今天心情不错，打算周末约几个好朋友一起去郊区露营，顺便拍一些照片。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl636.wav", "scene": "passthrough", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 359, "source_record_id": "recheck.jsonl:636"}
{"oral": "将这个函数重构为异步模式，并确保在调用 API 时添加 500 毫秒的超时处理。", "clean": "将这个函数重构为异步模式，并确保在调用 API 时添加 500 毫秒的超时处理。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl641.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 363, "source_record_id": "recheck.jsonl:641"}
{"oral": "关于深度学习中的梯度消失问题，我们应该讨论一下使用 ReLU 激活函数与 LSTM 结构的对比。", "clean": "关于深度学习中的梯度消失问题，我们应该讨论一下使用 ReLU 激活函数与 LSTM 结构的对比。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl637.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 360, "source_record_id": "recheck.jsonl:637"}
{"oral": "在深度学习模型中，通过引入注意力机制可以有效解决长序列依赖问题，从而提升Transformer架构的性能。", "clean": "在深度学习模型中，通过引入注意力机制可以有效解决长序列依赖问题，从而提升Transformer架构的性能。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl639.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 362, "source_record_id": "recheck.jsonl:639"}
{"oral": "关于第二季度的市场推广计划，我认为我们需要增加在社交媒体端的投入，重点针对 18 到 25 岁的用户群。", "clean": "关于第二季度的市场推广计划，我认为我们需要增加在社交媒体端的投入，重点针对 18 到 25 岁的用户群。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl642.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 364, "source_record_id": "recheck.jsonl:642"}
{"oral": "明天下午3点你有空吗？我想约你一起去市中心那家新开的咖啡馆坐坐。", "clean": "明天下午3点你有空吗？我想约你一起去市中心那家新开的咖啡馆坐坐。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl643.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 365, "source_record_id": "recheck.jsonl:643"}
{"oral": "帮我把今天的待办事项记录下来：下午2点开周会，4点回复客户邮件，晚上记得给猫喂食。", "clean": "帮我把今天的待办事项记录下来：下午2点开周会，4点回复客户邮件，晚上记得给猫喂食。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl644.wav", "scene": "passthrough", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 366, "source_record_id": "recheck.jsonl:644"}
{"oral": "我想查询一下 2024 年巴黎奥运会开幕式的具体时间。", "clean": "我想查询一下 2024 年巴黎奥运会开幕式的具体时间。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl647.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 369, "source_record_id": "recheck.jsonl:647"}
{"oral": "我的订单号是 20240812993，我想申请办理退货，请问具体的流程是怎么样的？", "clean": "我的订单号是 20240812993，我想申请办理退货，请问具体的流程是怎么样的？", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl646.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 368, "source_record_id": "recheck.jsonl:646"}
{"oral": "关于第二季度的市场分析报告，我认为我们需要重点关注用户留存率的下滑原因。", "clean": "关于第二季度的市场分析报告，我认为我们需要重点关注用户留存率的下滑原因。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl651.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 371, "source_record_id": "recheck.jsonl:651"}
{"oral": "关于大语言模型在医疗诊断中的鲁棒性问题，我认为我们需要引入更多高质量的对抗性样本进行微调。", "clean": "关于大语言模型在医疗诊断中的鲁棒性问题，我认为我们需要引入更多高质量的对抗性样本进行微调。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl645.wav", "scene": "passthrough", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 367, "source_record_id": "recheck.jsonl:645"}
{"oral": "请将当前项目的所有 API 接口文档同步到 Confluence 页面，并通知前端开发团队。", "clean": "请将当前项目的所有 API 接口文档同步到 Confluence 页面，并通知前端开发团队。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl654.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 372, "source_record_id": "recheck.jsonl:654"}
{"oral": "关于这个季度的人员绩效考核，我认为应该增加一个关于团队协作的量化指标。", "clean": "关于这个季度的人员绩效考核，我认为应该增加一个关于团队协作的量化指标。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl650.wav", "scene": "passthrough", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 370, "source_record_id": "recheck.jsonl:650"}
{"oral": "我想查一下订单号 20240512998 的快递到哪了，为什么显示在上海停了 3 天？", "clean": "我想查一下订单号 20240512998 的快递到哪了，为什么显示在上海停了 3 天？", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl656.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 373, "source_record_id": "recheck.jsonl:656"}
{"oral": "我们需要在模型训练中引入 L2 正则化，以防止在验证集上出现过拟合现象。", "clean": "我们需要在模型训练中引入 L2 正则化，以防止在验证集上出现过拟合现象。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl658.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 375, "source_record_id": "recheck.jsonl:658"}
{"oral": "关于这篇论文的第三章，我认为目前的消融实验不足以支撑结论，需要增加对比基准模型。", "clean": "关于这篇论文的第三章，我认为目前的消融实验不足以支撑结论，需要增加对比基准模型。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl657.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 374, "source_record_id": "recheck.jsonl:657"}
{"oral": "最近压力有点大，想在这个周末去郊外徒步放松一下，你有兴趣一起去吗？", "clean": "最近压力有点大，想在这个周末去郊外徒步放松一下，你有兴趣一起去吗？", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl661.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 377, "source_record_id": "recheck.jsonl:661"}
{"oral": "最近在看一本关于量子力学的书，感觉里面的波函数概念非常有趣。", "clean": "最近在看一本关于量子力学的书，感觉里面的波函数概念非常有趣。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl659.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 376, "source_record_id": "recheck.jsonl:659"}
{"oral": "帮我查一下去上海虹桥机场最快的路线，现在堵车严重吗？", "clean": "帮我查一下去上海虹桥机场最快的路线，现在堵车严重吗？", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl663.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 378, "source_record_id": "recheck.jsonl:663"}
{"oral": "在 Transformer 架构中，多头注意力机制能够让模型在不同表示子空间中共同学习信息。", "clean": "在 Transformer 架构中，多头注意力机制能够让模型在不同表示子空间中共同学习信息。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl665.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 380, "source_record_id": "recheck.jsonl:665"}
{"oral": "关于这篇论文的第三章，我认为在实验部分需要增加更多关于消融研究的对比数据，以增强结论的说服力。", "clean": "关于这篇论文的第三章，我认为在实验部分需要增加更多关于消融研究的对比数据，以增强结论的说服力。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl664.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 379, "source_record_id": "recheck.jsonl:664"}
{"oral": "我想咨询一下订单号 20240512998 的物流状态，显示已经到达分拨中心三天了，请帮我核实一下。", "clean": "我想咨询一下订单号 20240512998 的物流状态，显示已经到达分拨中心三天了，请帮我核实一下。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl666.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 381, "source_record_id": "recheck.jsonl:666"}
{"oral": "帮我订一张明天上午 10 点去上海的高铁二等座票。", "clean": "帮我订一张明天上午 10 点去上海的高铁二等座票。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl670.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 385, "source_record_id": "recheck.jsonl:670"}
{"oral": "记得在购物清单里加上 2 升牛奶、一打鸡蛋和一袋全麦面包。", "clean": "记得在购物清单里加上 2 升牛奶、一打鸡蛋和一袋全麦面包。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl669.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 384, "source_record_id": "recheck.jsonl:669"}
{"oral": "关于第二季度的销售报告，我建议在 PPT 中增加一个对比图表，详细分析 5 月份的增长原因。", "clean": "关于第二季度的销售报告，我建议在 PPT 中增加一个对比图表，详细分析 5 月份的增长原因。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl667.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 382, "source_record_id": "recheck.jsonl:667"}
{"oral": "最近在读的一本书非常有启发，它讨论了认知心理学在现代产品设计中的实际应用。", "clean": "最近在读的一本书非常有启发，它讨论了认知心理学在现代产品设计中的实际应用。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl668.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 383, "source_record_id": "recheck.jsonl:668"}
{"oral": "关于下个季度的市场推广计划，我认为我们应该增加在短视频平台的投放预算，以吸引更多 18 到 25 岁的年轻用户。", "clean": "关于下个季度的市场推广计划，我认为我们应该增加在短视频平台的投放预算，以吸引更多 18 到 25 岁的年轻用户。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl671.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 386, "source_record_id": "recheck.jsonl:671"}
{"oral": "记得买牛奶、鸡蛋和洗洁精，顺便问一下妈妈周末要不要一起去公园野餐。", "clean": "记得买牛奶、鸡蛋和洗洁精，顺便问一下妈妈周末要不要一起去公园野餐。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl673.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 387, "source_record_id": "recheck.jsonl:673"}
{"oral": "在 Python 中，如何使用 pandas 库实现对一个 2GB 大小的 CSV 文件进行分块读取？", "clean": "在 Python 中，如何使用 pandas 库实现对一个 2GB 大小的 CSV 文件进行分块读取？", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl676.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 388, "source_record_id": "recheck.jsonl:676"}
{"oral": "关于第二季度的营收增长，我们目前的初步分析显示，主要驱动力来自于东南亚市场的 B2C 业务扩张。", "clean": "关于第二季度的营收增长，我们目前的初步分析显示，主要驱动力来自于东南亚市场的 B2C 业务扩张。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl678.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 390, "source_record_id": "recheck.jsonl:678"}
{"oral": "周末有空吗？我们一起去那家新开的日料店试试吧，听说他们的Omakase性价比很高。", "clean": "周末有空吗？我们一起去那家新开的日料店试试吧，听说他们的Omakase性价比很高。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl679.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 391, "source_record_id": "recheck.jsonl:679"}
{"oral": "关于这个季度的KPI，我认为我们在用户留存率方面还需要进一步优化，建议增加针对新用户的引导流程。", "clean": "关于这个季度的KPI，我认为我们在用户留存率方面还需要进一步优化，建议增加针对新用户的引导流程。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl681.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 392, "source_record_id": "recheck.jsonl:681"}
{"oral": "记得明天早上 8 点去超市买牛奶、鸡蛋和 2 公斤苹果。", "clean": "记得明天早上 8 点去超市买牛奶、鸡蛋和 2 公斤苹果。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl685.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 394, "source_record_id": "recheck.jsonl:685"}
{"oral": "帮我记录一条备忘录：记得在周五下班前把项目进度表发给王经理。", "clean": "帮我记录一条备忘录：记得在周五下班前把项目进度表发给王经理。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl684.wav", "scene": "passthrough", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 393, "source_record_id": "recheck.jsonl:684"}
{"oral": "帮我预定明天下午 2 点的会议室，记得邀请产品组的所有成员。", "clean": "帮我预定明天下午 2 点的会议室，记得邀请产品组的所有成员。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl687.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 396, "source_record_id": "recheck.jsonl:687"}
{"oral": "记得在 2025 年 3 月 15 日之前提交所有的报销凭证，否则无法在当月结算。", "clean": "记得在 2025 年 3 月 15 日之前提交所有的报销凭证，否则无法在当月结算。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl686.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 395, "source_record_id": "recheck.jsonl:686"}
{"oral": "最近心情挺不错的，感觉生活节奏慢了下来，打算周末去郊区徒步放松一下。", "clean": "最近心情挺不错的，感觉生活节奏慢了下来，打算周末去郊区徒步放松一下。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl691.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 398, "source_record_id": "recheck.jsonl:691"}
{"oral": "我们在部署 K8s 集群时遇到了 Pod 调度问题，需要检查一下节点资源配额和 Taints 设置。", "clean": "我们在部署 K8s 集群时遇到了 Pod 调度问题，需要检查一下节点资源配额和 Taints 设置。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl689.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 397, "source_record_id": "recheck.jsonl:689"}
{"oral": "关于这篇论文的第三章，我认为目前的消融实验不足以证明该模块的有效性，需要增加对比基准。", "clean": "关于这篇论文的第三章，我认为目前的消融实验不足以证明该模块的有效性，需要增加对比基准。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl692.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 399, "source_record_id": "recheck.jsonl:692"}
{"oral": "提醒我明天上午 10 点给王经理发邮件，把上周的销售报告附件加上。", "clean": "提醒我明天上午 10 点给王经理发邮件，把上周的销售报告附件加上。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl694.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 401, "source_record_id": "recheck.jsonl:694"}
{"oral": "我想在 Python 中使用 FastAPI 构建一个异步的 REST API，应该如何配置中间件？", "clean": "我想在 Python 中使用 FastAPI 构建一个异步的 REST API，应该如何配置中间件？", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl693.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 400, "source_record_id": "recheck.jsonl:693"}
{"oral": "我想咨询一下订单号 2024051299 的物流状态，为什么已经在转运中心停留 3 天了？", "clean": "我想咨询一下订单号 2024051299 的物流状态，为什么已经在转运中心停留 3 天了？", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl695.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 402, "source_record_id": "recheck.jsonl:695"}
{"oral": "周末有空吗？我们要不要一起去那家新开的 brunch 餐厅试试，听说他们的班尼迪克蛋很出名。", "clean": "周末有空吗？我们要不要一起去那家新开的 brunch 餐厅试试，听说他们的班尼迪克蛋很出名。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl696.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 403, "source_record_id": "recheck.jsonl:696"}
{"oral": "在 Python 中，如何使用 asyncio 库来并发处理多个 HTTP 请求以提高爬虫的效率？", "clean": "在 Python 中，如何使用 asyncio 库来并发处理多个 HTTP 请求以提高爬虫的效率？", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl697.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 404, "source_record_id": "recheck.jsonl:697"}
{"oral": "关于这个分布式系统的共识算法，我认为 Raft 比 Paxos 更容易在实际工程中实现，因为它的领导者选举机制更清晰。", "clean": "关于这个分布式系统的共识算法，我认为 Raft 比 Paxos 更容易在实际工程中实现，因为它的领导者选举机制更清晰。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl700.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 405, "source_record_id": "recheck.jsonl:700"}
{"oral": "帮我写个接口，嗯，与其用 Zustand 存状态，不如直接在 Nuxt 后端处理，感觉比 Tiptap 传参要快得多，我想用 LlamaIndex 挂个索引，哦不对，应该是用 Tauri 端的本地存储，这样性能更强吧。", "clean": "帮我写个接口，与其用 Zustand 存状态，不如直接在 Nuxt 后端处理，感觉比 Tiptap 传参要快得多，我想用 Tauri 端的本地存储，这样性能更强吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl704.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 408, "source_record_id": "recheck.jsonl:704"}
{"oral": "那个，如果咱们用 Tauri 把这个前端给包起来，然后把 hidden_dim 设成五百一十二，怎么说呢，要是咱们能把 Keras 那套逻辑迁移到 Julia 里面去，那我们就能把 alpha 参数调得更灵活一些。", "clean": "如果咱们用 Tauri 把前端给包起来，然后把 hidden_dim 设成 512，要是咱们能把 Keras 那套逻辑迁移到 Julia 里面去，那我们就能把 alpha 参数调得更灵活一些。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl709.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 411, "source_record_id": "recheck.jsonl:709"}
{"oral": "诶，你说这个 cache ttl 设置成六十秒会不会太短了？我想想啊，我现在是用 Julia 调 Scala 接口，结果发现 cache ttl 缓存失效太快，太快了，怎么说呢，我怀疑 Lua 脚本那边逻辑有问题，有问题，应该是把 TTL 读成毫秒吧。", "clean": "你说这个 cache_ttl 设置成60秒会不会太短了？我想想，我现在是用 Julia 调 Scala 接口，结果发现 cache_ttl 缓存失效太快，我怀疑 Lua 脚本那边逻辑有问题，应该是把 TTL 读成毫秒吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl706.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 409, "source_record_id": "recheck.jsonl:706"}
{"oral": "诶，我现在在用 Tauri 搞一个 Clojure 写的后端，就是这个 Astro 框架那边，Astro 框架那边好像有个 ConnectionTimeout 问题，我想想啊，帮我写个单元测试，测试一下 compression level 设成九，不对，设成五的时候能不能正常跑通吧。", "clean": "我现在在用 Tauri 搞一个 Clojure 写的后端，Astro 框架那边好像有个 ConnectionTimeout 问题，帮我写个单元测试，测试一下 compression_level 设成 5 的时候能不能正常跑通吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl708.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 410, "source_record_id": "recheck.jsonl:708"}
{"oral": "哎呀太夸张了，嗯，这段代码竟然在用 Elixir 处理数据，哦不是，是 Julia，怎么说呢，我明明配置了 Prisma 但现在一直报 ConnectionRefused，而且还跳出个 NullPointerException，你帮我看看怎么回事吧。", "clean": "太夸张了，这段代码竟然在用 Julia 处理数据，我明明配置了 Prisma 但现在一直报 ConnectionRefused，而且还跳出个 NullPointerException，你帮我看看怎么回事吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl702.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 407, "source_record_id": "recheck.jsonl:702"}
{"oral": "诶那个，与其用 Scala 这种，额，不如直接用 Haskell 重构，比那个，怎么说呢，我想想，比用 Lua 写的逻辑更函数式一点，你帮我把这块代码在，再优化下，用 FastHTML 那个框架，对，FastHTML，得把接口改掉", "clean": "与其用 Scala，不如直接用 Haskell 重构，比用 Lua 写的逻辑更函数式一点，你帮我把这块代码再优化下，用 FastHTML 框架，得把接口改掉", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl712.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 413, "source_record_id": "recheck.jsonl:712"}
{"oral": "我想写个 regex 匹配 Knative 那个 config 里的 lora alpha 参数, 应该是用 Scala 写, 啊不对, 是用 Lua 写, 这样能直接在 runtime 跑, 你觉得呢?", "clean": "我想写个 regex 匹配 Knative config 里的 lora_alpha 参数，应该是用 Lua 写，这样能直接在 runtime 跑，你觉得呢？", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl714.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 414, "source_record_id": "recheck.jsonl:714"}
{"oral": "嗯，先是我想在 HuggingFace 上拉个模型，然后用 Julia 跑了一下 beta2 版本，就是 compression ratio 设成了百分之零点五，不对，应该是百分之零点八，最后在 Kotlin 这一层报错了呀。", "clean": "我想在 HuggingFace 上拉个模型，然后用 Julia 跑了一下 beta2 版本，compression_ratio 设成了 0.8%，最后在 Kotlin 这一层报错了。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl719.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 417, "source_record_id": "recheck.jsonl:719"}
{"oral": "对了，我觉得我觉得这个 Drizzle 适配得不太好，嗯，要是把……怎么说呢，把 compression ratio 调到零点五，不对，是零点八，应该能优化性能，话说咱们这个 Mistral 模型，模型是用 Xformers 跑，跑在 Prisma 上吧？", "clean": "我觉得 Drizzle 适配得不太好，要是把 compression_ratio 调到 0.8，应该能优化性能。话说咱们这个 Mistral 模型是用 Xformers 跑在 Prisma 上吧？", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl718.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 416, "source_record_id": "recheck.jsonl:718"}
{"oral": "诶，你说把这个 Elixir 逻辑迁移到 Zig 能不能解决内存碎片问题？我想想啊，我现在的内存占用到了八百兆，不对，是六百四十兆，我觉得这个迁移，迁移到 Zig 应该能优化，能优化掉很多冗余，反正 Vercel 那边部署也支持，应该可以吧。", "clean": "你说把 Elixir 逻辑迁移到 Zig 能不能解决内存碎片问题？我现在的内存占用到了640MB，我觉得迁移到 Zig 应该能优化掉很多冗余，反正 Vercel 那边部署也支持，应该可以吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl715.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 415, "source_record_id": "recheck.jsonl:715"}
{"oral": "嗯，如果 timeout ms 设成五千毫秒的话，那我们就能排除掉，排除掉网络抖动问题，我想想啊，要是后端用 Scala 写的…不对，应该是用 Haskell 写的，那逻辑层可能哈，可能就卡在内存回收那块吧。", "clean": "如果 timeout_ms 设成5000毫秒的话，那我们就能排除掉网络抖动问题。要是后端用 Haskell 写的，那逻辑层可能就卡在内存回收那块吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl722.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 420, "source_record_id": "recheck.jsonl:722"}
{"oral": "哦对了,帮我写个正则表达式,要把 DeepSpeed 里面那段 Pydantic 校验逻辑,校验逻辑给提取出来,对了,要把那个 Olive 框架, Olive 框架里定义的一百二十八位长度字符串给匹配到,哎不对,应该是两百五十六位,你得帮我兼容下 Xformers 那个配置项,配置项得支持 Elixir 这种格式,懂我意思吧?", "clean": "帮我写个正则表达式，要把 DeepSpeed 里面那段 Pydantic 校验逻辑提取出来，要把 Olive 框架里定义的 256 位长度字符串给匹配到，你得帮我兼容下 Xformers 那个配置项，配置项得支持 Elixir 这种格式，懂我意思吧？", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl721.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 419, "source_record_id": "recheck.jsonl:721"}
{"oral": "嗯, 如果这块内存泄漏在 Zustand 状态管理里, 那我们就能快速定位, 怎么说呢, 只要把 tRPC 的那个接口超时时间改成五秒, 哦不是, 三秒, 应该就能排查出问题吧。", "clean": "如果这块内存泄漏在 Zustand 状态管理里，那我们就能快速定位，只要把 tRPC 的接口超时时间改成3秒，应该就能排查出问题吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl723.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 421, "source_record_id": "recheck.jsonl:723"}
{"oral": "嗯，帮我把这个 query 换成 index seek 吧，怎么说呢，现在这个 query plan 跑出来是 index scan，导致在三千五百毫秒左右才返回，得优化一下得太慢了哈。", "clean": "帮我把这个 query 换成 index seek 吧，现在这个 query plan 跑出来是 index scan，导致在 3500 毫秒左右才返回，得优化一下，太慢了。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl732.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 425, "source_record_id": "recheck.jsonl:732"}
{"oral": "你说用 Ollama 跑那个 Olive 优化过模型能不能配 Drizzle？嗯，我想想，把 lora rank 设成六十四，batch_size 调到三十二，不对，应该是十六，应该没问题吧。", "clean": "你说用 Ollama 跑 Olive 优化过模型能不能配 Drizzle？我想想，把 lora_rank 设成64，batch_size 调到16，应该没问题吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl725.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 422, "source_record_id": "recheck.jsonl:725"}
{"oral": "助手，我想在用 Dart 和 XState 写的状态机里优化下 Drizzle 的查询，现在频繁报 IndexOutOfBounds 还有 ConnectionTimeout，帮我看看怎么调。", "clean": "助手，我想在用 Dart 和 XState 写的状态机里优化下 Drizzle 的查询，现在频繁报 IndexOutOfBounds 还有 ConnectionTimeout，帮我看看怎么调。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl729.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 423, "source_record_id": "recheck.jsonl:729"}
{"oral": "帮我把这段 Python 代码，翻译成 Java，哦不是，是 Go 语言，把那个 list 转换成 slice 就行。", "clean": "帮我把这段 Python 代码翻译成 Go 语言，把 list 转换成 slice 就行。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl734.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 427, "source_record_id": "recheck.jsonl:734"}
{"oral": "嗯，你帮我把 Docker 镜像重新打一遍，因为我想想啊，我想想啊，我得把 Dart 加上 Lua 还有 Solidity 全给打包进去，就是，为了确保环境一致，那个……把 Ollama 端口改成，改成一万一千零一号，诶，对了，应该是十一千一百一十一号吧。", "clean": "你帮我把 Docker 镜像重新打一遍，我得把 Dart 加上 Lua 还有 Solidity 全给打包进去，为了确保环境一致，把 Ollama 端口改成 11111 号。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl731.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 424, "source_record_id": "recheck.jsonl:731"}
{"oral": "嗯，之前写的那个 Modal 组件，Modal 组件逻辑有点乱，怎么说呢，我想把 z index 改成九百九十九，改成九百九十九吧。", "clean": "之前写的那个 Modal 组件逻辑有点乱，我想把 z-index 改成 999。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl737.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 429, "source_record_id": "recheck.jsonl:737"}
{"oral": "嗯，我现在的索引索引优化是这样，把那个 B tree 换成 Hash 索引，哦不是，应该是覆盖索引，覆盖索引这样能快一点吧，你觉得呢？", "clean": "我现在的索引优化是这样，把 B-tree 换成覆盖索引，这样能快一点吧，你觉得呢？", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl740.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 431, "source_record_id": "recheck.jsonl:740"}
{"oral": "对了，那个…把…把 batch size 改成八", "clean": "把batch size改成8", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl741.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 432, "source_record_id": "recheck.jsonl:741"}
{"oral": "帮我写个数据迁移脚本，先是把旧库里版本一点零的数据全部导出，导出到临时表，然后检查一下字段，只要状态不等于零且等级大于等于五的记录才迁移，迁移这些记录，最后用C加加写个校验逻辑，调用那个双冒号后面的静态方法核对一遍，核对一遍总数对不对。", "clean": "帮我写个数据迁移脚本：\n1. 把旧库里v1.0的数据全部导出到临时表\n2. 检查一下字段，只要状态!=0且等级>=5的记录才迁移\n3. 用C++写个校验逻辑，调用双冒号后面的静态方法核对一遍总数对不对", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl736.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 428, "source_record_id": "recheck.jsonl:736"}
{"oral": "对了，帮我写个 Regex 匹配那个 email，嗯，长度得大于等于十个字符，不对，应该是八个字符吧，得支持那种带点号的域名。", "clean": "帮我写个 Regex 匹配 email，长度得 >= 8 个字符，得支持那种带点号的域名。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl743.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 434, "source_record_id": "recheck.jsonl:743"}
{"oral": "写个v一点二，不对，v一点三 API？行吧。", "clean": "写个v1.3 API？行吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl747.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 436, "source_record_id": "recheck.jsonl:747"}
{"oral": "我那个Docker镜像现在在跑v一点二，啊不对，是v一点五，我想把那个...呃...怎么说呢，把端口映射改成八千零八十二，不对，是八千零八十", "clean": "我Docker镜像现在在跑v1.5，我想把端口映射改成8080", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl745.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 435, "source_record_id": "recheck.jsonl:745"}
{"oral": "正则正则…写个正则…匹配不等于零，你觉得呢？", "clean": "写个正则匹配!=0，你觉得呢？", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl759.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 442, "source_record_id": "recheck.jsonl:759"}
{"oral": "你帮我弄下CI CD流水线，先是触发Git push，触发Git push之后跑单元测试，然后把镜像推到Docker hub，推到Docker hub，最后部署到k8s集群。", "clean": "你帮我弄下CI/CD流水线：\n1. 触发Git push\n2. 跑单元测试\n3. 把镜像推到Docker Hub\n4. 部署到K8s集群", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl753.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 438, "source_record_id": "recheck.jsonl:753"}
{"oral": "说白了，我想想啊，这段逻辑逻辑太冗余，帮我把这个 loop 循环，循环改成 map 映射，把时间复杂度降到 O 1。", "clean": "说白了，这段逻辑太冗余，帮我把这个 loop 循环改成 map 映射，把时间复杂度降到 O(1)。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl749.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 437, "source_record_id": "recheck.jsonl:749"}
{"oral": "哦对了，帮我看看这段C加加代码在干嘛，那个if判断里是不是写成x大于等于五了，不对，应该是感叹号等于零，那个箭头指向哪个member我不清楚。", "clean": "帮我看看这段C++代码在干嘛，if判断里是不是写成x!=0了，箭头指向哪个member我不清楚。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl757.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 441, "source_record_id": "recheck.jsonl:757"}
{"oral": "论文里说要把 learning rate 设成 0.01，不对，是 0.001。我觉得这太保守，咱们能不能...呃...怎么说呢...直接用 C加加 写个 loop，判断 loss 大于等于 0.5 就跳出", "clean": "论文里说要把 learning rate 设成 0.001。我觉得这太保守，咱们能不能直接用 C++ 写个 loop，判断 loss >= 0.5 就跳出", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl754.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 439, "source_record_id": "recheck.jsonl:754"}
{"oral": "说白了，帮我写个 regex，匹配那个长度大于五位，不对，大于八位，且不包含感叹号等于号 这种特殊字符 字符串", "clean": "说白了，帮我写个regex，匹配长度大于8位，且不包含!=这种特殊字符的字符串", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl761.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 444, "source_record_id": "recheck.jsonl:761"}
{"oral": "嗯，如果把这个 API 接口改成 POST，啊不，是 PUT，那我们就能，就是，直接更新数据库里那三千五百条记录吧。", "clean": "如果把这个API接口改成PUT，那我们就能直接更新数据库里那3500条记录吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl760.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 443, "source_record_id": "recheck.jsonl:760"}
{"oral": "我也不太确定，但是感觉把 batch size 改成 512 之后，训练速度能快 百分之三点五 左右。", "clean": "我也不太确定，但是感觉把 batch size 改成 512 之后，训练速度能快 3.5% 左右。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl774.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 449, "source_record_id": "recheck.jsonl:774"}
{"oral": "嗯，组长说这个接口得覆盖边界值，我觉得我觉得得写个 unit test，就是把输入设成十个，哦不是，设成一百个吧，哈。", "clean": "组长说这个接口得覆盖边界值，我觉得得写个 unit test，把输入设成100个。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl763.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 446, "source_record_id": "recheck.jsonl:763"}
{"oral": "把这个，把这个 Loop, 把 Loop 抽出来, 怎么说, 抽成函数, 你觉得呢？", "clean": "把Loop抽成函数，你觉得呢？", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl776.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 450, "source_record_id": "recheck.jsonl:776"}
{"oral": "他说用C sharp, 哦不是, C加加写个test", "clean": "他说用C++写个test", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl782.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 453, "source_record_id": "recheck.jsonl:782"}
{"oral": "哦对了，我想给那个模块加个 cache 机制，就是把所有大于等于五百个元素 列表存到 Redis 里，诶，不对，应该是大于等于五百一十二个，然后用双冒号调用那个 Singleton 类，哈。", "clean": "我想给那个模块加个 cache 机制，把所有 >= 512 个元素的列表存到 Redis 里，然后用 :: 调用 Singleton 类。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl768.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 448, "source_record_id": "recheck.jsonl:768"}
{"oral": "帮我看看这段C加加代码，看看这段C加加代码在干嘛，然后这里有个指针箭头指向 member 变量，还有这个 double 类型的 threshold 阈值阈值怎么设，另外判断 if 语句里那个不等于零逻辑对不对", "clean": "帮我看看这段C++代码在干嘛，然后这里有个指针->指向member变量，还有这个double类型的threshold阈值怎么设，另外判断if语句里那个!=0逻辑对不对", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl777.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 451, "source_record_id": "recheck.jsonl:777"}
{"oral": "对了，帮我 check 一下那个 deployment 配置，我觉得我觉得那个内存限制设得太低，得把 memory limit 调高点。那个值设成四 G…不对，八 G…等等错了，是十六 G！话说咱们那个…要是把…怎么说呢…如果把 C plus plus 编译器版本换成十四号的话，会不会导致 binary 兼容性问题？你检查检查这个版本，检查检查一下环境变量里有没有那个 Path 路径。", "clean": "帮我check一下deployment配置，我觉得内存限制设得太低，得把memory limit调高点。那个值设成16G！如果把C++编译器版本换成14的话，会不会导致binary兼容性问题？你检查一下环境变量里有没有Path路径。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl762.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 445, "source_record_id": "recheck.jsonl:762"}
{"oral": "老大说要把那个接口改成C井号，哦不是，是C加加，我觉得那个…怎么说呢…如果用双冒号来调用，可能性能会更好。", "clean": "老大说要把接口改成C++，我觉得如果用::来调用，可能性能会更好。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl781.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 452, "source_record_id": "recheck.jsonl:781"}
{"oral": "嗯，我想用 Redis 缓存，不对，应该用 Memcached 比较快，怎么说呢，这个 latency 在十毫秒以内，能到道，性能得提升百分之二十吧。", "clean": "我想用 Memcached 比较快，latency 在 10毫秒 以内，性能得提升 百分之二十 吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl784.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 454, "source_record_id": "recheck.jsonl:784"}
{"oral": "嗯，帮我写个API，写个API接口，参数要是大于等于十，不对，大于等于二十，就是用C加加实现吧。", "clean": "帮我写个API接口，参数要>=20，用C++实现吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl785.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 455, "source_record_id": "recheck.jsonl:785"}
{"oral": "嗯，我也不太确定，但是感觉这个指针指针指向不对，怎么说呢，感觉应该用双冒号双冒号去调用，诶，可能得改一下吧。", "clean": "我也不太确定，但感觉这个指针指向不对，感觉应该用::去调用，可能得改一下吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl788.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 457, "source_record_id": "recheck.jsonl:788"}
{"oral": "老大，我觉得这个内存泄漏内存泄漏应该在那个…怎么说呢…如果检查一下 cache 缓存的话，应该是五百兆…不对，是八百兆，赶紧看看那个内存内存泄漏再到哪儿触发的。", "clean": "老大，我觉得这个内存泄漏应该检查一下cache缓存的话，应该是800MB，赶紧看看内存泄漏是在哪儿触发的。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl786.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 456, "source_record_id": "recheck.jsonl:786"}
{"oral": "嗯，如果这个 cluster 节点数设成三，不对，五的话，那咱们就... 怎么说呢... 咱们就能实现高可用，实现高可用就到道了吧。", "clean": "如果这个 cluster 节点数设成 5 的话，咱们就能实现高可用，实现高可用就达到了吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl793.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 460, "source_record_id": "recheck.jsonl:793"}
{"oral": "老王，我觉得我觉得用 Redis 缓存，缓存这个方案比较快，然后那个...要是把...怎么说呢...如果把 TTL 设为 三百秒，不对，是 六百秒，这样压力压力会小点。", "clean": "老王，我觉得用 Redis 缓存这个方案比较快，如果把 TTL 设为 600 秒，这样压力会小点。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl789.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 458, "source_record_id": "recheck.jsonl:789"}
{"oral": "嗯，不是让我去知乎搜那个，其实是我想听Adele，不对，我想听邓紫棋在Discord上播的白噪音吧，怎么说呢，帮我放三分钟的。", "clean": "不是让我去知乎搜，其实是我想听邓紫棋在Discord上播的白噪音，帮我放3分钟的。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl803.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 465, "source_record_id": "recheck.jsonl:803"}
{"oral": "嗯，用Redis存...我想想啊，其实直接走MySQL可能更好，还有就是判断那个status不等于零得写清楚，另外再把C加加的版本号升到v一点零吧，诶，就是要把索引优化好呀。", "clean": "其实直接走MySQL可能更好，判断status!=0得写清楚，另外把C++的版本号升到v1.0吧，要把索引优化好。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl794.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 461, "source_record_id": "recheck.jsonl:794"}
{"oral": "帮我搜搜，与其看蔡依林，蔡依林，不如看哔哩哔哩，哔哩哔哩，比Zara更时尚 电影，不对，比Zara更时尚 纪录片。", "clean": "帮我搜搜，与其看蔡依林，不如看哔哩哔哩，比Zara更时尚纪录片。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl805.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 466, "source_record_id": "recheck.jsonl:805"}
{"oral": "我现在这个 query 慢得离谱，就是那个……我想把 index 重新建一遍，要是让 status 不等于零，再……不对，应该是在 where 子句里加个作用域限制，你觉得这样能把 latency 降下来吗？", "clean": "我现在这个 query 慢得离谱，我想把 index 重新建一遍，应该是在 where 子句里加个 :: 限制，你觉得这样能把 latency 降下来吗？", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl796.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 462, "source_record_id": "recheck.jsonl:796"}
{"oral": "小助手，嗯，帮我写个 data migration 脚本，把 MySQL 那个表迁移到 PostgreSQL，我想想啊，字段要把 varchar 改成 text，不对，其实是得对齐那个 schema，诶，还有两个索引要删掉，不对不对，是加一个 unique constraint，就是，那个二零二四年十月一号之前的数据不用搬，总之你帮我弄个 python 脚本吧。", "clean": "小助手，帮我写个 data migration 脚本，把 MySQL 表迁移到 PostgreSQL。字段得对齐 schema，加一个 unique constraint。2024年10月1日之前的数据不用搬，总之你帮我弄个 Python 脚本吧。", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl791.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 459, "source_record_id": "recheck.jsonl:791"}
{"oral": "嗯，帮我看看这段代码在干嘛，就是，诶，我想想啊，帮我分析下这个 pointer 后面接双冒号作用域得逻辑，应该在这一块吧？", "clean": "帮我看看这段代码在干嘛，帮我分析下这个 pointer 后面接 :: 的逻辑，应该在这一块吧？", "usage_text_words": null, "audio": "audio/vibe_coding__recheckjsonl799.wav", "scene": "vibe_coding", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 463, "source_record_id": "recheck.jsonl:799"}
{"oral": "哦对了，帮我给蔡依林发个微信，就说，就说我想买台 Tesla Model 3，诶，我想买买台 Tesla Model 3，颜色要白色哈。", "clean": "帮我给蔡依林发个微信，就说我想买台 Tesla Model 3，颜色要白色哈。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl807.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 467, "source_record_id": "recheck.jsonl:807"}
{"oral": "诶，不是让我订盒马鲜生，就是提醒我半个小时后开会哈。我想想啊，开会地点在百胜餐饮，哦不是，在Tesla Model 3展厅吧。", "clean": "不是让我订盒马鲜生，是提醒我30分钟后开会哈。开会地点在Tesla Model 3展厅吧。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl809.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 469, "source_record_id": "recheck.jsonl:809"}
{"oral": "嗯，我最近在看 Kindle Paperwhite，还买了 Lululemon，就是 Lululemon 衣服，诶，我想听听听 Louis Vuitton 那个播客，播客最新一期帮我放一下吧。", "clean": "我最近在看 Kindle Paperwhite，还买了 Lululemon 衣服，我想听 Louis Vuitton 那个播客，最新一期帮我放一下吧。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl808.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 468, "source_record_id": "recheck.jsonl:808"}
{"oral": "诶，帮我把空调打开，就是，我想想啊，刚才在外面拍 GoPro Hero 12 拍太久，热得像在海底捞里坐着一样吧。", "clean": "帮我把空调打开，刚才在外面拍 GoPro Hero 12 拍太久，热得像在海底捞里坐着一样吧。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl811.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 471, "source_record_id": "recheck.jsonl:811"}
{"oral": "诶，我想把闹钟定在明天早上六点…不对，七点…等等错了，是八点！就是我想想啊，全家全家早点开门，我得赶紧买个早餐，然后听首Billie Eilish，你觉得呢吧？", "clean": "我想把闹钟定在明天早上8:00，全家早点开门，我得赶紧买个早餐，然后听首Billie Eilish，你觉得呢？", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl810.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 470, "source_record_id": "recheck.jsonl:810"}
{"oral": "诶，搜一下那个DJI Mini 四 Pro，不对，我想找 宜家 拍 那个 MacBook Air 广告电影吧。", "clean": "搜一下宜家拍的MacBook Air广告电影吧。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl813.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 473, "source_record_id": "recheck.jsonl:813"}
{"oral": "诶，快帮我把空调打开，就是我想想啊，刚才在京东买 Billie Eilish 联名款，还有 Decathlon 运动服，跑了太久，现在热死我了，吧。", "clean": "快帮我把空调打开，刚才在京东买 Billie Eilish 联名款，还有 Decathlon 运动服，跑了太久，现在热死我了。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl812.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 472, "source_record_id": "recheck.jsonl:812"}
{"oral": "如果明天早上七点半，啊不对，七点能起来，那我就能先用 Dyson V15 把地吸一遍，帮我定个七点闹钟。", "clean": "如果明天早上7:00能起来，那我就能先用 Dyson V15 把地吸一遍，帮我定个7:00闹钟。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl819.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 476, "source_record_id": "recheck.jsonl:819"}
{"oral": "诶，如果我想在抖音搜刘若英，啊不对，是盒马鲜生，那能告诉我 lululemon 怎么拼吗？", "clean": "如果我想在抖音搜盒马鲜生，那能告诉我 lululemon 怎么拼吗？", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl816.wav", "scene": "voice_search", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 475, "source_record_id": "recheck.jsonl:816"}
{"oral": "张杰，帮我算下Justin Bieber在Slack发得那个，把三千五百块换成百分之十五，啊不对，百分之二十，得多少钱？", "clean": "张杰，帮我算下Justin Bieber在Slack发的那个，把3500元换成20%，得多少钱？", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl822.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 477, "source_record_id": "recheck.jsonl:822"}
{"oral": "不是让我播 Adele, 啊不对, 播 Netflix, 帮我把空调打开", "clean": "播 Netflix，帮我把空调打开", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl826.wav", "scene": "voice_search", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 479, "source_record_id": "recheck.jsonl:826"}
{"oral": "诶，与其听Justin Bieber新歌，不如看看Amazon的新闻，就是说Xiaomi 14 Ultra比Dyson V15更值得关注吧，我想想啊，帮我播报到今天的摘要哈。", "clean": "与其听Justin Bieber新歌，不如看看Amazon的新闻，Xiaomi 14 Ultra比Dyson V15更值得关注吧，帮我播报到今天的摘要哈。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl824.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 478, "source_record_id": "recheck.jsonl:824"}
{"oral": "诶，帮我开下空调，温度调到二十六度，不对，二十四度，然后播放播放 Ariana Grande 的歌，就是，另外帮我连接 AirPods Pro 吧。", "clean": "帮我开下空调，温度调到24度，然后播放 Ariana Grande 的歌，另外帮我连接 AirPods Pro 吧。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl834.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 481, "source_record_id": "recheck.jsonl:834"}
{"oral": "我待会儿得去趟盒马鲜生，帮我搜附近三公里内，啊不对，五公里内加油站", "clean": "我待会儿得去趟盒马鲜生，帮我搜附近5公里内加油站", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl843.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 484, "source_record_id": "recheck.jsonl:843"}
{"oral": "哦对了，帮我算算算算，一个路易威登包包三万六千块，换成特斯拉 Model 3 多少辆 Model 3，换成多少辆", "clean": "对了，帮我算算，一个路易威登包包36000元，换成特斯拉 Model 3 是多少辆？", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl837.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 482, "source_record_id": "recheck.jsonl:837"}
{"oral": "我朋友说 Bruno Mars 好像比 Billie Eilish 大很多，我觉得不对，帮我查下 Bruno Mars 今年几岁，再查下他 Instagram 资料，不对，查 Adele 的资料。", "clean": "我朋友说 Bruno Mars 好像比 Billie Eilish 大很多，我觉得不对，帮我查下 Bruno Mars 今年几岁，再查下 Adele 的资料。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl827.wav", "scene": "voice_search", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 480, "source_record_id": "recheck.jsonl:827"}
{"oral": "小助手，搜搜附近加油站，搜搜附近的加油站，我要买个TikTok，啊不对，是Nintendo Switch，还有个GoPro Hero 12，快点帮我找找找找加油站。", "clean": "小助手，搜搜附近的加油站，我要买个Nintendo Switch，还有个GoPro Hero 12，快点帮我找找加油站。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl839.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 483, "source_record_id": "recheck.jsonl:839"}
{"oral": "诶，如果把闲鱼把闲鱼卖给Amazon的话，那我们是不是就能用TikTok买FamilyMart的东西呀，怎么说呢，张惠妹也会在上面直播吗？", "clean": "如果把闲鱼卖给Amazon的话，那我们是不是就能用TikTok买FamilyMart的东西呀，张惠妹也会在上面直播吗？", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl848.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 487, "source_record_id": "recheck.jsonl:848"}
{"oral": "诶，Justin Bieber 现在的资料说他二十岁，不对，应该说他现在二十一岁吧，就是，你觉得他这个年纪成名太早吗？", "clean": "Justin Bieber 现在的资料说他21岁，你觉得他这个年纪成名太早吗？", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl849.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 488, "source_record_id": "recheck.jsonl:849"}
{"oral": "嗯，我也不太确定，但是感觉感觉客厅灯光调到百分之五十吧，我想想啊，要把林俊杰全家GoPro Hero 12拍 Amazon百度网盘备份，调调到百分之五十。", "clean": "我也不太确定，但是感觉客厅灯光调到50%吧，要把林俊杰全家GoPro Hero 12拍的Amazon百度网盘备份，调到50%。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl845.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 485, "source_record_id": "recheck.jsonl:845"}
{"oral": "哦对了，帮我播播，播放一下Billie Eilish，啊不对，是Justin Bieber那个播客，那个播客最新一期，我想想，怎么说，直接播最新的那一集就行", "clean": "帮我播放Justin Bieber播客最新一期，直接播最新的那一集就行", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl851.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 489, "source_record_id": "recheck.jsonl:851"}
{"oral": "诶，李荣浩不是那个唱特斯拉 Model 3 广告歌吧，就是，他其实是写歌才子，查下他今年三十九岁，啊不对，四十岁，还有具体资料。", "clean": "李荣浩不是唱特斯拉 Model 3 广告歌吧，他其实是写歌才子，查下他今年40岁，还有具体资料。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl854.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 490, "source_record_id": "recheck.jsonl:854"}
{"oral": "帮我设个十分钟计时器，我想查查用 Xiaomi 14 Ultra 拍 Tesla Model 3 效果怎么样，哦不是，是用 Sony WH-1000XM5 录音效果怎么样，要在知乎上看看。", "clean": "帮我设个10分钟计时器，我想查查用 Sony WH-1000XM5 录音效果怎么样，要在知乎上看看。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl847.wav", "scene": "voice_search", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 486, "source_record_id": "recheck.jsonl:847"}
{"oral": "给李宇春发微信，说匹克那个，不对，给陈粒发微信，帮我关闭洗衣机并把声音调小。", "clean": "给陈粒发微信，帮我关闭洗衣机并把声音调小。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl862.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 493, "source_record_id": "recheck.jsonl:862"}
{"oral": "诶，如果我想听听刘若英，啊不对，是Charlie Puth，就是听听Charlie Puth的话，那我们就能睡着吧。", "clean": "如果我想听Charlie Puth的话，那我们就能睡着吧。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl864.wav", "scene": "voice_search", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 494, "source_record_id": "recheck.jsonl:864"}
{"oral": "诶，帮我搜搜毛不易，就是他演的那个电影，啊不，是演的金铲铲之战广告，我想用Xiaomi 14 Pro投屏看，因为 Firefox 浏览器打不开吧。", "clean": "帮我搜搜毛不易演的金铲铲之战广告，我想用Xiaomi 14 Pro投屏看，因为Firefox浏览器打不开。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl870.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 496, "source_record_id": "recheck.jsonl:870"}
{"oral": "诶，帮我给林俊杰发个微信，就说我今天不能去逛ZARA了，嗯，因为我得赶紧回去看看智能门锁关锁没，而且我刚买的那个OnePlus Open，怎么说呢，我想想啊，是打算拿它试一下香奈儿新出那个APP，啊不对，是买香奈儿新出那个包，吧。", "clean": "帮我给林俊杰发个微信，就说我今天不能去逛ZARA了，因为我得赶紧回去看看智能门锁关锁没，而且我刚买的OnePlus Open，打算拿它买香奈儿新出那个包。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl866.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 495, "source_record_id": "recheck.jsonl:866"}
{"oral": "诶，帮我查查Troye Sivan资料，我想知道他比Justin Bieber更出名吗，怎么说呢，与其看他现在开蔚来汽车，不如看看他以前代言爱马仕，不对，是代言巴奴毛肚火锅，他现在是不是二十九岁，啊不，三十一岁，帮我确认下。", "clean": "帮我查查Troye Sivan资料，我想知道他比Justin Bieber更出名吗，与其看他现在开蔚来汽车，不如看看他以前代言巴奴毛肚火锅，他现在是不是31岁，帮我确认下。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl858.wav", "scene": "voice_search", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 491, "source_record_id": "recheck.jsonl:858"}
{"oral": "诶，先切换下一首陈奕迅，然后…帮我搜，不对，找附近加油站，最后把Li Auto L7静音电视吧。", "clean": "先切换下一首陈奕迅，然后找附近加油站，最后把Li Auto L7静音电视吧。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl874.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 498, "source_record_id": "recheck.jsonl:874"}
{"oral": "我把电磁炉关闭关闭，想去山姆会员店买个特步，买个特步，你觉得怎么着？", "clean": "我把电磁炉关闭，想去山姆会员店买个特步，你觉得怎么着？", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl875.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 499, "source_record_id": "recheck.jsonl:875"}
{"oral": "小爱，名创优品那个，MINISO 成立几年了？不对，是问名创优品这个公司，二零一七年成立是不是？", "clean": "小爱，名创优品这个公司，2017年成立是不是？", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl881.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 501, "source_record_id": "recheck.jsonl:881"}
{"oral": "帮我放个白噪音，就是我想想啊，不是那个，放个像在汉庭酒店或者万宁药店那种安静环境下适合睡觉的，不对，直接播个米哈游风格的轻音乐吧，声音调到三十二", "clean": "帮我放个米哈游风格的轻音乐吧，声音调到32", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl876.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 500, "source_record_id": "recheck.jsonl:876"}
{"oral": "不是申通那个拼写，我想问 Xiaomi Vacuum Mop 2 怎么拼，还有李玉刚设定烤箱温度 180度 怎么写，薛之谦名字怎么拼。", "clean": "我想问Xiaomi Vacuum Mop 2怎么拼，还有李玉刚设定烤箱温度180度怎么写，薛之谦名字怎么拼。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl887.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 502, "source_record_id": "recheck.jsonl:887"}
{"oral": "我想听听赵雷，赵雷，不对，听听Photoshop播客最新一期", "clean": "我想听听Photoshop播客最新一期", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl897.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 506, "source_record_id": "recheck.jsonl:897"}
{"oral": "如果我现在开着Li Auto L7在去富士山路上，那帮我播首Halsey，啊不对，播首富士山下的歌，顺便用京东到家点杯喜茶。", "clean": "如果我现在开着Li Auto L7在去富士山路上，那帮我播首富士山下的歌，顺便用京东到家点杯喜茶。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl889.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 503, "source_record_id": "recheck.jsonl:889"}
{"oral": "哦对了，帮我设定设定提醒，半小时后开会，诶，帮我关掉雷神，啊不对，是关闭电风扇，我想想，开会开会时间记得提醒我哈。", "clean": "帮我设定提醒，30分钟后开会，帮我关闭电风扇，开会时间记得提醒我哈。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl892.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 504, "source_record_id": "recheck.jsonl:892"}
{"oral": "帮我关闭电风扇，诶，我想想啊，林俊杰林俊杰，比李宇春更...唔，与其搜李宇春，不如比比林俊杰 Mariah Carey 谁唱功更强吧。", "clean": "帮我关闭电风扇。与其搜李宇春，不如比比林俊杰和Mariah Carey谁唱功更强吧。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl896.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 505, "source_record_id": "recheck.jsonl:896"}
{"oral": "先播赵雷，然后播五月天，不对，播苏打绿，最后放个Marshmello。", "clean": "先播赵雷，然后播苏打绿，最后放个Marshmello。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl898.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 507, "source_record_id": "recheck.jsonl:898"}
{"oral": "小爱，设个十分钟，啊不，五分钟计时器哈，再帮我查查空气质量。", "clean": "小爱，设个5分钟计时器，再帮我查查空气质量。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl900.wav", "scene": "voice_search", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 509, "source_record_id": "recheck.jsonl:900"}
{"oral": "诶，不是要打开电饭煲，就是帮我找个凑凑火锅附近停车场，我想想啊，离那儿五百米以内吧。", "clean": "帮我找个凑凑火锅附近停车场，离那儿500米以内吧。", "usage_text_words": null, "audio": "audio/voice_search__recheckjsonl899.wav", "scene": "voice_search", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 508, "source_record_id": "recheck.jsonl:899"}
{"oral": "So looking at the results, the FID score is actually quite low, which is good, but we need to double check if it's statistically significant. Also, I'm seeing some weird spikes in perplexity when we run Qwen-two, and then we really need a proper evaluation of reward hacking in RLHF pipelines to see if the model is just gaming the metric. Plus, we should probably use BERT for some baseline semantic analysis just to be safe.", "clean": "So looking at the results, the FID score is actually quite low, which is good, but we need to double check if it's statistically significant. Also, I'm seeing some weird spikes in perplexity when we run Qwen2, and then we really need a proper evaluation of reward hacking in RLHF pipelines to see if the model is just gaming the metric. Plus, we should probably use BERT for some baseline semantic analysis just to be safe.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl1180.wav", "scene": "academic", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 512, "source_record_id": "assigned_records.jsonl:1180"}
{"oral": "Reviewer says... says the learning rate is too high. But we... wait... we we tuned it... tuned it across ten ten different seeds... actually it is optimal.", "clean": "Reviewer says the learning rate is too high. But we tuned it across 10 different seeds; actually it is optimal.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl1597.wav", "scene": "academic", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 516, "source_record_id": "assigned_records.jsonl:1597"}
{"oral": "So if we look at long-context retrieval in needle-in-a-haystack, we might... well, maybe we should first consider the scaling laws for mixture of experts, specifically referencing Hoffmann et al. twenty twenty-two. I mean, if we apply those same principles to... actually, let's pivot to mixture of experts load balancing strategies. We know BERT handles local context well, but for MoE we need to ensure the router isn't just hitting one expert. If we scale it to, say, eight experts, the... the efficiency might drop if we don't fix the load balancing first.", "clean": "So if we look at long-context retrieval in needle-in-a-haystack, maybe we should first consider the scaling laws for mixture of experts, specifically referencing Hoffmann et al. 2022. Let's pivot to mixture of experts load balancing strategies. We know BERT handles local context well, but for MoE we need to ensure the router isn't just hitting one expert. If we scale it to, say, 8 experts, the efficiency might drop if we don't fix the load balancing first.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl1195.wav", "scene": "academic", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 513, "source_record_id": "assigned_records.jsonl:1195"}
{"oral": "Thanks for the talk. I was thinking about the sparse attention mechanism efficiency, like, specifically how Gemini one point five handles long contexts. I wonder if multi-token prediction for autoregressive models, multi-token prediction could actually offset some of that overhead. Also, regarding the learning rate, I think we saw a drop at ten percent... oh no, it was twelve point five percent in the baseline. I'm curious if applying similar logic to diffusion model sampling efficiency, sampling efficiency would yield similar gains. What's your take on that?", "clean": "Thanks for the talk. I was thinking about the sparse attention mechanism efficiency, specifically how Gemini 1.5 handles long contexts. I wonder if multi-token prediction for autoregressive models could actually offset some of that overhead. Also, regarding the learning rate, I think we saw a drop at 12.5% in the baseline. I'm curious if applying similar logic to diffusion model sampling efficiency would yield similar gains. What's your take on that?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl1382.wav", "scene": "academic", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 514, "source_record_id": "assigned_records.jsonl:1382"}
{"oral": "So uh I was thinking maybe we apply this to... uh... Qwen... Qwen-one... no sorry, Qwen-two... wait that's wrong, it's Qwen-two point five! Basically, it's more like a memory optimization trick than a full architectural change. Like, we could use gradient checkpointing for billion parameter models to fit the batch, right? But instead of standard ViT, we focus on the attention layers. I'm worried about the... uh... perplexity. It might be... four point two... no, four point five... wait, I got it wrong, it's four point eight! If we can push that down, it'd be huge.", "clean": "I was thinking maybe we apply this to Qwen-2.5! It's more like a memory optimization trick than a full architectural change. We could use gradient checkpointing for billion parameter models to fit the batch, right? But instead of standard ViT, we focus on the attention layers. I'm worried about the perplexity. It might be 4.8! If we can push that down, it'd be huge.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl1100.wav", "scene": "academic", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 511, "source_record_id": "assigned_records.jsonl:1100"}
{"oral": "Thanks for talk. I wonder if the latent space, the latent space is more like a manifold projection than a linear mapping, a linear mapping of the five hundred twenty eight dimensions? Just wondering if that holds.", "clean": "Thanks for the talk. I wonder if the latent space is more like a manifold projection than a linear mapping of the 528 dimensions? Just wondering if that holds.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl1604.wav", "scene": "academic", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 517, "source_record_id": "assigned_records.jsonl:1604"}
{"oral": "So for the multimodal part, I was thinking about... uh... we could apply parameter efficient fine tuning for multi modal, but... wait, actually, if we just use LORA on the vision encoder it might be better. Now, looking at the weight tying, we see some instability. The... uh... the quantization error in four bit weight matrices is basically... it's like... let me rephrase... it's causing a significant drift in the latent space. This hits the FID score pretty hard, like, it jumped by twelve point five points. And if we check the FLOPS, it's down to about two point one teraflops, so it's efficient, but the quality trade off is... well, it's not ideal.", "clean": "So for the multimodal part, I was thinking we could apply parameter efficient fine tuning for multi modal, but if we just use LoRA on the vision encoder it might be better. Now, looking at the weight tying, we see some instability. The quantization error in 4-bit weight matrices is causing a significant drift in the latent space. This hits the FID score pretty hard, it jumped by 12.5 points. And if we check the FLOPS, it's down to about 2.1 teraflops, so it's efficient, but the quality trade off is not ideal.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl103.wav", "scene": "academic", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 510, "source_record_id": "assigned_records.jsonl:103"}
{"oral": "So, the... the... the ablation on the LORA rank shows... uh... a two point four percent drop, which is... is... is basically negligible. Do you think that's significant?", "clean": "The ablation on the LoRA rank shows a 2.4% drop, which is negligible. Do you think that's significant?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl1696.wav", "scene": "academic", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 519, "source_record_id": "assigned_records.jsonl:1696"}
{"oral": "So uh regarding the methodology we basically looked at algorithmic alignment via constitutional ai but we wanted to see how... I mean... if we could isolate the impact of quantization on emergent abilities. We ran this on Claude three point four... oh no, Claude three point five and we noticed that the latency spikes were actually consistent with some of the early observations in Sutskever et al. twenty twelve... no wait, twenty fourteen. So basically the compression doesn't seem to kill the reasoning but it does... uh... it kind of shifts the activation patterns. Does that sound reasonable to you or do you think we need more benchmarks?", "clean": "Regarding the methodology, we looked at algorithmic alignment via Constitutional AI, but we wanted to see if we could isolate the impact of quantization on emergent abilities. We ran this on Claude 3.5 and we noticed that the latency spikes were consistent with some of the early observations in Sutskever et al. 2014. The compression doesn't seem to kill the reasoning, but it shifts the activation patterns. Does that sound reasonable to you or do you think we need more benchmarks?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl1430.wav", "scene": "academic", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 515, "source_record_id": "assigned_records.jsonl:1430"}
{"oral": "So for the setup we are using a five hundred twelve gigabyte V one hundred GPU cluster and we set the learning rate to three point five percent though actually we might... let me rephrase that we are utilizing a dynamic scheduler to stabilize the gradient flow during the first ten epochs.", "clean": "For the setup, we are using a 512GB V100 GPU cluster and we are utilizing a dynamic scheduler to stabilize the gradient flow during the first 10 epochs.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl1876.wav", "scene": "academic", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 520, "source_record_id": "assigned_records.jsonl:1876"}
{"oral": "Dr. Sarah, I wonder if the loss function, the loss function you used for the multi-head attention, maybe it could overfit when sample size is only two hundred forty eight, or I mean, it might be, it might be too sensitive to outliers, so I was wondering if applying a weight decay of zero point zero one might actually stabilize the gradient, you know, in a more robust way?", "clean": "Dr. Sarah, I wonder if the loss function you used for the multi-head attention maybe it could overfit when sample size is only 248, or it might be too sensitive to outliers, so I was wondering if applying a weight decay of 0.01 might actually stabilize the gradient in a more robust way?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl1691.wav", "scene": "academic", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 518, "source_record_id": "assigned_records.jsonl:1691"}
{"oral": "So I think we should switch to the FID score for this baseline, uh, because the current accuracy... wait, maybe better to say, it just doesn't capture the distribution shift we saw in those five hundred twenty samples, and if we look at the lons of the loss curve, it's clear that the model is just... uh... overfiting on the training set, so we need something more robust to evaluate the generative quality.", "clean": "I think we should switch to the FID score for this baseline, because it just doesn't capture the distribution shift we saw in those 520 samples, and if we look at the logs of the loss curve, it's clear that the model is just overfitting on the training set, so we need something more robust to evaluate the generative quality.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2177.wav", "scene": "academic", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 522, "source_record_id": "assigned_records.jsonl:2177"}
{"oral": "So how do we handle the latent space? Well, we use... I mean, we employ a VAE with five hundred twelve dimensions, just to... uh, ensure the manifold is... is stable in the new domain.", "clean": "So how do we handle the latent space? Well, we employ a VAE with 512 dimensions, just to ensure the manifold is stable in the new domain.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2330.wav", "scene": "academic", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 526, "source_record_id": "assigned_records.jsonl:2330"}
{"oral": "Aha! So look, contrastive learning basically pulls positive pairs together and pushes negatives apart, like, you take a batch of five hundred twelve samples and use a contrastive loss, wait, I mean a InfoNCE loss, to maximize similarity, just... basically mapping them to a hypersphere where the distance is, well, minimized for similar images, right?", "clean": "Contrastive learning pulls positive pairs together and pushes negatives apart. You take a batch of 512 samples and use an InfoNCE loss to maximize similarity, mapping them to a hypersphere where the distance is minimized for similar images, right?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2039.wav", "scene": "academic", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 521, "source_record_id": "assigned_records.jsonl:2039"}
{"oral": "I m not totally sure but I think l...learning rate is too high. The l...loss just... it just plateaus at zero point zero five. Maybe we should try a b...batch size of two hundred fifty six instead?", "clean": "I'm not totally sure, but I think learning rate is too high. The loss just plateaus at 0.05. Maybe we should try a batch size of 256 instead?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2333.wav", "scene": "academic", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 527, "source_record_id": "assigned_records.jsonl:2333"}
{"oral": "So I checked the loss curves and it looks like we have some serious gradient vanishing, maybe maybe maybe because the learning rate is too high, like maybe ten to the minus four, or uh, maybe the weight initialization is just off. I tried reducing batch size to one hundred twenty eight but it still not converging. Do you think we should try a different optimizer or maybe just add some batch normalization?", "clean": "I checked the loss curves and it looks like we have some serious gradient vanishing, maybe because the learning rate is too high, maybe 10^-4, or maybe the weight initialization is just off. I tried reducing batch size to 128 but it's still not converging. Do you think we should try a different optimizer or maybe just add some batch normalization?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2322.wav", "scene": "academic", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 525, "source_record_id": "assigned_records.jsonl:2322"}
{"oral": "So for the baseline, I think we should use the ResNet fifty... oh no, it is ResNet one hundred eight, and just like, uh, keep the learning rate at ten to the minus four. Actually... wait, did you see the latest paper from DeepMind on the latent diffusion? I feel like their sampling method might be more efficient for our current bottleneck, so... uh, what do you think about swapping the architecture entirely?", "clean": "For the baseline, I think we should use ResNet 108, and keep the learning rate at 10^-4. Did you see the latest paper from DeepMind on the latent diffusion? I feel like their sampling method might be more efficient for our current bottleneck, so what do you think about swapping the architecture entirely?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2272.wav", "scene": "academic", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 523, "source_record_id": "assigned_records.jsonl:2272"}
{"oral": "So basically the core idea is we want to push similar samples together and push dissimilar ones apart. Um, we do this because... well... if we just use like a cross entropy loss it might... uh... let me rephrase that... it tends to collapse if we don't have enough labels. So we use a contrastive loss with maybe five hundred twelve negative samples to force model learn more robust representations without needing labels for every single image.", "clean": "The core idea is we want to push similar samples together and push dissimilar ones apart. We do this because it tends to collapse if we don't have enough labels. So we use a contrastive loss with maybe 512 negative samples to force the model to learn more robust representations without needing labels for every single image.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2301.wav", "scene": "academic", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 524, "source_record_id": "assigned_records.jsonl:2301"}
{"oral": "I think the batch size is too small. Since... you used eight... I mean, given that you only used eight samples per iteration, it might not converge.", "clean": "I think the batch size is too small. Given that you only used 8 samples per iteration, it might not converge.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2602.wav", "scene": "academic", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 529, "source_record_id": "assigned_records.jsonl:2602"}
{"oral": "P value... zero point zero five... oh no, zero point zero one. Not... uh... significant enough.", "clean": "P-value 0.01. Not significant enough.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2905.wav", "scene": "academic", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 533, "source_record_id": "assigned_records.jsonl:2905"}
{"oral": "Need new metric, new metric. F-one score... no, BLEU score fails on ten percent accuracy.", "clean": "Need new metric. BLEU score fails on 10% accuracy.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2917.wav", "scene": "academic", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 534, "source_record_id": "assigned_records.jsonl:2917"}
{"oral": "Thanks for talk. I think the gradient clipping... clipping might be key here. The threshold was ten... no, twenty... wait wrong, it was fifty! Fifty! I wonder if fifty fifty is optimal for convergence, so what do you think about this specific value?", "clean": "Thanks for the talk. I think the gradient clipping might be key here. The threshold was 50! I wonder if 50 is optimal for convergence, so what do you think about this specific value?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2807.wav", "scene": "academic", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 532, "source_record_id": "assigned_records.jsonl:2807"}
{"oral": "So like if we look at the baseline we got about eighty four point two percent accuracy but the memory footprint is just too high for the edge device. I mean we tried increasing the hidden dimension to five hundred twelve and it just... well maybe we should say it didn't scale linearly. Um so I think we should... uh... if we just... actually let's just swap the whole MLP block for a LoRA adapter to keep it lightweight.", "clean": "If we look at the baseline, we got about 84.2% accuracy, but the memory footprint is just too high for the edge device. We tried increasing the hidden dimension to 512 and it didn't scale linearly. I think we should swap the whole MLP block for a LoRA adapter to keep it lightweight.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2775.wav", "scene": "academic", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 530, "source_record_id": "assigned_records.jsonl:2775"}
{"oral": "So Vaswani said it is essentially a weighted average. But my take is... my take is it is more like... well, a soft lookup table. We set the head count to eight... no, twelve... wait wrong, it is sixteen! It is just... how to say... if we treat queries as keys... if we treat queries as keys, we get this dynamic routing, yeah.", "clean": "Vaswani said it is essentially a weighted average. But my take is it is more like a soft lookup table. We set the head count to 16! If we treat queries as keys, we get this dynamic routing, yeah.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2776.wav", "scene": "academic", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 531, "source_record_id": "assigned_records.jsonl:2776"}
{"oral": "Basically not significant, yeah. p-value is... p-value is zero point zero six, which just misses threshold. I think we need more samples, more samples, maybe five hundred instead of three hundred, to really confirm if gain is real, you know?", "clean": "Not significant. p-value is 0.06, which just misses threshold. I think we need more samples, maybe 500 instead of 300, to really confirm if gain is real.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2945.wav", "scene": "academic", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 535, "source_record_id": "assigned_records.jsonl:2945"}
{"oral": "Honestly, contrastive learning is basically just... uh, pushing positive pairs together and pulling negatives apart in the embedding space. So, like, if we take a batch of five hundred twelve samples, we use a contrastive loss, maybe NT-Xent, to ensure the augmented views of the same image are close. It is sort of... wait, actually, the key is the temperature hyperparameter, maybe zero point zero seven, which controls how sharply the model penalizes hard negatives. It can be tricky to tune, but that is essentially how we get those robust representations without labels.", "clean": "Honestly, contrastive learning is just pushing positive pairs together and pulling negatives apart in the embedding space. So, if we take a batch of 512 samples, we use a contrastive loss, maybe NT-Xent, to ensure the augmented views of the same image are close. The key is the temperature hyperparameter, maybe 0.07, which controls how sharply the model penalizes hard negatives. It can be tricky to tune, but that is essentially how we get those robust representations without labels.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl2525.wav", "scene": "academic", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 528, "source_record_id": "assigned_records.jsonl:2525"}
{"oral": "I mean, if we just, uh, stick with the current transformer block and maybe, uh, scale the hidden dimension to five hundred twelve... oh no, it was seven hundred sixty eight, then we could probably, you know, mitigate some of that gradient vanishing issue without needing to, like, re-initialize the whole weights matrix, right?", "clean": "If we just stick with the current transformer block and maybe scale the hidden dimension to 768, then we could probably mitigate some of that gradient vanishing issue without needing to re-initialize the whole weights matrix, right?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl3016.wav", "scene": "academic", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 536, "source_record_id": "assigned_records.jsonl:3016"}
{"oral": "So, the reviewer said the baseline is too weak, but, uh, I think the p-value of zero point zero five is sufficient for this proof of concept, right?", "clean": "The reviewer said the baseline is too weak, but I think the p-value of 0.05 is sufficient for this proof of concept, right?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl3299.wav", "scene": "academic", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 541, "source_record_id": "assigned_records.jsonl:3299"}
{"oral": "So looking at the ablation, we saw the FID score dropped to twelve point four, which is good, and then, uh, we tried adding stochastic depth for vision transformers but it didn't really help with out of distribution robustness in medical imaging, plus we checked the FLOPs, just to see if it scaled like Kaplan et al. twenty twenty, but it's just... how should I put it... not quite there yet.", "clean": "Looking at the ablation, we saw the FID score dropped to 12.4, which is good, and then we tried adding stochastic depth for Vision Transformers but it didn't really help with out-of-distribution robustness in medical imaging, plus we checked the FLOPs, just to see if it scaled like Kaplan et al. 2020, but it's just not quite there yet.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl306.wav", "scene": "academic", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 537, "source_record_id": "assigned_records.jsonl:306"}
{"oral": "Honestly, I think we should use a transformer, a transformer for this. The latency is... oh no, it is not twelve milliseconds, it is fifteen milliseconds. we need to optimize the weights, optimize the weights in that new garden... I mean, the new Garden project.", "clean": "Honestly, I think we should use a transformer for this. The latency is 15 milliseconds. We need to optimize the weights in the new Garden project.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl3086.wav", "scene": "academic", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 539, "source_record_id": "assigned_records.jsonl:3086"}
{"oral": "Aha! So contrastive learning basically pulls positive pairs closer, pulls positive pairs closer in latent space, while pushing negatives apart. We usually use a batch size of two hundred fifty six... oh no, five hundred twelve, to stabilize gradients, just to stabilize gradients, you know?", "clean": "Contrastive learning pulls positive pairs closer in latent space, while pushing negatives apart. We usually use a batch size of 512 to stabilize gradients.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl3280.wav", "scene": "academic", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 540, "source_record_id": "assigned_records.jsonl:3280"}
{"oral": "Hmm, first the baseline performance... actually, what I wanted to ask is, then you mentioned you mentioned the learning rate being zero point zero zero one, how does that affect the convergence? And I think, how to say, the loss curve loss curve looks a bit noisy, so finally, did you try any other optimizers, like AdamW, to see if it helps stabilize the training process, ya?", "clean": "You mentioned the learning rate being 0.001, how does that affect the convergence? The loss curve looks a bit noisy, so finally, did you try any other optimizers, like AdamW, to see if it helps stabilize the training process?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl3070.wav", "scene": "academic", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 538, "source_record_id": "assigned_records.jsonl:3070"}
{"oral": "Wait! So the current F-one score is... uh... it is twenty-two percent... oh no, twenty-eight point five percent. Just, we need... I mean, if we integrate this new metric in the... in the new system, it will be... how to say... more robust for long-tail distributions, ha.", "clean": "The current F1 score is 28.5%. If we integrate this new metric in the new system, it will be more robust for long-tail distributions.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl3324.wav", "scene": "academic", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 542, "source_record_id": "assigned_records.jsonl:3324"}
{"oral": "So we started by using contrastive learning for the initial representation, but then we tried to, uh, sort of... actually we shifted to a curriculum learning approach to stabilize training. We're seeing perplexity drop to about four point two, which is great, but we're still trying to, how should I put it, optimize the knowledge distillation process from PaLM two. We used a teacher model with five hundred forty four billion parameters and... well, we're basically compressing it into a smaller student, you know, right?", "clean": "So we started by using contrastive learning for the initial representation, but then we shifted to a curriculum learning approach to stabilize training. We're seeing perplexity drop to about 4.2, which is great, but we're still trying to optimize the knowledge distillation process from PaLM 2. We used a teacher model with 544 billion parameters and we're compressing it into a smaller student, right?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl405.wav", "scene": "academic", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 543, "source_record_id": "assigned_records.jsonl:405"}
{"oral": "Wait! So we're seeing huge bottleneck in inference latency for quantized weights, and honestly, just implementing Grouped Query Attention... oh no, I mean group query attention... is helping, but we need better metric. Like, how do we measure tokens per second when we're dealing with, how should I put it, effective data curation for synthetic pretraining? I think we can hit eighty five tokens per second, maybe ninety two... no, wait, ninety two tokens per second, if we optimize the cache, right?", "clean": "Wait! So we're seeing a huge bottleneck in inference latency for quantized weights, and honestly, just implementing Grouped Query Attention is helping, but we need a better metric. How do we measure tokens per second when we're dealing with effective data curation for synthetic pretraining? I think we can hit 85 tokens per second, maybe 92 tokens per second, if we optimize the cache, right?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl699.wav", "scene": "academic", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 549, "source_record_id": "assigned_records.jsonl:699"}
{"oral": "So I been thinking about this... I'm not totally sure but I think we could use Grok-one as base model. Maybe apply Curriculum Learning, apply Curriculum Learning to stabilize training. For the alignment part, we might see RLHF reward hacking in reasoning tasks. I remember the paper said reward hacking occurs at... ten percent... no, twenty percent... wait wrong, is thirty percent of samples! Then we could monitor KL divergence to keep it tight. I was looking at DALL-E three... no, DALL-E two... hold on wrong, is DALL-E three for some multimodal intuition. We need to solve solve this stability issue.", "clean": "So I've been thinking about this... I'm not totally sure but I think we could use Grok-1 as a base model. Maybe apply Curriculum Learning to stabilize training. For the alignment part, we might see RLHF reward hacking in reasoning tasks. I remember the paper said reward hacking occurs at 30% of samples! Then we could monitor KL divergence to keep it tight. I was looking at DALL-E 3 for some multimodal intuition. We need to solve this stability issue.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl541.wav", "scene": "academic", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 545, "source_record_id": "assigned_records.jsonl:541"}
{"oral": "Actually, people think DeepSeek-V3 just scales tokens per second through raw compute, but it's more about the architectural efficiency. The real key is how they handled gradient descent stability in deep networks, which, uh, I mean, it's similar to the early insights from Radford et al. twenty nineteen but applied to a much larger reward model. It's like... wait, the actual throughput is more like... what was it... maybe five hundred tokens per second per user, but the stability part is what actually prevents the divergence.", "clean": "Actually, people think DeepSeek-V3 just scales tokens per second through raw compute, but it's more about the architectural efficiency. The real key is how they handled gradient descent stability in deep networks, which is similar to the early insights from Radford et al. 2019 but applied to a much larger reward model. The actual throughput is maybe 500 tokens per second per user, but the stability part is what actually prevents the divergence.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl512.wav", "scene": "academic", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 544, "source_record_id": "assigned_records.jsonl:512"}
{"oral": "So, how does our current baseline actually stack up against the literature, right? Well, just... if we look at the classic Sutskever et al. twenty fourteen approach, it's obviously foundational, but we're seeing a bit of a gap in long context handling. I mean, we tried implementing a sliding window attention, but the Recall at ten is still... uh, it's not quite where we want it, maybe? Like, if you compare it to Wei et al. twenty twenty two, they're leveraging chain of thought, which is great, but we're trying to optimize via mixture of experts. So, just... let me think, we might need to tweak the gating mechanism to actually hit those higher recall numbers, ya know?", "clean": "So, how does our current baseline actually stack up against the literature, right? Well, if we look at the classic Sutskever et al. 2014 approach, it's obviously foundational, but we're seeing a bit of a gap in long context handling. We tried implementing a sliding window attention, but the Recall@10 is still not quite where we want it, maybe? If you compare it to Wei et al. 2022, they're leveraging chain of thought, which is great, but we're trying to optimize via mixture of experts. So, we might need to tweak the gating mechanism to actually hit those higher recall numbers.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl595.wav", "scene": "academic", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 547, "source_record_id": "assigned_records.jsonl:595"}
{"oral": "Sarah, I think we need to double check the p values. The contrastive learning for self supervised vision... the contrastive learning for self supervised vision results look promising but inference latency is way too high. It's like... how should I put it... if we integrate stochastic gradient Langevin dynamics for Bayesian neural nets it might slow down everything. We saw GPT-four o hit eighty-eight point two percent on MMLU... eighty-eight point two percent on MMLU which is great but our current throughput is only twelve frames per second. I mean it's only twelve frames per second.", "clean": "Sarah, I think we need to double check the p-values. The contrastive learning for self-supervised vision results look promising but inference latency is way too high. If we integrate Stochastic Gradient Langevin Dynamics for Bayesian neural nets it might slow down everything. We saw GPT-4o hit 88.2% on MMLU, which is great but our current throughput is only 12 frames per second.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl557.wav", "scene": "academic", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 546, "source_record_id": "assigned_records.jsonl:557"}
{"oral": "So, uh, I remember Sarah mentioned, I think, that we should focus on knowledge distillation from frontier models to SLMs to save on compute, right? But, how should I put it, I'm seeing some weird... uh... interference between multitasking heads when we scale down. Like, the Recall at K is looking okay, but the ROC AUC is just... it's dipping, maybe? I mean, if we look at Hoffmann et al. twenty twenty two, the scaling laws suggest we need more tokens, but, you know, we're capped. So yeah, I think we need to re-evaluate the head architecture, maybe?", "clean": "I remember Sarah mentioned that we should focus on knowledge distillation from frontier models to SLMs to save on compute, right? But I'm seeing some weird interference between multitasking heads when we scale down. The Recall@K is looking okay, but the ROC AUC is dipping. If we look at Hoffmann et al. 2022, the scaling laws suggest we need more tokens, but we're capped. So I think we need to re-evaluate the head architecture.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl621.wav", "scene": "academic", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 548, "source_record_id": "assigned_records.jsonl:621"}
{"oral": "So, uh, the PI mentioned that we should focus on the emergence of reasoning in large scale models, like, specifically looking at how MCTS, you know, Monte Carlo Tree Search, might act as a verifier. But, well, my take is that we are seeing too much instability. I mean, the perplexity is still spiking, so I tried implementing Flash Attention to speed up the iterations, and maybe, let me think, stochastic weight averaging for the final ensemble. It should stabilize the loss, right? Just, like, it is a bit tricky with the batch size being two thousand forty eight.", "clean": "The PI mentioned that we should focus on the emergence of reasoning in large scale models, specifically looking at how MCTS (Monte Carlo Tree Search) might act as a verifier. But my take is that we are seeing too much instability. The perplexity is still spiking, so I tried implementing FlashAttention to speed up the iterations, and stochastic weight averaging for the final ensemble. It should stabilize the loss, right? It is a bit tricky with the batch size being 2,048.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl72.wav", "scene": "academic", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 550, "source_record_id": "assigned_records.jsonl:72"}
{"oral": "So uh I'm thinking we could implement quantization aware training for four bit weights to shrink the footprint and then also do some knowledge distillation from proprietary to open models. Plus I want to integrate FlashAttention two... oh no, FlashAttention three for the speedup, and maybe use Grouped Query Attention to keep KV cache manageable. We can evaluate quality using FID score, probably around... uh... twenty four point five, I think.", "clean": "I'm thinking we could implement quantization aware training for 4-bit weights to shrink the footprint and then also do some knowledge distillation from proprietary to open models. Plus I want to integrate FlashAttention 3 for the speedup, and maybe use Grouped Query Attention to keep KV cache manageable. We can evaluate quality using FID score, probably around 24.5, I think.", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl874.wav", "scene": "academic", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 552, "source_record_id": "assigned_records.jsonl:874"}
{"oral": "喂，那个设计方案得改改，客厅颜色太深太深了，还有那个地址地址是成都市武侯区锦绣路十号，啊不对，是锦绣路十一号，然后我想问下分期乐分期乐能不能走，另外我顺丰单号一零二五八七三六四二五零一四七发你，你查查收没收到。", "clean": "设计方案得改改，客厅颜色太深了，还有地址是成都市武侯区锦绣路11号，然后我想问下分期乐能不能走，另外我顺丰单号102587364250147发你，你查查收没收到。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3388.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 554, "source_record_id": "assigned_records.jsonl:3388"}
{"oral": "喂物业吗，如果你们现在能赶紧过来处理的话，那我们今晚就能睡个好觉，我住在武汉市江汉区建设大道五十五号，啊不对，是建设大道五十号光辉大厦，那个邻居一直吵，而且……怎么说呢……如果你们不能快速解决，我就得请求升级版本，直接找社区投诉了！", "clean": "喂物业吗，如果你们现在能赶紧过来处理的话，那我们今晚就能睡个好觉，我住在武汉市江汉区建设大道50号光辉大厦，邻居一直吵，而且如果你们不能快速解决，我就得请求升级版本，直接找社区投诉了！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3398.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 555, "source_record_id": "assigned_records.jsonl:3398"}
{"oral": "So I was thinking we might need a new metric to track alignment stability in RLHF pipelines because we are seeing some weird variance. Like if we look at Grok-one it is huge so the total FLOPS are just... uh... massive. I wonder if we can leverage stochastic depth to regularize this or maybe account for the attention sink phenomenon in streaming LLMs to keep the KV cache stable. I mean it is basically like... uh... let me rephrase... it is about maintaining distribution shift over time. Do you think that would actually capture the drift or is it too noisy?", "clean": "So I was thinking we might need a new metric to track alignment stability in RLHF pipelines because we are seeing some weird variance. If we look at Grok-1 it is huge so the total FLOPS are just massive. I wonder if we can leverage stochastic depth to regularize this or maybe account for the attention sink phenomenon in streaming LLMs to keep the KV cache stable. It is about maintaining distribution shift over time. Do you think that would actually capture the drift or is it too noisy?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl78.wav", "scene": "academic", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 551, "source_record_id": "assigned_records.jsonl:78"}
{"oral": "诶，你说这携号转网能不能快点办完？我想想啊，我得先查询账单详情，查询账单详情看看还欠多少钱。地址是西安市雁塔区曲江新区一百零二号，西安市雁塔区曲江新区一百零二号，你们能帮我预约上门办吗？我想想，应该能预约上门吧。", "clean": "你说这携号转网能不能快点办完？我得先查询账单详情，看看还欠多少钱。地址是西安市雁塔区曲江新区102号，你们能帮我预约上门办吗？应该能预约上门吧。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3406.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 556, "source_record_id": "assigned_records.jsonl:3406"}
{"oral": "Wow! So you mentioned that the model outperformed CLIP, uh, contrastive language image pretraining, on MMLU and HumanEval. But, uh, if we look at... actually, let me rephrase... if we consider the convergence rates of AdamW in high-dimensional landscapes, it seems... I mean... the way the loss curves behave... it feels like the HumanEval results might be biased. Do you think the optimizer's behavior in those high dimensions affects the generalizability across those two benchmarks?", "clean": "Wow! So you mentioned that the model outperformed CLIP, Contrastive Language-Image Pre-training, on MMLU and HumanEval. But if we consider the convergence rates of AdamW in high-dimensional landscapes, it seems the way the loss curves behave... it feels like the HumanEval results might be biased. Do you think the optimizer's behavior in those high dimensions affects the generalizability across those two benchmarks?", "usage_text_words": null, "audio": "audio/academic__assigned_recordsjsonl890.wav", "scene": "academic", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 553, "source_record_id": "assigned_records.jsonl:890"}
{"oral": "我这屋里水管突然爆了，水漫全地，诶，快帮我报修，地址是西安市雁塔区雁塔路二十号，二十号哈，快点来。", "clean": "我这屋里水管突然爆了，水漫全地，快帮我报修，地址是西安市雁塔区雁塔路20号，快点来。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3685.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 560, "source_record_id": "assigned_records.jsonl:3685"}
{"oral": "诶，我这外卖快超时一小时哈！就是我下单地址填成苏州工业园区金风路九十九号，不对，是青岛市市南区香港中路二十号，你们帮我看看怎么回事，单号P五三二七六一八二四三零六九，快给我查查怎么还没送到！", "clean": "我这外卖快超时1小时哈！我下单地址填成青岛市市南区香港中路20号，你们帮我看看怎么回事，单号P5327618243069，快给我查查怎么还没送到！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3418.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 557, "source_record_id": "assigned_records.jsonl:3418"}
{"oral": "哎呀真是没法住了！那个住在西安市雁塔区曲江新区大雁塔北路八十八号的邻居，天天在那儿吵，就在那儿吵！我真的受不了了，你们赶紧管管，就在大雁塔北路八十八号，快点来处理，再不来我得直接报警了！", "clean": "真是没法住了！住在西安市雁塔区曲江新区大雁塔北路88号的邻居，天天在那儿吵！我真的受不了了，你们赶紧管管，就在大雁塔北路88号，快点来处理，再不来我得直接报警了！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3469.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 558, "source_record_id": "assigned_records.jsonl:3469"}
{"oral": "喂你好，我想查查快递，与其一直等不如直接问你们，我这个包裹比急件更急，快快快帮我查查。地址地址是郑州市金水区花园路两百号，没到吗？我那个航空电子客票号十三位数字，是七零三四五七六八九零零幺五，票号票号七零三四五七六八九零零幺五，快帮我看看现在在哪！", "clean": "喂你好，我想查查快递，与其一直等不如直接问你们，我这个包裹比急件更急，快帮我查查。地址是郑州市金水区花园路200号，没到吗？我那个航空电子客票号13位数字，是7034576890015，票号7034576890015，快帮我看看现在在哪！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3572.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 559, "source_record_id": "assigned_records.jsonl:3572"}
{"oral": "不是说用信用分就行，其实得用权益兑换券，权益兑换券。我的兑换码是G9SPB37BZON4，这个码这个码得手动输。另外配送地址是杭州市滨江区西兴路三百号，三百号，送到这个道。我的信用卡号是6238509130843821，那个卡号麻烦再确认下。", "clean": "不是说用信用分就行，其实得用权益兑换券。我的兑换码是G9SPB37BZON4，这个码得手动输。另外配送地址是杭州市滨江区西兴路300号，送到这个道。我的信用卡号是6238509130843821，卡号麻烦再确认下。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3700.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 561, "source_record_id": "assigned_records.jsonl:3700"}
{"oral": "嗯，我之前在你们行办了财富通，财富通存了不少钱，我想想啊，我想升级升级会员等级哈。就是我想办那个年金险，年金险得怎么办理呀？诶，我得把地址核对下，地址是，地址是昆明市五华区翠竹路十五号。对了，我再充个电信充值卡，卡号是，19835198635198463543，帮我看看能不能办到，能不能办道。", "clean": "我之前在你们行办了财富通，存了不少钱，我想升级会员等级。我想办年金险，年金险怎么办理呀？我得把地址核对下，地址是昆明市五华区翠竹路15号。对了，我再充个电信充值卡，卡号是19835198635198463543，帮我看看能不能办到。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3702.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 562, "source_record_id": "assigned_records.jsonl:3702"}
{"oral": "诶，如果我能把那个办理信用卡注销办完，再把分期乐处理好，那我们就能在周六上午十点到店，我的电信宽带账号是519360827413，我在郑州市金水区花园道一百三十二号，我想想啊，如果能在那个…要是能在周六上午把保养做了，我就能准时出发了哈。", "clean": "如果我能把办理信用卡注销办完，再把分期乐处理好，那我们就能在周六10:00到店，我的电信宽带账号是519360827413，我在郑州市金水区花园道132号，如果能在周六上午把保养做了，我就能准时出发了哈。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3752.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 565, "source_record_id": "assigned_records.jsonl:3752"}
{"oral": "诶，对了，我想问问办理进度，查询办理进度，怎么说呢，我之前在西安市雁塔区曲江新区雁塔路一百号提交申请，那个申请…我想想啊，我申请请求升级账户，升级账户的事儿怎么还没弄好呀？哦对了，回单号是BRCTVPXQWYZSS，回单号BRCTVPXQWYZSS，麻烦帮我查查哈。", "clean": "对了，我想问问办理进度，我之前在西安市雁塔区曲江新区雁塔路100号提交申请，我申请升级账户的事儿怎么还没弄好呀？回单号是BRCTVPXQWYZSS，麻烦帮我查查哈。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3707.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 563, "source_record_id": "assigned_records.jsonl:3707"}
{"oral": "喂物业啊，我这情况是这样，我家里水管漏水漏得厉害，我得赶紧处理，那个地址在成都市武侯区人民南路四段二十二号，对，就是二十二号，我想着能不能先把大额存单那个申请开通功能在理财通里操作一下，我想把钱弄出来修水管，然后我这还有个EMS单号，是EMS开头加上382704915628，你帮我看看这事怎么弄，你觉得呢？我想知道现在能赶紧来个人帮我看看吗？那个，再赶紧给我个答复，行吗？", "clean": "喂物业啊，我这情况是这样，我家里水管漏水漏得厉害，我得赶紧处理，地址在成都市武侯区人民南路四段22号，我想着能不能先把大额存单那个申请开通功能在理财通里操作一下，我想把钱弄出来修水管，然后我这还有个EMS单号，是EMS382704915628，你帮我看看这事怎么弄，你觉得呢？我想知道现在能赶紧来个人帮我看看吗？再赶紧给我个答复，行吗？", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3733.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 564, "source_record_id": "assigned_records.jsonl:3733"}
{"oral": "诶，如果你们能把那个云存储空间挪到，我想想啊，要是把那个控制面板移到武汉市江汉区建设大道一百一号那个样板间那样的话，吧，那我们就能省掉很多走线，就是那个订单号二八四六幺九零三七五二八，你们对对一下哈。", "clean": "如果你们能把云存储空间挪到，要是把控制面板移到武汉市江汉区建设大道11号那个样板间那样的话，那我们就能省掉很多走线，订单号284619037528，你们对一下哈。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3904.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 568, "source_record_id": "assigned_records.jsonl:3904"}
{"oral": "诶，我想问问哈，那个闪电贷，闪电贷能不能分期付，就是怎么说呢，我想想啊，我想把那个分期付的额度给调一下。哎不对，我先改地址，地址是郑州市金水区花园路一百号，花园路一百号。然后那个快递单号，EMS开头，后面是零七四二八五幺九三六零七，EMS零七四二八五幺九三六零七。对了，我还有一个沈阳市沈河区青年大街十号，沈阳市沈河区青年大街十号这边的账户，得怎么处理吧？", "clean": "我想问问，闪电贷能不能分期付，我想把分期付的额度给调一下。我先改地址，地址是郑州市金水区花园路100号。然后快递单号是EMS074285193607。对了，我还有一个沈阳市沈河区青年大街10号这边的账户，得怎么处理吧？", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3792.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 566, "source_record_id": "assigned_records.jsonl:3792"}
{"oral": "诶，我想预约个皮肤护理，我皮肤最近起皮很严重，然后我想想，就是那个…皮肤有点泛红，另外我想确认下地址，是沈阳市沈河区青年大街十八号吗？", "clean": "我想预约个皮肤护理，我皮肤最近起皮很严重，皮肤有点泛红，另外我想确认下地址，是沈阳市沈河区青年大街18号吗？", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4194.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 570, "source_record_id": "assigned_records.jsonl:4194"}
{"oral": "诶我的外卖超时太久了！就是我住武汉市江汉区建设大道五百六十八号，等了快一个小时没到。我想用那个优惠券码98CSHIDUB3，还有面单号S四零七二九五一六三八四零，你们赶紧帮我查查怎么回事呀！", "clean": "我的外卖超时太久了！我住武汉市江汉区建设大道568号，等了快1个小时没到。我想用优惠券码98CSHIDUB3，还有面单号S407295163840，你们赶紧帮我查查怎么回事呀！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4201.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 571, "source_record_id": "assigned_records.jsonl:4201"}
{"oral": "诶，我真的是年费会员哈，本来觉得你们配送快，结果我这单JD七五幺三八零四九二六七五，就是送到郑州市金水区花园路五零号这单，怎么说呢，怎么还没到呀！", "clean": "我真的是年费会员哈，本来觉得你们配送快，结果我这单JD751380492675，送到郑州市金水区花园路50号这单，怎么还没到呀！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4223.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 572, "source_record_id": "assigned_records.jsonl:4223"}
{"oral": "王工，那个方案还得改，我跟你说那个单号JD六二零九四七幺五三八二六，你赶紧查下，还有我那个流量加油包没到账，地址在长沙市芙蓉区五一路十号，你再确认下，顺丰单号是SF九幺五三七零六二八四九幺，银行借记卡号是6223918606801308，你赶紧帮我弄到道，这事儿太急了，赶紧弄到道！", "clean": "王工，方案还得改，我跟你说单号JD620947153826，你赶紧查下。还有我那个流量加油包没到账，地址在长沙市芙蓉区五一路10号，你再确认下。顺丰单号是SF915370628491，银行借记卡号是6223918606801308，你赶紧帮我弄到账，这事儿太急了，赶紧弄到账！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl3844.wav", "scene": "customer_service", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 567, "source_record_id": "assigned_records.jsonl:3844"}
{"oral": "喂！我包裹丢了，丢了！你们赶紧给我申请退款申请退款，还有就是我想请求更改套餐，然后另外我想问问会员权益，诶，我地址是广州市天河区珠江新城花城大道六十六号，六十六号！你们怎么回事啊？", "clean": "喂！我包裹丢了！你们赶紧给我申请退款，我想请求更改套餐，另外我想问问会员权益。我地址是广州市天河区珠江新城花城大道66号！你们怎么回事啊？", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4228.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 574, "source_record_id": "assigned_records.jsonl:4228"}
{"oral": "明早 8 点记得把垃圾带下去，然后顺便检查一下阳台的花浇水了没有。", "clean": "明早 8 点记得把垃圾带下去，然后顺便检查一下阳台的花浇水了没有。", "usage_text_words": null, "audio": "audio/passthrough__recheckjsonl677.wav", "scene": "passthrough", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 389, "source_record_id": "recheck.jsonl:677"}
{"oral": "我先是用了快捷支付，快捷支付订单号A9S16E1Z8516，然后转账到沈阳市沈河区青年大街三十号，沈阳市沈河区青年大街三十号，最后发给我两个回单，回单号一个OHUIBHJSCDJIOZHUBRW，另一个回单号CSDIBNECPJIHODSUB，钱转到郑州市金水区花园路一百号，郑州市金水区花园路一百号，你们快帮我查查怎么回事！", "clean": "我先是用了快捷支付，快捷支付订单号A9S16E1Z8516，然后转账到沈阳市沈河区青年大街30号，最后发给我2个回单，回单号一个OHUIBHJSCDJIOZHUBRW，另一个回单号CSDIBNECPJIHODSUB，钱转到郑州市金水区花园路100号，你们快帮我查查怎么回事！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4074.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 569, "source_record_id": "assigned_records.jsonl:4074"}
{"oral": "诶，我情况就 是这样，我信用卡号六二幺七八五零四九三二六零幺八三，地址在长沙市芙蓉区五一广场十号，我想用这个超级会员权益预约挂号，你觉得行吗？", "clean": "我情况是这样，我信用卡号6217 8504 9326 0183，地址在长沙市芙蓉区五一广场10号，我想用这个超级会员权益预约挂号，你觉得行吗？", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4396.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 578, "source_record_id": "assigned_records.jsonl:4396"}
{"oral": "我想问下这个理财，我也不太确定，但是感觉那个风险还是高了点，那个，我想查一下我之前充值那张卡，卡号是六二二二五八一零三三四一九零二八七六，啊不对，是六二二二五八一零三三四一九零二八五六，然后我想确认下，那个，我的订单状态，怎么说呢，要是现在查的话能不能看到，我那个地址写的是长沙市芙蓉区五一广场三十三号，是不是还没发货？", "clean": "我想问下这个理财，我也不太确定，但是感觉风险还是高了点。我想查一下我之前充值那张卡，卡号是622258103341902856，然后我想确认下我的订单状态，要是现在查的话能不能看到？我那个地址写的是长沙市芙蓉区五一广场33号，是不是还没发货？", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4225.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 573, "source_record_id": "assigned_records.jsonl:4225"}
{"oral": "诶，你们物业怎么管的，邻居吵得没法住，就是我之前催单发货催了好几次没反应，还有我那张白金信用卡白金信用卡，卡号是6231 9840 3649 2501，我想核对下超级会员权益，另外地址地址是西安市雁塔区曲江新区大雁塔北路八十八号，你们赶紧帮我处理一下吧！", "clean": "你们物业怎么管的，邻居吵得没法住。我之前催单发货催了好几次没反应，还有我那张白金信用卡，卡号是6231 9840 3649 2501，我想核对下超级会员权益。另外地址是西安市雁塔区曲江新区大雁塔北路88号，你们赶紧帮我处理一下吧！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4311.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 576, "source_record_id": "assigned_records.jsonl:4311"}
{"oral": "前台小姐姐，诶，我想订个房间，地址是，地址是沈阳市沈河区青年大街五十号，啊不对，是重庆市渝中区解放碑步行街一号哈，就是得要个安静点房间，安静点房间，在顶层能行吗？", "clean": "前台小姐姐，我想订个房间，地址是重庆市渝中区解放碑步行街1号，得要个安静点房间，在顶层能行吗？", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4436.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 579, "source_record_id": "assigned_records.jsonl:4436"}
{"oral": "诶，我想说方案里这个颜色，与其用这种深灰，不如换成米白色，米白色会比深灰更亮堂。地址地址是，成都市武侯区锦绣路十号锦绣花园三栋一百零二室，锦绣花园三栋一百零二室，我想想啊，咱们得把，把那个客厅墙面，重新调一下，重新调一下颜色哈。", "clean": "我想说方案里这个颜色，与其用这种深灰，不如换成米白色，米白色会比深灰更亮堂。地址是成都市武侯区锦绣路10号锦绣花园3栋102室。咱们得把客厅墙面重新调一下颜色哈。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4369.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 577, "source_record_id": "assigned_records.jsonl:4369"}
{"oral": "诶，说白了我就是想问问，我这宽带无法上网，宽带无法上网，怎么说呢，我记得我办了年费会员，年费会员应该没到期吧。我想想啊，我账号是100963916503272，100963916503272，地址是，地址在郑州市金水区花园路两百号，不对，是花园路两百号那个小区，哈。然后我卡号是6220916791439865，6220916791439865，你们帮我查查怎么回事嘛。", "clean": "说白了我就是想问问，我这宽带无法上网，我记得我办了年费会员，年费会员应该没到期吧。我账号是100963916503272，地址在郑州市金水区花园路200号那个小区。然后我卡号是6220916791439865，你们帮我查查怎么回事嘛。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4248.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 575, "source_record_id": "assigned_records.jsonl:4248"}
{"oral": "诶，我跟你说，与其让我在这儿干等，不如你赶紧帮我查查怎么回事！比超时更让我生气的是你们根本没反应，我地址就在西安市雁塔区曲江新区新开门南路八十八号，对，八十八号！我刚才还想查账单详情看怎么扣费，结果你们这配送太慢，我想反馈产品质量问题都找不到人，我就想问问这订单怎么回事哈！", "clean": "我跟你说，与其让我在这儿干等，不如你赶紧帮我查查怎么回事！比超时更让我生气的是你们根本没反应，我地址就在西安市雁塔区曲江新区新开门南路88号！我刚才还想查账单详情看怎么扣费，结果你们这配送太慢，我想反馈产品质量问题都找不到人，我就想问问这订单怎么回事哈！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4468.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 580, "source_record_id": "assigned_records.jsonl:4468"}
{"oral": "诶，不是说查询余额不足就没法操作，其实是我想咨询账单详情，就是那个扣费金额不对，一共扣了三百五十块，啊不是，是三百八十块，我质疑这个金额怎么算得，赶紧给我请求人工接听，我想看看到底扣哪再了。", "clean": "不是说查询余额不足就没法操作，其实是我想咨询账单详情，扣费金额不对，一共扣了380元，我质疑这个金额怎么算得，赶紧给我请求人工接听，我想看看到底扣在哪了。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4606.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 582, "source_record_id": "assigned_records.jsonl:4606"}
{"oral": "诶我先是叫了闪送，然后我想想啊，我想把那个…就是把一千条短信特惠包发过去，最后帮我收个韵达快递，单号是二零二四零五一一八三零五二七幺四六九八三哈。", "clean": "我先叫了闪送，我想把1000条短信特惠包发过去，最后帮我收个韵达快递，单号是20240511830527146983。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4790.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 585, "source_record_id": "assigned_records.jsonl:4790"}
{"oral": "哦对了，就是我这单送到北京市海淀区中关村东路一号北大校园内二号楼一百零一室，怎么还没到呀？", "clean": "我这单送到北京市海淀区中关村东路1号北大校园内2号楼101室，怎么还没到呀？", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4888.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 586, "source_record_id": "assigned_records.jsonl:4888"}
{"oral": "柜员，我我我想办，大额存单，一年期，啊不对，三年期，这个，办办办一个。", "clean": "柜员，我想办三年期大额存单，办一个。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4961.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 587, "source_record_id": "assigned_records.jsonl:4961"}
{"oral": "诶，我想想啊，与其让我等得那么久，不如直接帮我把房间升级升级，毕竟我住过很多酒店，安静的房间比地段更重要哈。我想确认下，我的订单地址是，地址是重庆市渝中区解放碑步行街一百零一号，在那儿，帮我看看能不能安排在高层，高层安静点嘛。", "clean": "与其让我等得那么久，不如直接帮我把房间升级，毕竟我住过很多酒店，安静的房间比地段更重要。我想确认下，我的订单地址是重庆市渝中区解放碑步行街101号，帮我看看能不能安排在高层，高层安静点嘛。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4723.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 583, "source_record_id": "assigned_records.jsonl:4723"}
{"oral": "诶，我要报告服务故障，收到的东西坏了坏了！就是我用了优惠券代码9KJSDU07D，结果寄过来破了，快递是中通快递ZTO，单号是ZTO1006297168439078，单号ZTO1006297168439078。我社保卡号是六二幺四八六零九三二七五幺八四零六二九，地址地址是成都市武侯区人民南路四段二十二号哈。", "clean": "我要报告服务故障，收到的东西坏了！我用了优惠券代码9KJSDU07D，结果寄过来破了，快递是中通快递ZTO，单号是ZTO1006297168439078。我社保卡号是6214 8609 3275 1840 629，地址是成都市武侯区人民南路四段22号。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4552.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 581, "source_record_id": "assigned_records.jsonl:4552"}
{"oral": "我那个订单号六二二...不对，六二三...等等错了，是6249070063841！那个...我想说...怎么说呢...把那个广西壮族自治区南宁市青秀区人民路一号政务中心填地址，你觉得行不？", "clean": "我订单号6249070063841，把广西壮族自治区南宁市青秀区人民路1号政务中心填地址，你觉得行不？", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4732.wav", "scene": "customer_service", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 584, "source_record_id": "assigned_records.jsonl:4732"}
{"oral": "物业管家，我那个工商银行个人住房按揭贷款还没办完，哎呀不对，我得赶紧跟你说，我楼上太吵了！那个广东省深圳市南山区粤海街道海天二路科技园一栋三百零二室，这房号你记一下，三百零二室！真的吵死了，快过来处理一下！", "clean": "物业管家，我楼上太吵了！广东省深圳市南山区粤海街道海天二路科技园一栋302室，这房号你记一下，302室！真的吵死了，快过来处理一下！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl4994.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 588, "source_record_id": "assigned_records.jsonl:4994"}
{"oral": "诶，与其等这个超时单不如直接退款吧，比等配送员更省心。地址地址是陕西省西安市雁塔区小寨东路一百号，一百号，我想想啊，单号是圆通单号圆通单号10098562846513，我就想要个说法哈。", "clean": "与其等这个超时单不如直接退款吧，比等配送员更省心。地址是陕西省西安市雁塔区小寨东路100号，单号是圆通单号10098562846513，我就想要个说法哈。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5021.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 589, "source_record_id": "assigned_records.jsonl:5021"}
{"oral": "先是填了江苏省苏州市工业园区金风路十号软件园一栋五百零二室，然后是淘宝订单号，就是最后报了招行卡号六二幺九三八零四七二五幺六九三八零四七哈。", "clean": "先是填了江苏省苏州市工业园区金风路10号软件园1栋502室，然后是淘宝订单号，最后报了招行卡号6219 3804 7251 6938 047哈。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5039.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 590, "source_record_id": "assigned_records.jsonl:5039"}
{"oral": "不是说下周二考，不是说下周二考！我看了通知其实是二零二五年六月一号，地址地址在湖北省武汉市江汉区之孝路十二号，我那个Apple Order ID，那个Apple Order ID发你确认下，赶紧确认下！", "clean": "不是说下周二考！我看了通知其实是2025年6月1日，地址在湖北省武汉市江汉区之孝路12号，我的Apple Order ID发你确认下，赶紧确认下！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5073.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 591, "source_record_id": "assigned_records.jsonl:5073"}
{"oral": "那个我想约个保养，我这车在，不对，我想约后天，地址在山东省青岛市市南区香港中路十号海信广场四楼，那个海信广场四楼，你帮我记在，记再系统里。", "clean": "我想约个保养，我想约后天，地址在山东省青岛市市南区香港中路10号海信广场4楼，你帮我记在系统里。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5158.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 594, "source_record_id": "assigned_records.jsonl:5158"}
{"oral": "那个全家桶宽带加手机融合套餐，与其等你们慢慢查，不如直接看地址，湖北省武汉市江岸区之江路十八号光谷广场C座八楼，啊不对，是之孝路十二号，那个…怎么说呢…比之前那个海口湾大厦更急，赶紧帮我再看下到哪了！", "clean": "全家桶宽带加手机融合套餐，与其等你们慢慢查，不如直接看地址，湖北省武汉市江岸区之孝路12号光谷广场C座8楼，比之前那个海口湾大厦更急，赶紧帮我再看下到哪了！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5099.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 592, "source_record_id": "assigned_records.jsonl:5099"}
{"oral": "我想续租，那个上海市静安区南京西路一千六百九十一号恒隆广场A座十五楼，还有涨价太高，另外那个合同，得改改啊", "clean": "我想续租，上海市静安区南京西路1691号恒隆广场A座15楼，涨价太高，另外合同得改改啊", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5254.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 595, "source_record_id": "assigned_records.jsonl:5254"}
{"oral": "我要办短信特惠包，短信特惠包一千条，还有语音通话时长包，通话时长包五百分钟，然后地址地址是福建省厦门市思明区鹭江道十五号。", "clean": "我要办短信特惠包1000条，还有语音通话时长包500分钟，地址是福建省厦门市思明区鹭江道15号。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5286.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 596, "source_record_id": "assigned_records.jsonl:5286"}
{"oral": "说白了，我想办转网，卡号是招行卡号六二二二，不对，六二二六四幺八三零九六二五七四八幺三六，地址是广东省广州市天河区珠江新城花城大道六十六号哈，快点办吧。", "clean": "说白了，我想办转网，卡号是招行卡号6226 4183 0962 5748 136，地址是广东省广州市天河区珠江新城花城大道66号，快点办吧。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5119.wav", "scene": "customer_service", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 593, "source_record_id": "assigned_records.jsonl:5119"}
{"oral": "哎呀太贵太贵了！两百九十九两百九十九，电信电信，融合宽带，在西湖区西湖区文三路四百五十六号创意产业苑，那个涨租涨租太离谱！", "clean": "太贵了！299元，电信融合宽带，在西湖区文三路456号创意产业苑，涨租太离谱！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5372.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 599, "source_record_id": "assigned_records.jsonl:5372"}
{"oral": "诶，情况就是我订了广西壮族自治区南宁市青秀区人民路二号，不对，一号政务中心，餐都超时半小时嘛，你现在怎么处理哈？", "clean": "我订了广西壮族自治区南宁市青秀区人民路1号政务中心，餐都超时30分钟了，你现在怎么处理哈？", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5655.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 601, "source_record_id": "assigned_records.jsonl:5655"}
{"oral": "不是说邻居装修哈，其实是他在搞那个平安银行财富管理计划，吵死人！我住贵州省贵阳市南明区南明大路八十八号商务中心四零二室，你们赶紧来看看！", "clean": "不是说邻居装修哈，其实是他在搞平安银行财富管理计划，吵死人！我住贵州省贵阳市南明区南明大路88号商务中心402室，你们赶紧来看看！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5349.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 597, "source_record_id": "assigned_records.jsonl:5349"}
{"oral": "哎呀太夸张了！我家里水管漏水漏得厉害，地址在浙江省金华市婺城区金东路一百号，那个…我想说那个水阀门…怎么说呢，好像完全关不掉，快帮我看看，金东路一百号！", "clean": "太夸张了！我家里水管漏水漏得厉害，地址在浙江省金华市婺城区金东路100号，水阀门好像完全关不掉，快帮我看看，金东路100号！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5360.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 598, "source_record_id": "assigned_records.jsonl:5360"}
{"oral": "美团订单号美团订单号是二零二三...不对，二零二四...等等错了，是二零二五七二零九三八四六幺五九零七二八！然后商品破损破损，还有地址...上海市浦东新区陆家嘴环路一千号，另外，另外别发错货！", "clean": "美团订单号是2025720938461590728，然后商品破损，还有地址上海市浦东新区陆家嘴环路1000号，另外别发错货！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5524.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 600, "source_record_id": "assigned_records.jsonl:5524"}
{"oral": "诶，对了，我那个订单，送到上海市静安区南京西路一二三四号恒隆广场B座十五楼一五零八室，怎么还没到呀？就是那个iPad Pro M2十二点九英寸，得赶紧送到，再慢我就得投诉了哈。", "clean": "我那个订单，送到上海市静安区南京西路1234号恒隆广场B座15楼1508室，怎么还没到呀？那个iPad Pro M2 12.9英寸，得赶紧送到，再慢我就得投诉了哈。", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5787.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 603, "source_record_id": "assigned_records.jsonl:5787"}
{"oral": "那个建行卡号...哎呀不对，我得赶紧报修，我家水管漏水了，浙江省宁波市海曙区鼓楼路二十号一百零二室，还有那个虚拟运营商校园特惠卡的事儿，然后另外那个亚马逊订单号，那个订单号是BRPSTX六二幺三八六零五幺四七二九三八六零五一，快帮我看看！", "clean": "我得赶紧报修，我家水管漏水了，浙江省宁波市海曙区鼓楼路20号102室。还有虚拟运营商校园特惠卡的事儿，另外亚马逊订单号是BRPSTX6213860514729386051，快帮我看看！", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5697.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 602, "source_record_id": "assigned_records.jsonl:5697"}
{"oral": "i wanna go Toronto... oh no Vancouver for holiday and then get AirPods Pro... get AirPods Pro plus maybe some MAC makeup... makeup from The Organic Shop... also Malik... Malik said go there", "clean": "I wanna go to Vancouver for holiday and then get AirPods Pro, maybe some makeup from The Organic Shop. Also, Malik said go there.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl10045.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 605, "source_record_id": "assigned_records.jsonl:10045"}
{"oral": "man i got this deal... i got this deal for a trip to Leeds and like maybe Trafalgar Square too. im not totally shore but i think it was bang on... like maybe five hundred pounds for everything. we go to Blue Mountains... we go to the Blue Mountains after... i mean we visit the mountains then hit Popeyes. it was just... just way too cheap.", "clean": "I got this deal for a trip to Leeds and maybe Trafalgar Square too. I'm not totally sure, but I think it was bang on, maybe £500 for everything. We visit the Blue Mountains then hit Popeyes. It was just way too cheap.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl10275.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 607, "source_record_id": "assigned_records.jsonl:10275"}
{"oral": "诶，情况是这样，我京东自营订单号四零七二九五一六三八四零，里面有华为 Mate 60 Pro 月色银，还有耐克 Air Max 270 运动鞋，还有雅诗兰黛小棕瓶精华液 五十毫升，结果现在快递站说找不到了，就是丢了，我银行卡号六二二九七四六幺零三五八二九四七幺六三都付钱了，怎么说呢，这事儿你们得给我个交代吧？", "clean": "情况是这样，我京东自营订单号407295163840，里面有华为 Mate 60 Pro 月色银，还有耐克 Air Max 270 运动鞋，还有雅诗兰黛小棕瓶精华液 50ml，结果现在快递站说找不到了，就是丢了，我银行卡号6229 7461 0358 2947 163都付钱了，这事儿你们得给我个交代吧？", "usage_text_words": null, "audio": "audio/customer_service__assigned_recordsjsonl5829.wav", "scene": "customer_service", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 604, "source_record_id": "assigned_records.jsonl:5829"}
{"oral": "so i usually hit gym late afternoon late afternoon but i was thinking maybe... oh wait i actually gotta spend time coordinating a dog walking schedule coordinating a dog walking schedule for my sister in brixton around five o clock and honestly man i really miss those old dubai workouts old dubai workouts where i used to lift heavy stuff before moving here it was just different you know", "clean": "I usually hit the gym late afternoon, but I actually gotta spend time coordinating a dog walking schedule for my sister in Brixton around 5 o'clock. Honestly man, I really miss those old Dubai workouts where I used to lift heavy stuff before moving here; it was just different.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl10366.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 608, "source_record_id": "assigned_records.jsonl:10366"}
{"oral": "so like around turn of year i was thinking about discussing fitness challenge... discussing fitness challenge with you. if we hit gym after The Daily Grind... then we could maybe grab Subway. it is like... wait... it is like deciding on name for new pet... deciding on name... just takes forever to pick one.", "clean": "Around the turn of the year, I was thinking about discussing a fitness challenge with you. If we hit the gym after The Daily Grind, then we could maybe grab Subway. It is like deciding on a name for a new pet; it just takes forever to pick one.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl10495.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 609, "source_record_id": "assigned_records.jsonl:10495"}
{"oral": "im not totally sure but i think things used to be like... way cheaper back in the day. i remember my dell xps was... no wait it was a dell inspiron... oh not a dell xps actually it was a lenovo. anyway like if i was deciding on a name for a new puppy i'd probably spend more on dog treats than the pup it self. its like i was talking about a confusing dream where everything cost ten thousand dollars and i was just like im down for that... wait i mean im not down for that at all. it's just crazy how much pricess have gone up", "clean": "I'm not totally sure, but I think things used to be way cheaper back in the day. I remember it was a Lenovo. Anyway, if I was deciding on a name for a new puppy, I'd probably spend more on dog treats than the pup itself. It's like I was talking about a confusing dream where everything cost $10,000 and I was just like I'm not down for that at all. It's just crazy how much prices have gone up", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl10141.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 606, "source_record_id": "assigned_records.jsonl:10141"}
{"oral": "like he was... uh oh right he asks if i fancy a trip to the cotswolds planning a visit to a historic castle and it ten minutes past eight he start complaining bout typo in professional document like seriously who does that", "clean": "He asks if I fancy a trip to the Cotswolds, planning a visit to a historic castle. At 8:10, he started complaining about a typo in a professional document. Like seriously, who does that?", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl10745.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 612, "source_record_id": "assigned_records.jsonl:10745"}
{"oral": "man those neighbors are crazy. im not totally sure but i think they sharing a story about a lucky charm... like like talking way too loud. i was just vibing and deciding on a workout split... deciding on a workout split and then boom. they ghosted my request to be quiet. we planning a meal prep sunday... planning a meal prep sunday but i cant even focus.", "clean": "Man, those neighbors are crazy. I'm not totally sure, but I think they're sharing a story about a lucky charm, talking way too loud. I was just vibing and deciding on a workout split, and then boom. They ghosted my request to be quiet. We're planning a meal prep Sunday, but I can't even focus.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl10655.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 610, "source_record_id": "assigned_records.jsonl:10655"}
{"oral": "um basically just stayed home mostly but like uh we went to Alexanderplatz and then Tesco Express i think... i was like about five minutes late meetin them and honestly i felt kinda dizzy... we spent like hours just uh debating best type of pasta shape like which one holds sauce better you know... so yeah mostly just that", "clean": "Just stayed home mostly, but we went to Alexanderplatz and then Tesco Express I think. I was about 5 minutes late meeting them and honestly I felt dizzy. We spent hours just debating the best type of pasta shape, like which one holds sauce better. So yeah, mostly just that.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl10739.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 611, "source_record_id": "assigned_records.jsonl:10739"}
{"oral": "um people think Camden has bad clinics but actually they're great... uh we were discussing a favorite childhood cartoon like... wait anyway you oughta see Hana. she used to work in Cape Town and she's like the best dentist around here", "clean": "People think Camden has bad clinics but actually they're great. Anyway, you oughta see Hana. She used to work in Cape Town and she's the best dentist around here", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl10819.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 613, "source_record_id": "assigned_records.jsonl:10819"}
{"oral": "you think he cares? i mean... uh... honestly no. like the day before yesterday we were discussing a new workout app and he just... he started complaining about misleading advertisement for Pasta Palace or something about the Maasai Mara. just totally ignored you.", "clean": "You think he cares? Honestly no. The day before yesterday we were discussing a new workout app and he just started complaining about misleading advertisement for Pasta Palace or something about the Maasai Mara. Just totally ignored you.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl11030.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 615, "source_record_id": "assigned_records.jsonl:11030"}
{"oral": "um Sofia she... she pissed cause we skint", "clean": "Sofia is pissed cause we're skint", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl11645.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 620, "source_record_id": "assigned_records.jsonl:11645"}
{"oral": "dreamt I was in Paris... no London... wait wrong it's Prague!", "clean": "Dreamt I was in Prague!", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl11478.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 619, "source_record_id": "assigned_records.jsonl:11478"}
{"oral": "i mean like you gotta stop complaining about slow checkout line at Pret A Manger and then... i dont know... maybe just... like tell Liam he is being kinda cringe also you could just buy him a Nintendo Switch plus... uh... you should just... wait... maybe just try talking it out first", "clean": "You gotta stop complaining about the slow checkout line at Pret A Manger and then maybe just tell Liam he is being cringe, also you could just buy him a Nintendo Switch, but maybe just try talking it out first", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl11361.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 618, "source_record_id": "assigned_records.jsonl:11361"}
{"oral": "um if Lachlan wants that movie about Mrs Rossi talking bout stressful family dinner in Leeds then we could watch the Amalfi Coast one instead... it's like a way better choice for tonight", "clean": "If Lachlan wants that movie about Mrs. Rossi talking about stressful family dinner in Leeds, then we could watch the Amalfi Coast one instead... it's a way better choice for tonight", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl11669.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 621, "source_record_id": "assigned_records.jsonl:11669"}
{"oral": "Actually oh by the way I was like just v...venting about this passive aggressive coworker and then I start thinking bout planning trip to Christmas market maybe Iceland? Like the Blue Lagoon and stuff but man everything is just so expensive now like imagine paying five hundred dollars for some flight. I mean I was debating train versus plane but trains are like so slow. It's almost as much as like a ticket to the Uffizi museum if you go during peak season. Just... it's just too much.", "clean": "By the way, I was venting about this passive aggressive coworker and then I start thinking about planning a trip to a Christmas market, maybe Iceland? The Blue Lagoon and stuff, but man everything is just so expensive now, imagine paying $500 for some flight. I was debating train versus plane, but trains are so slow. It's almost as much as a ticket to the Uffizi museum if you go during peak season. It's just too much.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl10867.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 614, "source_record_id": "assigned_records.jsonl:10867"}
{"oral": "living here is more like paying for luxury than just survival. i mean i am keen keen on shopping at Aldi. cost is... cost is ten dollars... no twenty... wait wrong it is thirty dollars for basic stuff. anyway you should see Dr Nguyen. oh and that movie plot hole... plot hole is just wild.", "clean": "Living here is more like paying for luxury than just survival. I am keen on shopping at Aldi. Cost is $30 for basic stuff. Anyway, you should see Dr. Nguyen. Oh, and that movie plot hole is just wild.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl11257.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 616, "source_record_id": "assigned_records.jsonl:11257"}
{"oral": "omg you gotta hear this like uh first i was discussing a viral TikTok trend with Jax about ten minutes ago and he told me bout this place... then we decided we should go to the Getty Museum... oh no i mean the Getty Center since it is right there and we can see the view. then finally we thought maybe we can like... planning a visit to a planetarium after that just to make it a whole day trip. honestly this restaurant he mentioned sounds like it is insane like just totally out of this world and i really wanna try it tomorrow maybe... yeah it sounds like a plan", "clean": "OMG you gotta hear this. First, I was discussing a viral TikTok trend with Jax about 10 minutes ago and he told me about this place. Then we decided we should go to the Getty Center since it is right there and we can see the view. Then finally we thought maybe we can plan a visit to a planetarium after that just to make it a whole day trip. Honestly this restaurant he mentioned sounds like it is insane, just totally out of this world, and I really wanna try it tomorrow maybe. Yeah, it sounds like a plan.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl11305.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 617, "source_record_id": "assigned_records.jsonl:11305"}
{"oral": "hey mom... yeah i just got home like... roughly ten minutes ago actually. so i was talking to kenji about that recent trip to yellow stone... oh wait no i mean yosemite national park. anyway he is still like... totally stressing over his new office in SOHO. and uh... did you hear about the kids? they were literally just arguing about whose turn it is to do laundry... like for an hour. honestly if they just made a schedule then we could probably avoid all this drama next time.", "clean": "Hey mom, yeah I just got home roughly 10 minutes ago. So I was talking to Kenji about that recent trip to Yosemite National Park. Anyway, he is still totally stressing over his new office in SoHo. And did you hear about the kids? They were literally just arguing about whose turn it is to do laundry for an hour. Honestly, if they just made a schedule then we could probably avoid all this drama next time.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl6389.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 622, "source_record_id": "assigned_records.jsonl:6389"}
{"oral": "Honestly it was like such a mess. I spent like half Saturday asking for refund on faulty charger and then uh asking for refund for damaged delivery from Instacart. Oh and I had to call Soren... wait no sorry I mean Sarah to like book table for baby shower. Just total chaos really.", "clean": "Honestly it was such a mess. I spent half Saturday asking for a refund on a faulty charger and then asking for a refund for a damaged delivery from Instacart. Oh and I had to call Sarah to book a table for a baby shower. Just total chaos really.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl6606.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 624, "source_record_id": "assigned_records.jsonl:6606"}
{"oral": "so i read this book right and honestly it just flipped my whole world upside down. like if i had met Catherine and Yasmine and Mei Ling back then then we could have started that business together way sooner. it was just... how do i put it... just eye opening. i even told Amara and Sephora about it and they read it in like two days. it really changed how i see things, ya know?", "clean": "So I read this book, and honestly it just flipped my whole world upside down. If I had met Catherine, Yasmine, and Mei Ling back then, then we could have started that business together way sooner. It was just eye opening. I even told Amara and Sephora about it and they read it in 2 days. It really changed how I see things.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl6758.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 625, "source_record_id": "assigned_records.jsonl:6758"}
{"oral": "so like who is actually coming to the thing? oh wait i think it is just ananya and priya and alastair for now. um we should probably... i mean if we... how should i put it... if we try to do something like the louvre theme it might be too much. and uh we still gotta figure out how we're trying to split a complicated dinner bill from last time because that was like forty five dollars each but someone paid the whole thing and i think there was some sort of miss understanding with the total. wait did i say the total? i mean the amount. anyway let me know if that works for you guys.", "clean": "Who is actually coming to the thing? I think it is just Ananya and Priya and Alastair for now. We should probably... if we try to do something like the Louvre theme it might be too much. And we still gotta figure out how we're trying to split a complicated dinner bill from last time because that was $45 each but someone paid the whole thing and I think there was some sort of misunderstanding with the amount. Anyway let me know if that works for you guys.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl6449.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 623, "source_record_id": "assigned_records.jsonl:6449"}
{"oral": "Dude I had like the weirdest dream last night. Just totally random. So like I'm in this car with Mei Lin and it's a Tesla but we're just stuck in traffic and she starts venting about a slow internet connection, which is just weird since we're in a car. Then suddenly we're at Five Guys eating burgers but for some reason we're playing Nintendo on the table. I mean, how does that even happen? It was just... yeah, super trippy.", "clean": "Dude I had the weirdest dream last night. Just totally random. So I'm in this car with Mei Lin and it's a Tesla but we're just stuck in traffic and she starts venting about a slow internet connection, which is just weird since we're in a car. Then suddenly we're at Five Guys eating burgers but for some reason we're playing Nintendo on the table. How does that even happen? It was just super trippy.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl6780.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 626, "source_record_id": "assigned_records.jsonl:6780"}
{"oral": "so i was talking to yara and she said she tried booking a table for brunch at that place in chelsea... oh wait no kensington. anyway she was like totally obsessed with talking about a recent gym workout routine and like how imogen is doing like five... um... i mean she is basically doing five sets of everything now. it sounds kinda exhausted honestly. do you think they are just like trying too hard or what", "clean": "So I was talking to Yara and she said she tried booking a table for brunch at that place in Kensington. Anyway, she was totally obsessed with talking about a recent gym workout routine and how Imogen is doing 5 sets of everything now. It sounds exhausting honestly. Do you think they are just trying too hard or what?", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl6907.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 627, "source_record_id": "assigned_records.jsonl:6907"}
{"oral": "Oh my god! So like this book totally flipped my world upside down. I remember reading it around midnight on New Year's Eve while I was planning a route for road trip. Just before that I was busy organizing a surprise birthday party for a colleague in sales... oh no not sales marketing department. Anyway I was also dealing with this nightmare asking for refund on faulty gadget and I just felt so burnt out. Then I met Hiroshi and he recommended this book. I mean it just made everything click for me ya.", "clean": "Oh my god! This book totally flipped my world upside down. I remember reading it around midnight on New Year's Eve while I was planning a route for a road trip. Just before that I was busy organizing a surprise birthday party for a colleague in the marketing department. Anyway I was also dealing with this nightmare asking for a refund on a faulty gadget and I just felt so burnt out. Then I met Hiroshi and he recommended this book. It just made everything click for me ya.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl7219.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 628, "source_record_id": "assigned_records.jsonl:7219"}
{"oral": "Omg you wont believe this dream i had! So like i was in London right and i was just hanging out in Piccadilly Circus... oh no wait Leicester Square and i keep seeing Saoirse Ronan everywhere. She was like actually asking for a recommendation for skincare routine and then we start planning a trip to local farmers market for the day after tomorrow. It was just so random like totally weird.", "clean": "Omg you won't believe this dream I had! So I was in London right and I was just hanging out in Leicester Square and I keep seeing Saoirse Ronan everywhere. She was actually asking for a recommendation for a skincare routine and then we start planning a trip to a local farmers market for the day after tomorrow. It was just so random, totally weird.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl7338.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 630, "source_record_id": "assigned_records.jsonl:7338"}
{"oral": "Hey Yuna, I gotta tell you about this dream I had last night. So I was at Borough Market... Borough Market in London, right? And I really wanna find some food but I was actually coordinating a pickup at the airport... oh no, coordinating a pickup at the train station. And I was like I gotta get outta there, I gotta get outta there fast or I'm gonna miss it. It was so weird.", "clean": "Hey Yuna, I gotta tell you about this dream I had last night. So I was at Borough Market in London, right? And I really wanna find some food but I was coordinating a pickup at the train station. And I gotta get outta there fast or I'm gonna miss it. It was so weird.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl7364.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 631, "source_record_id": "assigned_records.jsonl:7364"}
{"oral": "Honestly I tried doing the whole resolution thing but it just not working. Like Priya and Kenji they were organizing a carpool... organizing a carpool for a concert last week and it was a total mess. We booked an AirBnB in Trafalgar Square... oh no not Trafalgar Square I mean Leicester Square and it was way too far. I really need to get organized get organized with my life but yeah its hard.", "clean": "Honestly I tried doing the whole resolution thing but it's just not working. Priya and Kenji were organizing a carpool for a concert last week and it was a total mess. We booked an Airbnb in Leicester Square and it was way too far. I really need to get organized with my life but yeah it's hard.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl7554.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 633, "source_record_id": "assigned_records.jsonl:7554"}
{"oral": "Priya, uh, i gotta tell you about this book that totally changed my life. it was like... it was like a complete wake up call. i remember i was reading it at Venice Beach, eh, right before i went to Sephora. wait, i was at Santa Monica Pier... oh no, Venice Beach was right. anyway, i was so focused i almost forgot i was requesting refund for flight delay, i mean, requesting refund for flight delay about twenty minutes ago. it just makes you realize how small our problems are, you know, just kinda puts things in perspective, i guess.", "clean": "Priya, i gotta tell you about this book that totally changed my life. it was a complete wake up call. i remember i was reading it at Venice Beach right before i went to Sephora. anyway, i was so focused i almost forgot i was requesting a refund for flight delay about 20 minutes ago. it just makes you realize how small our problems are, just puts things in perspective, i guess.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl7462.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 632, "source_record_id": "assigned_records.jsonl:7462"}
{"oral": "So like I was just recounting a weird dream from last night where we were somehow at The Louvre but then it turned into planning a route for a road trip... actually wait we need to hit the store first. I was thinking maybe we could do the big shop bi-weekly on Thursdays? Oh and I wanted to ask for a recommendation for a skincare routine because my skin is just... anyway we can talk about that after we get the milk.", "clean": "I was just recounting a weird dream from last night where we were somehow at The Louvre, but then it turned into planning a route for a road trip... we need to hit the store first. I was thinking maybe we could do the big shop bi-weekly on Thursdays? Oh and I wanted to ask for a recommendation for a skincare routine because my skin is just... anyway we can talk about that after we get the milk.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl7628.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 634, "source_record_id": "assigned_records.jsonl:7628"}
{"oral": "hey man how was your weekend? oh i actually saw thalia and zayn on saturday. wait no... not zayn... it was saoirse. we were all just hanging out playing nintendo switch for like four hours. then we went out for food but man it was such a mess trying to split a dinner bill between the three of us. total nightmare.", "clean": "Hey man, how was your weekend? Oh, I actually saw Thalia and Saoirse on Saturday. We were all just hanging out playing Nintendo Switch for 4 hours. Then we went out for food, but man, it was such a mess trying to split a dinner bill between the three of us. Total nightmare.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl7633.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 635, "source_record_id": "assigned_records.jsonl:7633"}
{"oral": "Yara, listen, like, I saw Chloe about a couple hours ago at The Louvre, oh no, I mean The Shard! And she was telling me... I think... well, she mentioned that she's basically moving to Paris, just... let's say, she wants to be closer to the actual art museums and stuff, ya know?", "clean": "Yara, listen, I saw Chloe about a couple hours ago at The Shard! And she was telling me... she mentioned that she's moving to Paris, let's say, she wants to be closer to the actual art museums and stuff.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl7737.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 637, "source_record_id": "assigned_records.jsonl:7737"}
{"oral": "Hey Sarah, so like I had this crazy dream last night. I was just... I mean I was standing on Brooklyn Bridge, right? And for some reason I was holding a Kindle and just complaining about late Amazon delivery, just like... you know, just venting. Then suddenly I'm in Costco, uh, and I start complaining about slow Wi-Fi connection, just... it was just so random, I guess.", "clean": "Hey Sarah, I had this crazy dream last night. I was standing on Brooklyn Bridge, right? And for some reason I was holding a Kindle and complaining about late Amazon delivery, just venting. Then suddenly I'm in Costco, and I start complaining about slow Wi-Fi connection; it was just so random, I guess.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl7731.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 636, "source_record_id": "assigned_records.jsonl:7731"}
{"oral": "hey do you know any good dentists around here? i mean i was talking to dmitri and beatrix and fatima about it like when we were splitting splitting the bill at tapas restaurant on january first and... well i just need someone who wont rip me off. i think i think i need a cleaning soon, how should i say it, maybe a checkup too.", "clean": "Hey, do you know any good dentists around here? I was talking to Dmitri and Beatrix and Fatima about it when we were splitting the bill at Tapas restaurant on January 1, and I just need someone who won't rip me off. I think I need a cleaning soon, maybe a checkup too.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl8389.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 643, "source_record_id": "assigned_records.jsonl:8389"}
{"oral": "Honestly I just cannot deal with that guy upstairs anymore like he is just constant venting about a noisy neighbor is an understatement he is like a human jackhammer. Anyway uh we gotta figure out grocery list. I was thinking about... maybe we could... well let's just say we need a better way to meal prep for next week so we dont spend like forty dollars a day on lunch. Maybe we get those big containers. Oh and did you see that AirBNB Saoirse sent us for the trip? It looks sick. Maybe we can just go Wagamama for dinner tonight and do the big shop tomorrow morning around ten.", "clean": "Honestly I just cannot deal with that guy upstairs anymore; he is just constant venting about a noisy neighbor is an understatement, he is a human jackhammer. Anyway we gotta figure out grocery list. I was thinking maybe we need a better way to meal prep for next week so we don't spend $40 a day on lunch. Maybe we get those big containers. Oh and did you see that Airbnb Saoirse sent us for the trip? It looks sick. Maybe we can just go Wagamama for dinner tonight and do the big shop tomorrow morning around 10.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl7930.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 638, "source_record_id": "assigned_records.jsonl:7930"}
{"oral": "so like i was on phone around lunch time like just totally explaining this like super complex recipe over phone to dimitri cuz we were discussing recipe for potluck and then i had to stop by sephora for a bit and i think i spent like twenty dollars... oh no wait twenty five dollars on some stuff anyway like i think we should... uh... we could maybe... how do i put this... just go see that new horror movie tonight", "clean": "I was on the phone around lunch time explaining this super complex recipe over the phone to Dimitri because we were discussing a recipe for a potluck, and then I had to stop by Sephora for a bit and I think I spent $25 on some stuff. Anyway, I think we could maybe just go see that new horror movie tonight.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl8047.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 641, "source_record_id": "assigned_records.jsonl:8047"}
{"oral": "so i had this crazy dream last night and like i was venting about a long line at the DMV to some people i dont even know. wait it was a bank... oh no it was the DMV. any way we were all just standing there and then suddenly we were at in n out eating burgers. beatrix was there and saoirse and ji hu. it was so random. i think i think we were arguing about the fries. and i remember thinking if we just left the line then we could go to the beach. we could go to the beach and just forget about the paperwork. it was just just so weird ya.", "clean": "I had this crazy dream last night and I was venting about a long line at the DMV to some people I don't even know. Anyway, we were all just standing there and then suddenly we were at In-N-Out eating burgers. Beatrix was there and Saoirse and Ji Hu. It was so random. I think we were arguing about the fries. And I remember thinking if we just left the line then we could go to the beach and just forget about the paperwork. It was just so weird.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl8021.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 640, "source_record_id": "assigned_records.jsonl:8021"}
{"oral": "Man! I mean everything just costs way too much now. I was talking to Kwame and Aoife on WhatsApp and we were like... uh... we were just saying how... well... it's like we can't even afford basic stuff anymore. Like Xavier tried to get that Tesla and the price is just... it's insane. I saw some thing yesterday that was like eighty nine ninety-nine for a small bag of coffee. I mean who does that? It's just totaly ridiculous how fast prices go up. I can't believe it's been two years since things were actually cheap. It's just not right.", "clean": "Man! Everything just costs way too much now. I was talking to Kwame and Aoife on WhatsApp and we were just saying how we can't even afford basic stuff anymore. Xavier tried to get that Tesla and the price is just insane. I saw something yesterday that was $89.99 for a small bag of coffee. Who does that? It's just totally ridiculous how fast prices go up. I can't believe it's been 2 years since things were actually cheap. It's just not right.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl8370.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 642, "source_record_id": "assigned_records.jsonl:8370"}
{"oral": "so i had this crazy dream last night and it started like i was in london right i was walking around piccadilly circus and i saw priya there. we decided we needed coffee so we went to costco... oh no not costco it was starbucks and we were just waiting in line. i remember i was ordering a latte with oat milk and syrup and i was ordering a latte with oat milk and syrup like five times cause the guy couldnt hear me. then suddenly this tesla just drives through the wall and the tesla just drives through the wall right into the cafe and it was just totally random.", "clean": "So I had this crazy dream last night and it started I was in London, right? I was walking around Piccadilly Circus and I saw Priya there. We decided we needed coffee so we went to Starbucks and we were just waiting in line. I remember I was ordering a latte with oat milk and syrup 5 times cause the guy couldn't hear me. Then suddenly this Tesla just drives through the wall right into the cafe and it was just totally random.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl7959.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 639, "source_record_id": "assigned_records.jsonl:7959"}
{"oral": "um i'm not totally sure but i think Alistair and Mei Lin were talking about like comparing different electric vehicle models earlier today oh wait no it was just Alistair and like some guy from work anyway we still gotta figure out splitting bill after dinner like at Pike Place Market... wait no it was Pike Place Fish Market oh no sorry Pike Place Public Market actually and then we can just do it on venmo or whatever since it's gonna be like twenty people or so and it's gonna be a lot of money to keep track of so we should just do it the second we get out of there", "clean": "I'm not totally sure but I think Alistair and some guy from work were talking about comparing different electric vehicle models earlier today. Anyway, we still gotta figure out splitting the bill after dinner at Pike Place Public Market, and then we can just do it on Venmo or whatever since it's gonna be 20 people or so and it's gonna be a lot of money to keep track of, so we should just do it the second we get out of there.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl8401.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 644, "source_record_id": "assigned_records.jsonl:8401"}
{"oral": "so i read this book by alastair and man it totally changed how i see things. i remember i was eating at in n out... i mean i was at the met... oh no not the met i was at the met museum with mei ling. i think i think it was back in twenty twenty two. i m not totally sure but i think we spent like four hours just talking about it. it made me realize that life life is just about perspective. i actually tried to explain it to her while we were walking and i think i think it clicked for her too.", "clean": "So I read this book by Alastair and man it totally changed how I see things. I remember I was at the Met Museum with Mei Ling. I think it was back in 2022. I'm not totally sure but I think we spent 4 hours just talking about it. It made me realize that life is just about perspective. I actually tried to explain it to her while we were walking and I think it clicked for her too.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl8516.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 645, "source_record_id": "assigned_records.jsonl:8516"}
{"oral": "hey so uh did you do anything fun like oh Osei and Anne went to the Peak District... oh no sorry the Cotswolds and it was more or less goated if you free next weekend then we could go too", "clean": "Hey, did you do anything fun? Osei and Anne went to the Cotswolds and it was more or less goated. If you're free next weekend, then we could go too.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl9063.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 649, "source_record_id": "assigned_records.jsonl:9063"}
{"oral": "ya i wanna visit Ms Gable in Melbourne Laneys and then maybe go to The Rusty Anchor Bar... oh no I mean The Rusty Anchor Pub plus like... let me think... how should i say... planning a visit to water park once a quarter i guess ha. also i gotta go to that garden... wait i mean garden court park ma", "clean": "I want to visit Ms Gable in Melbourne Laneys and then maybe go to The Rusty Anchor Pub. Planning a visit to the water park once a quarter. Also I gotta go to Garden Court Park.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl9086.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 650, "source_record_id": "assigned_records.jsonl:9086"}
{"oral": "im not totally sure but i think mr henderson... uh... he lives in upper west side... no upper east side... wait wrong... it's the upper east side! and he was... like... talking bout... how do i put this... discussing pottery at the blue lagon... oh wait the blue lagoon... and he had some hydro flask in his hand i think he bought it at the mall in new york city", "clean": "I'm not totally sure, but I think Mr. Henderson lives in the Upper East Side, and he was discussing pottery at the Blue Lagoon, and he had some Hydro Flask in his hand; I think he bought it at the mall in New York City.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl9014.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 648, "source_record_id": "assigned_records.jsonl:9014"}
{"oral": "dude i swear those people next door are really driving me crazy like i'm not totally sure but i think they're planning a surprise twenty fifth... oh no sorry thirtyth birthday party or something and i could hear zahra and oluchi just like shouting and sharing a life hack for cleaning the carpets at three in the morning like who even does that right i mean i tried to ignore it but it's just way too loud for a tuesday night", "clean": "Dude, I swear those people next door are really driving me crazy. I'm not totally sure, but I think they're planning a surprise 30th birthday party or something, and I could hear Zahra and Oluchi shouting and sharing a life hack for cleaning the carpets at 3:00am. Who even does that, right? I tried to ignore it, but it's just way too loud for a Tuesday night.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl8900.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 647, "source_record_id": "assigned_records.jsonl:8900"}
{"oral": "Man! you were venting venting about missed alarm... just like ten... five minutes ago... no wait two minutes ago... wait wrong it was about ten minutes ago! eh we were planning planning visit to flea market... at the... the Savoy Hotel... no wait at Burger King... wait wrong it is Savoy Hotel! i mean you just... just need some space maybe ya", "clean": "Man! You were venting about a missed alarm about 10 minutes ago. We were planning a visit to the flea market at the Savoy Hotel. You just need some space maybe.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl9110.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 651, "source_record_id": "assigned_records.jsonl:9110"}
{"oral": "hey so uh what'd you do this weekend? i actually went to singapore like... well i mean i just flew in and spent some time at clarke quay and uh had kofi with a friend while talking about a new fashion trend and like... at the same time i was thinking bout the sydney opera house since i want to go there too but anyway i just... stayed in the city center for two days", "clean": "Hey, what'd you do this weekend? I flew in and spent some time at Clarke Quay and had coffee with a friend while talking about a new fashion trend. At the same time, I was thinking about the Sydney Opera House since I want to go there too, but anyway I just stayed in the city center for 2 days.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl8862.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 646, "source_record_id": "assigned_records.jsonl:8862"}
{"oral": "hey so uh i went to like... the Lincoln Memorial... no wait the Washington Monument... uh actually it was the National Mall! and then spent like two hours planning a picnic in park with Poppy plus also keep complaining about loud air conditioner at home before we get together", "clean": "I went to the National Mall and then spent 2 hours planning a picnic in the park with Poppy, plus also kept complaining about the loud air conditioner at home before we got together", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl9592.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 654, "source_record_id": "assigned_records.jsonl:9592"}
{"oral": "so i found this place in Chelsea... oh no Kensington a few weeks ago that fancy a Lake Como vibe you know? like we were deciding which sneakers buy for marathon and i saw it. you fancy a go?", "clean": "I found this place in Kensington a few weeks ago that has a Lake Como vibe. We were deciding which sneakers to buy for the marathon and I saw it. Fancy a go?", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl9612.wav", "scene": "daily_chat", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 655, "source_record_id": "assigned_records.jsonl:9612"}
{"oral": "Hey Sarah, eh I gotta tell you I got like a crazy deal today. So I was scheduling dentist appointment over phone and basically multitasking while organizing surprise welcome home party for Chioma, you know? Then I saw this voucher for Pasta Palace and it was like fifty percent off but you get tickets to CN Tower too, just... how do you say... it was a total steal for only twenty dollars, like it was just too good to pass up right now in the new city, ya.", "clean": "Hey Sarah, I gotta tell you I got a crazy deal today. So I was scheduling a dentist appointment over the phone and multitasking while organizing a surprise welcome home party for Chioma. Then I saw this voucher for Pasta Palace and it was 50% off but you get tickets to CN Tower too; it was a total steal for only $20, it was just too good to pass up right now in the new city, ya.", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl9292.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 653, "source_record_id": "assigned_records.jsonl:9292"}
{"oral": "Oh man! So uh I was hangin out with Kofi on Saturday and we was like planning dinner menu for guests... wait no I mean lunch menu for guests and then yall know how it is we spent like two hours just discussing cliffhanger in that TV series we keep talking about... I was tryna explain the twist to him but like I totally forgot the main part of the plot in the new episode... it was so wild man I really think the lead character is dead in the new season", "clean": "Oh man! I was hanging out with Kofi on Saturday and we were planning the lunch menu for guests, and then you all know how it is, we spent 2 hours just discussing the cliffhanger in that TV series we keep talking about... I was trying to explain the twist to him but I totally forgot the main part of the plot in the new episode... it was so wild man, I really think the lead character is dead in the new season", "usage_text_words": null, "audio": "audio/daily_chat__assigned_recordsjsonl9281.wav", "scene": "daily_chat", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 652, "source_record_id": "assigned_records.jsonl:9281"}
{"oral": "嗯，充电线拿了没，对了，吴经理说文件在百度网盘，百度网盘里那个文件夹，我得赶紧下好。话说房东那边得打个招呼，哦对了，钥匙给孙姐，给孙姐吧。诶，Slack消息发给财务了没，发给财务三号，啊不，是五号，哈。我想想啊，行程单打印两份，两份才稳妥嘛。", "clean": "充电线拿了没？对了，吴经理说文件在百度网盘，百度网盘里那个文件夹，我得赶紧下好。话说房东那边得打个招呼，哦对了，钥匙给孙姐吧。Slack消息发给财务了没，发给财务5号。我想想，行程单打印2份，2份才稳妥嘛。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11734.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 657, "source_record_id": "assigned_records.jsonl:11734"}
{"oral": "说白了就是得赶紧把东西塞完，嗯，审稿审稿得带平板，还得跟钱助理同步一下，同步一下行程。我想想啊，Notion 里的对账单还得再核一遍，对账单核一遍才放心。诶，充电线好像忘在卧室在卧室了，赶紧拿上吧。", "clean": "说白了得赶紧把东西塞完，审稿得带平板，还得跟钱助理同步一下行程。Notion里的对账单还得再核一遍，对账单核一遍才放心。充电线好像忘在卧室了，赶紧拿上吧。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11778.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 659, "source_record_id": "assigned_records.jsonl:11778"}
{"oral": "嗯，我想想啊，婚礼请帖与其找外面公司做，不如直接用Canva自己搞，比花三千五百块找人设计更省心。顺便问问周经理，不对，得问周姐，孩子能不能带去，要是带孩子比不带孩子更麻烦吧。", "clean": "婚礼请帖与其找外面公司做，不如直接用Canva自己搞，比花3500元找人设计更省心。顺便问问周姐，孩子能不能带去，要是带孩子比不带孩子更麻烦吧。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11779.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 660, "source_record_id": "assigned_records.jsonl:11779"}
{"oral": "嗯，先是把韩医生那个病例发给吴老师，吴老师那个最新的病例，然后把Slack上那个链接，不对，是知乎上那个链接发过去哈，最后再跟进跟进一下进度吧，我想想啊，大概在下午三点半左右发到道他邮箱里呀。", "clean": "先是把韩医生的病例发给吴老师，吴老师最新的病例，然后把知乎上那个链接发过去，最后再跟进一下进度，大概在15:30左右发到他邮箱里。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11764.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 658, "source_record_id": "assigned_records.jsonl:11764"}
{"oral": "嗯，年终总结怎么写嘛。我想想啊，与其把所有细节都堆在石墨文档里，不如先在Trello列个清单。比单纯列时间线更重要的是突出成果，突出成果哈。诶，之前知乎上看到的那个方法，就是把数据对比写得具体一点，得写得具体一点。接着，吴老师提到的那三个核心指标，得在二零二三年十二月三十一号之前跑完。我想想，其实比追求完美更关键是赶紧写完吧，在石墨文档里直接敲就行了。", "clean": "年终总结怎么写。与其把所有细节都堆在石墨文档里，不如先在Trello列个清单。比单纯列时间线更重要的是突出成果。之前知乎上看到的那个方法，就是把数据对比写得具体一点。接着，吴老师提到的那三个核心指标，得在2023年12月31日之前跑完。其实比追求完美更关键是赶紧写完吧，在石墨文档里直接敲就行了。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11702.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 656, "source_record_id": "assigned_records.jsonl:11702"}
{"oral": "诶，我想把周主管这块的任务直接交给赵总，就是因为赵总他他有更多资源。嗯，原本打算用Zoom开会，但现在觉得企业微信企业微信更方便，怎么说呢，这样黄总那边…要是能直接在群里同步…这样黄总能随时看进度吧。", "clean": "我想把周主管这块的任务直接交给赵总，因为赵总有更多资源。原本打算用Zoom开会，但现在觉得企业微信更方便，这样黄总那边要是能直接在群里同步，黄总能随时看进度吧。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11809.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 661, "source_record_id": "assigned_records.jsonl:11809"}
{"oral": "今天练腿状态太差，主要是昨天预约预约太晚，没睡够。深蹲做了四组，每组十二个，不对，是十个。把这些记录记录到Trello里，一会儿回家续费会员。刚才滴滴司机问我怎么走，我正把百度网盘链接发他，完全没心思训练。", "clean": "今天练腿状态太差，主要是昨天预约太晚，没睡够。深蹲做了4组，每组10个。把这些记录到Trello里，一会儿回家续费会员。刚才滴滴司机问我怎么走，我正把百度网盘链接发他，完全没心思训练。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11813.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 662, "source_record_id": "assigned_records.jsonl:11813"}
{"oral": "你说家里还剩多少蚝油？我想想啊，好像没多少蚝油，顺便把生抽也买了。诶，是美团快还是饿了么快？小王小王之前说美团送得快，林主管也这么说。买两瓶，买两瓶大号的，我想想啊，应该得花二十块左右吧。我觉得美团可以。", "clean": "你说家里还剩多少蚝油？我想想，好像没多少蚝油，顺便把生抽也买了。是美团快还是饿了么快？小王之前说美团送得快，林主管也这么说。买2瓶大号的，应该得花20元左右吧。我觉得美团可以。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11817.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 663, "source_record_id": "assigned_records.jsonl:11817"}
{"oral": "诶，今天的预定预定情况是这样的，我把具体明细发 Slack 还有 Trello 里面，钱姐说那边大概要花三千四百块，三千四百块左右，我想想啊，我待会把那个小宇宙播客推荐也发你，你觉得呢吧？", "clean": "今天的预定情况是这样的，我把具体明细发Slack还有Trello里面，钱姐说那边大概要花3400元左右，我待会把小宇宙播客推荐也发你，你觉得呢？", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11826.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 664, "source_record_id": "assigned_records.jsonl:11826"}
{"oral": "孙主任说得那个药得每天吃两回，然后，那个…怎么说呢…得在饭后半个钟头再吃，另外记得把那个腾讯会议链接发给赵总，还有让小王帮郑先生预约下周三下午两点半的复诊，记得得在那个预约苑里操作。", "clean": "孙主任说那个药得每天吃2回，得在饭后30分钟再吃，另外记得把腾讯会议链接发给赵总，还有让小王帮郑先生预约下周三14:30的复诊，记得得在预约院里操作。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11827.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 665, "source_record_id": "assigned_records.jsonl:11827"}
{"oral": "诶孙姐，企微那边…我想想啊，企微那边林医生发了三条消息，说快递小哥下午两点半到，结果结果没对上，我想想…要是直接在Slack上通知他，通知他应该更快吧，你觉得呢？", "clean": "孙姐，企微那边林医生发了3条消息，说快递小哥14:30到，结果没对上，要是直接在Slack上通知他应该更快吧，你觉得呢？", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11881.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 669, "source_record_id": "assigned_records.jsonl:11881"}
{"oral": "那个，先是得把马总预定那个鱼露买回来，然后，我想想，还得买那个，怎么说呢，把刚才抖音看到那个特种酱油加上，除此之外，钱助理说要买两瓶白醋，最后，最后得买三块钱一包那个盐，对了，得买两包。", "clean": "先得把马总预定的鱼露买回来，然后把刚才抖音看到的特种酱油加上，除此之外，钱助理说要买2瓶白醋，最后得买3元一包的盐，得买2包。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11832.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 667, "source_record_id": "assigned_records.jsonl:11832"}
{"oral": "说白了，嗯，这次流程得把Slack以及企微同步整合起来，就是把把同步整合起来，这样订票系统在再调用的时候才不会乱。我想想啊，最初预想是三个接口，哦不是，四个接口，然后通过Zoom开会定一下细节，细节，最后发给客户确认吧。", "clean": "说白了，这次流程得把Slack以及企微同步整合起来，这样订票系统在调用的时候才不会乱。最初预想是4个接口，然后通过Zoom开会定一下细节，最后发给客户确认吧。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11965.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 670, "source_record_id": "assigned_records.jsonl:11965"}
{"oral": "关于韩医生那个项目，目前进度是这样的，之前在Slack上跟周主管对过，整体差不多完成了百分之八十，但有个问题，那个…我想想…要是把知乎那块的权重…怎么说呢，如果把数据同步频率调高点，应该会更稳，钱经理那边要是…要是觉得太激进可能不答应，这个方案你觉得呢？我想再确认一下，怕得不到认可。", "clean": "关于韩医生那个项目，目前进度是这样的，之前在Slack上跟周主管对过，整体差不多完成了80%，但有个问题，要是把知乎那块的权重，如果把数据同步频率调高点，应该会更稳，钱经理那边要是觉得太激进可能不答应，这个方案你觉得呢？我想再确认一下，怕得不到认可。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11831.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 666, "source_record_id": "assigned_records.jsonl:11831"}
{"oral": "诶，年终总结不是写成工作流水账吧，其实得突出重点。我想想啊，前台接待这块，不是说只要接电话，其实核心是客情维护。比如郑女士，郑女士那次投诉，我处理得不错。怎么说呢，滴滴报销的事，小李小李得帮我核对下，总额大概是三千四百块。赵总赵总之前提过，要提高效率，哈，这个得重点写。", "clean": "年终总结不要写成工作流水账，其实得突出重点。前台接待这块，不是说只要接电话，其实核心是客情维护。比如郑女士那次投诉，我处理得不错。滴滴报销的事，小李得帮我核对下，总额大概是3400元。赵总之前提过，要提高效率，这个得重点写。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11841.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 668, "source_record_id": "assigned_records.jsonl:11841"}
{"oral": "嗯，周主管说这个鱼得先腌三十分钟，我觉得三十分钟三十分钟太久了，容易太咸吧。我想想啊，孙姐给我的那个，百度网盘里那个文档，说是得加百分之五的糖，郑先生在GitHub上面分享的配方也提到这点，得加糖加糖才行，哈。", "clean": "周主管说这个鱼得先腌30分钟，我觉得30分钟太久了，容易太咸吧。我想想，孙姐给我的百度网盘里那个文档，说是得加5%的糖，郑先生在GitHub上面分享的配方也提到这点，得加糖才行。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11998.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 673, "source_record_id": "assigned_records.jsonl:11998"}
{"oral": "关于那个续费逻辑，我刚才想，咱们能不能参考下知乎或者美团那种，呃，就是把那个...怎么说呢...如果把入口直接放在首页顶端，这样用户感知度更高。对了，这块儿得让小李那边核对下，你看这么搞行吗？", "clean": "关于续费逻辑，我刚才想，咱们能不能参考下知乎或者美团那种，如果把入口直接放在首页顶端，这样用户感知度更高。对了，这块儿得让小李那边核对下，你看这么搞行吗？", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12010.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 674, "source_record_id": "assigned_records.jsonl:12010"}
{"oral": "诶，明天到底要搞定多少事儿？我想想啊，上午十点半得先跟郑经理开个Zoom会，讨论下知乎那边推广方案，就是，得把预算调到三千五百块，哎呀，对了，还得看看美团外卖那个数据报告，抖音那边 短视频脚本得改一遍，我想应该能搞定吧。", "clean": "明天到底要搞定多少事儿？我想想，10:30得先跟郑经理开个Zoom会，讨论下知乎那边推广方案，得把预算调到3500元，还得看看美团外卖那个数据报告，抖音那边短视频脚本得改一遍，我想应该能搞定吧。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12019.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 675, "source_record_id": "assigned_records.jsonl:12019"}
{"oral": "哎呀太夸张了，今天深蹲居然推了一百二十公斤，嗯，冯总冯总之前说能到这个重量，我还没敢信。记录在Trello，记录在Trello里，得把数据传到Zoom会议那个文档里，我想想啊，周主任再三提醒过要量力而行，结果我今天竟然到点位了。", "clean": "太夸张了，今天深蹲居然推了120公斤，冯总之前说能到这个重量，我还没敢信。记录在Trello里，得把数据传到Zoom会议那个文档里，周主任再三提醒过要量力而行，结果我今天竟然到点位了。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11993.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 672, "source_record_id": "assigned_records.jsonl:11993"}
{"oral": "马总说GitHub那个权限得改改，得改改。本来想让小李给周姐开读写权限，啊不对，是只读权限。你那边Slack同步下，同步下这个事，你觉得呢？", "clean": "马总说GitHub那个权限得改，本来想让小李给周姐开只读权限。你那边Slack同步下这个事，你觉得呢？", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12124.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 677, "source_record_id": "assigned_records.jsonl:12124"}
{"oral": "说白了说白了，咱们得让郑主管把那个面试流程给改改，面试流程得简化。简单来说，把之前在企微发的文件，啊不对，是京东云盘里的那个模板，直接用Canva做成海报。然后那个…怎么说呢…如果能直接发给候选人，候选人直接在手机上就能看。咱们再再发一遍，把那个地址发到新苑里去，发到新苑里。", "clean": "说白了，咱们得让郑主管把面试流程给改改，面试流程得简化。简单来说，把之前在京东云盘里的那个模板，直接用Canva做成海报。如果能直接发给候选人，候选人直接在手机上就能看。咱们再发一遍，把那个地址发到新苑里去。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl11974.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 671, "source_record_id": "assigned_records.jsonl:11974"}
{"oral": "周姐，我刚才在 Slack 上跟你说那个请假申请，我下周二，啊不对，是下周三请假，我老婆那边房东要过来签合同，得在家盯着，你帮我看下行不行，我手机快没电关机了", "clean": "周姐，我刚才在 Slack 上跟你说请假申请，我下周三请假，我老婆那边房东要过来签合同，得在家盯着，你帮我看下行不行，我手机快没电关机了", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12136.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 678, "source_record_id": "assigned_records.jsonl:12136"}
{"oral": "嗯，今天深蹲做了四组，每组十二个。哦对了，刚才吴老师说得调低点重量，我想想啊，刚才那组好像有点吃力。话说其次得把心率调到一百四十，诶，刚才韩总在美团点外卖没，我看他在那刷美团吧。对了，得在明天之前把计划表发给教练呀。", "clean": "今天深蹲做了4组，每组12个。对了，刚才吴老师说得调低点重量，刚才那组好像有点吃力。其次得把心率调到140，刚才韩总在美团点外卖没，我看他在那刷美团吧。对了，得在明天之前把计划表发给教练呀。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12061.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 676, "source_record_id": "assigned_records.jsonl:12061"}
{"oral": "你说钱助理之前提 那个 Zoom 会议能不能改到下午三点半？嗯，得让前台帮我确认下，还有携程订票三千五百块钱报销单，我想想啊，应该是直接发给他，我觉得可以。", "clean": "你说钱助理之前提 Zoom 会议能不能改到 15:30？得让前台帮我确认下，还有携程订票 3500 元报销单，应该是直接发给他，我觉得可以。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12181.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 680, "source_record_id": "assigned_records.jsonl:12181"}
{"oral": "嗯，吴教授说那个药得续费，另外还得跟周主管汇报一下，我也不太确定具体时间，但是感觉得在下午三点半之前弄完，怎么说呢，要是时间来不及的话，可能得约个Zoom开会确认一下吧。", "clean": "吴教授说那个药得续费，另外还得跟周主管汇报一下，我也不太确定具体时间，但是感觉得在15:30之前弄完，要是时间来不及的话，可能得约个Zoom开会确认一下吧。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12193.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 681, "source_record_id": "assigned_records.jsonl:12193"}
{"oral": "如果钱管家能把那个房东给说通，说通的话，那我们就能在二零二五年三月一号之前把定金拿回来，那个…要是把…怎么说呢…如果把吴老师也拉进来帮着说，咱们就能把三千五百块快快拿回来，赶紧联系联系郑先生，问他能不能再到道那边帮个忙。", "clean": "如果钱管家能把房东给说通，说通的话，那我们就能在2025年3月1日之前把定金拿回来。如果把吴老师也拉进来帮着说，咱们就能把3500元快快拿回来，赶紧联系联系郑先生，问他能不能在道那边帮个忙。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12158.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 679, "source_record_id": "assigned_records.jsonl:12158"}
{"oral": "还得买瓶生抽，然后那个...怎么说呢...把小红书上那个低钠版记一下，哦不是，是买美团那个低钠版，另外得看看滴滴账单，我想想...那个...得把欠赵总钱的事儿给结了，然后等下Zoom会议结束赶紧下单。", "clean": "还得买瓶生抽，然后把美团那个低钠版记一下，另外得看看滴滴账单，得把欠赵总钱的事儿给结了，然后等下Zoom会议结束赶紧下单。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12200.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 682, "source_record_id": "assigned_records.jsonl:12200"}
{"oral": "对了，得赶紧把GitHub仓库同步一下，同步一下，不然到现场没网麻烦。嗯，钱助理说这次体检报告得带原件，带原件，诶，话说小王说帮我订了下午两点半的票，啊不对，是三点半的票，哈。哦对了，Canva那个演示文稿还没导出，得赶紧弄完吧。", "clean": "得赶紧把GitHub仓库同步一下，不然到现场没网麻烦。钱助理说这次体检报告得带原件。话说小王说帮我订了15:30的票。对了，Canva那个演示文稿还没导出，得赶紧弄完吧。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12278.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 684, "source_record_id": "assigned_records.jsonl:12278"}
{"oral": "嗯，这个做法比起小红书上那个教程更简单，与其在那儿死磕火候不如直接用空气炸锅，记得提醒小王把钱管家同步一下，对了最后得把截图发到Slack，我想想，应该是再加热两分钟就行，这样比之前那个版本好得多。", "clean": "这个做法比起小红书上的教程更简单，与其在那儿死磕火候不如直接用空气炸锅，记得提醒小王把钱管家同步一下，对了最后得把截图发到Slack，应该是再加热2分钟就行，这样比之前那个版本好得多。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12290.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 685, "source_record_id": "assigned_records.jsonl:12290"}
{"oral": "充电宝充好没，充电宝得带上，给郑经理准备文件，给郑经理准备的文件放包里没，不对，是给郑先生准备，周工说得带个U盘，U盘得带个备份的，滴滴约下午两点半，啊不，两点一刻到机场，郑女士那边会议时间没变吧。", "clean": "充电宝充好没，充电宝得带上。给郑先生准备的文件放包里没？周工说得带个U盘，U盘得带个备份的。滴滴约14:15到机场，郑女士那边会议时间没变吧。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12308.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 686, "source_record_id": "assigned_records.jsonl:12308"}
{"oral": "嗯，捐赠清单，先记一下。那个给赵总的两本杂志，不对，其实给吴老师更合适，就放一起吧。诶，还有小红书上刷到说要捐旧衣服，得把Trello里标记三件，不对，是三件长袖，得赶紧跟进一下，我想想啊，是不是得装在那个大纸箱子里呀。", "clean": "捐赠清单，先记一下。给吴老师的两本杂志，就放一起吧。还有小红书上刷到说要捐旧衣服，得把Trello里标记3件长袖，得赶紧跟进一下，我想想，是不是得装在那个大纸箱子里呀。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12275.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 683, "source_record_id": "assigned_records.jsonl:12275"}
{"oral": "诶，如果美团能把接机服务和对账流程给整合起来，我想想啊，要是能让快递小哥顺便说一下，帮用户把东西带到机场的话，那我们就能，怎么说呢，直接提升整体配送效率吧。", "clean": "如果美团能把接机服务和对账流程整合起来，要是能让快递小哥顺便帮用户把东西带到机场的话，那我们就能直接提升整体配送效率吧。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12354.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 688, "source_record_id": "assigned_records.jsonl:12354"}
{"oral": "孙姐，刚才看那套三室两厅，采光真不错，但那个客厅面积大概有二十五平米，哦不是，是十八平米，有点局促。吴老师推荐那套虽然贵，但地段没得说。对了，那个李经理说能打九五折，我得赶紧把这些对比项填到Trello里去，省得忘了。", "clean": "孙姐，刚才看那套三室两厅，采光真不错，但客厅面积大概有18平米，有点局促。吴老师推荐那套虽然贵，但地段没得说。对了，李经理说能打95折，我得赶紧把这些对比项填到Trello里去，省得忘了。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12312.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 687, "source_record_id": "assigned_records.jsonl:12312"}
{"oral": "嗯，今天这行程，我觉得与其在市区逛，不如直接打滴滴去那个博物馆，比挤地铁舒服多了。刚才算了下，打车花了五十八块，不对，是六十二块。再者，还得把之前在京东买的那个充电宝给周主任，我想想啊，明天见面对比发快递更稳当吧。对了，还得用Canva改一下行程单，到道酒店再发给他们，那个排版得弄好呀。", "clean": "今天这行程，我觉得与其在市区逛，不如直接打滴滴去博物馆，比挤地铁舒服多了。刚才算了下，打车花了62元。再者，还得把之前在京东买的充电宝给周主任，明天见面对比发快递更稳当吧。对了，还得用Canva改一下行程单，到酒店再发给他们，那个排版得弄好呀。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12357.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 689, "source_record_id": "assigned_records.jsonl:12357"}
{"oral": "诶，你说婚礼视觉评审得定在几号，我想想啊，得在电话回访完之后，然后记得预约美容，怎么说呢，还得准备个PPT，应该得定在下周三吧。", "clean": "你说婚礼视觉评审得定在几号，得在电话回访完之后，然后记得预约美容，还得准备个PPT，应该得定在下周三吧。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12449.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 693, "source_record_id": "assigned_records.jsonl:12449"}
{"oral": "吴总说这些旧设备捐掉捐掉比较好，我觉得挺好，得赶紧腾空间。先把那个小王给我的企业微信文档找出来，得核对核对清单。韩总之前提到审稿那个文件夹，好像在那个旧硬盘里，得一起捐了。那个硬盘大概有五百个G，五百个G，应该能用，先放进箱子吧。", "clean": "吴总说这些旧设备捐掉比较好，我觉得挺好，得赶紧腾空间。先把小王给我的企业微信文档找出来，得核对清单。韩总之前提到审稿那个文件夹，好像在旧硬盘里，得一起捐了。那个硬盘大概有500G，应该能用，先放进箱子吧。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12360.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 690, "source_record_id": "assigned_records.jsonl:12360"}
{"oral": "哎呀真是没完没了，还没收拾完就快下午三点半，嗯，得赶紧把同步API文档打印出来，怎么说呢，还得在Coursera下个课件，诶，对了，刚才是不是忘了遛狗，我想想啊，Signal收到消息说要重新绘制流程图，吧。", "clean": "真是没完没了，还没收拾完就快15:30，得赶紧把同步API文档打印出来，还得在Coursera下个课件。对了，刚才是不是忘了遛狗？Signal收到消息说要重新绘制流程图。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12423.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 692, "source_record_id": "assigned_records.jsonl:12423"}
{"oral": "老李，记得帮我买瓶生抽，还有那个白胡椒粉，不对，是黑胡椒粉，买两瓶。哦对了，韩医生说报销得在周五前提交，你记得在 GitHub 那个文档里对一下，再发到企微给我，记得发到新苑那个群里。", "clean": "老李，记得帮我买瓶生抽，还有黑胡椒粉，买2瓶。哦对了，韩医生说报销得在周五前提交，你记得在GitHub那个文档里对一下，再发到企微给我，记得发到新苑那个群里。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12386.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 691, "source_record_id": "assigned_records.jsonl:12386"}
{"oral": "如果打车能直接对接Foxmail，那…哎怎么说，如果能直接起草邮件的话，咱们就能在Microsoft Teams再撰写指南，省得得手动导出了。", "clean": "如果打车能直接对接Foxmail，如果能直接起草邮件的话，咱们就能在Microsoft Teams撰写指南，省得手动导出了。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12455.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 694, "source_record_id": "assigned_records.jsonl:12455"}
{"oral": "嗯，如果把这些数据先导进 Google Sheets 话，Google Sheets 吧，那咱们就能快速汇总，对了，办理年检，不对，是寄快递那个流程，我想想啊，如果把 Logseq 里的笔记，笔记直接用 Google Translate 翻译，翻译成英文的话，那我们就能直接发给外方了呀。", "clean": "如果把这些数据先导进Google Sheets，那咱们就能快速汇总。关于寄快递那个流程，如果把Logseq里的笔记直接用Google Translate翻译成英文的话，那我们就能直接发给外方了。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12480.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 695, "source_record_id": "assigned_records.jsonl:12480"}
{"oral": "明天早上九点得先看钉钉，钉钉上那个测试工程师发的消息，再一个，我得把 Bank of China 那个报表导出来，导出来核对一遍。我也不太确定，但是感觉数据好像有问题，怎么说呢，如果把对比维度换一下可能更准。哦不对，是换成月度维度。而且而且明天下午两点半还得开会，得赶紧准备准备。", "clean": "明天早上9:00得先看钉钉，钉钉上测试工程师发的消息。再一个，我得把 Bank of China 那个报表导出来核对一遍。我不太确定，但感觉数据好像有问题，如果把对比维度换成月度维度可能更准。而且明天14:30还得开会，得赶紧准备准备。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12482.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 696, "source_record_id": "assigned_records.jsonl:12482"}
{"oral": "对了，明天得处理处理工单，处理工单用 PDF Expert 导出，哦对了，下午三点半开会，开会之前给何司机打个电话，话说 ABC 那个文件得再传一遍，传一遍到邮箱里。", "clean": "明天得处理处理工单，用 PDF Expert 导出。下午15:30开会，开会之前给何司机打个电话。ABC那个文件得再传一遍到邮箱里。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12530.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 698, "source_record_id": "assigned_records.jsonl:12530"}
{"oral": "嗯，邻居那边设计原画，设计原画是用阿里云的，啊不对，是Google Cloud，就是...怎么说呢，包括包括所有资源都迁移，你觉得呢吧？", "clean": "邻居那边设计原画是用Google Cloud的，包括所有资源都迁移，你觉得呢？", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12542.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 699, "source_record_id": "assigned_records.jsonl:12542"}
{"oral": "周总说要重视项目复盘，项目复盘。我觉得首先得审核预算，审核预算。预算是五千块，不对，六千块，等等错了，是八千块！林护士那边得确认下。", "clean": "周总说要重视项目复盘。我觉得首先得审核预算，预算是8000元，林护士那边得确认下。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12568.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 701, "source_record_id": "assigned_records.jsonl:12568"}
{"oral": "其实那个系统巡检不是单纯靠数据分析师盯着看就行，啊不对，是得让产品助理先跑一遍流程。那个...我想想...要是能把WhatsApp的消息接口，怎么说呢，直接集成到整理笔记模块里，这样就不用每次手动导数据，整体效率起码能提高百分之二十五。", "clean": "其实系统巡检不是单纯靠数据分析师盯着看就行，是得让产品助理先跑一遍流程。要是能把WhatsApp的消息接口直接集成到整理笔记模块里，这样就不用每次手动导数据，整体效率起码能提高25%。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12563.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 700, "source_record_id": "assigned_records.jsonl:12563"}
{"oral": "我想把那个Lark复盘总结，复盘总结直接发给所有人，不对，先发给主管。其次，电话回访，电话回访这块得赶紧安排，安排个时间组织会议，大概下午三点半开，得在周五之前把这个流程跑通，这次项目到道了关键节点，得赶紧定下来。", "clean": "我想把Lark复盘总结先发给主管。其次，电话回访这块得赶紧安排个时间组织会议，大概15:30开，得在周五之前把这个流程跑通，这次项目到了关键节点，得赶紧定下来。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12677.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 703, "source_record_id": "assigned_records.jsonl:12677"}
{"oral": "嗯，先在Google Calendar标下跑步，不对，是先安排面试，然后对应的那个...我想想啊...最后得把火候调小吧。", "clean": "先安排面试，最后得把火候调小吧。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12756.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 705, "source_record_id": "assigned_records.jsonl:12756"}
{"oral": "诶，太离谱了，就是这家店在滨江路，啊不，在麓山路，我想想啊，得在三层，嗯，应该在三层，怎么说呢，在那家葑门路新开的那个商场里吧。", "clean": "太离谱了，这家店在麓山路，在3层，在那家葑门路新开的商场里吧。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl16927.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 707, "source_record_id": "assigned_records.jsonl:16927"}
{"oral": "嗯，第二套不行，就是那个，韩秘书韩秘书说，得交五百块...不对，四百块...等等错了，是三百块网费，嘛，而且老婆之前那个前任同事卫行政...卫行政在，哎，住这，太吵吧。", "clean": "第二套不行，韩秘书说得交300元网费，而且老婆之前的前任同事卫行政住这，太吵吧。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12590.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 702, "source_record_id": "assigned_records.jsonl:12590"}
{"oral": "说白了，去崟山，嗯，走寰宇大道，别走长安街，对了，改走隽永路吧。", "clean": "说白了，去崟山走隽永路，别走长安街。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl16930.wav", "scene": "navigation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 708, "source_record_id": "assigned_records.jsonl:16930"}
{"oral": "诶，我想起来，得赶紧更新简历，更新简历，把最近那个市场调研项目加上。嗯，其实主要 主要是想把怎么清理垃圾数据的逻辑写清楚，怎么说，总之得让老赵看到我处理海量数据能力强。诶不对，应该是写在第三页，第三页，把那个百分之十二的提升率标出来，这样更有说服力吧。", "clean": "我想起来，得赶紧更新简历，把最近那个市场调研项目加上。其实主要想把清理垃圾数据的逻辑写清楚，总之得让老赵看到我处理海量数据的能力强。应该是写在第三页，把12%的提升率标出来，这样更有说服力吧。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12724.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 704, "source_record_id": "assigned_records.jsonl:12724"}
{"oral": "如果把那个战略讨论的PDF发给Kevin的话，那我们就能在会议上管理库存，管理库存得更高效，哎，那个发给Kevin，不对，发给Linda，这样她能提前看。要是能把那个Outlook日程再，怎么说呢，如果把Outlook日程提前同步好，同步好，那我就不用在路上再查一遍，查一遍那个时间点在没在行程单里。", "clean": "如果把战略讨论的PDF发给Linda的话，那我们就能在会议上管理库存得更高效，这样她能提前看。如果把Outlook日程提前同步好，那我就不用在路上再查一遍时间点在没在行程单里。", "usage_text_words": null, "audio": "audio/dictation_memo__assigned_recordsjsonl12483.wav", "scene": "dictation_memo", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 697, "source_record_id": "assigned_records.jsonl:12483"}
{"oral": "诶，不是往南锣鼓巷走，出口出口其实在双棋路，啊不对，是栖霞路吧，您得往栖霞路栖霞路那边走哈。", "clean": "不是往南锣鼓巷走，出口其实在栖霞路，您得往栖霞路那边走哈。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl16935.wav", "scene": "navigation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 709, "source_record_id": "assigned_records.jsonl:16935"}
{"oral": "说白了，嗯，走璟睿大到吧，我想想啊，绕过汊河，不对，是走濉江，就是...直接开到裴家店那边，诶，应该快一点吧。", "clean": "说白了，走璟睿大道吧，绕过濉江，直接开到裴家店那边，应该快一点吧。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl16943.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 710, "source_record_id": "assigned_records.jsonl:16943"}
{"oral": "大爷您好，请问最近的医院怎么走，是从潖口方向还是往閖邑那边走，或者是经过琠山，不对，请问去南锣鼓巷最近的医院怎么走？", "clean": "大爷您好，请问去南锣鼓巷最近的医院怎么走？", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17006.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 712, "source_record_id": "assigned_records.jsonl:17006"}
{"oral": "诶，我这会儿在涠洲岛那边，我想想啊，在麒麟路一零二号，不对，一零三号，等等错了，是一零四号，您直接开到环球路御道街口哈！", "clean": "我这会儿在涠洲岛那边，在麒麟路104号，您直接开到环球路御道街口哈！", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17028.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 713, "source_record_id": "assigned_records.jsonl:17028"}
{"oral": "诶，请问出口在御道街，啊不是，麒麟路那边？我想再到...怎么说呢，往建国门外大结走走得通嘛？", "clean": "请问出口在麒麟路那边？我想往建国门外大街走走得通吗？", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17043.wav", "scene": "navigation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 714, "source_record_id": "assigned_records.jsonl:17043"}
{"oral": "王哥，我这会儿在汊河站，在北出站口，哦不是，是南出站口，你往这边走，我停在那个…怎么说呢…就在广济路那边那个掉头口，你快出来吧。", "clean": "王哥，我这会儿在汊河站南出站口，你往这边走，我停在广济路那边那个掉头口，你快出来吧。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17047.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 715, "source_record_id": "assigned_records.jsonl:17047"}
{"oral": "我登机口在曙光路，曙光路那边，得走麒麟路再绕道御道街，御道街走过去，你觉得这样走对不对？", "clean": "我登机口在曙光路，得走麒麟路再绕道御道街走过去，你觉得这样走对不对？", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17060.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 717, "source_record_id": "assigned_records.jsonl:17060"}
{"oral": "诶，车在梓州路，就是梓州路这边，因为檖梓街那边封路，怎么说呢，我想想啊，车停在汊河路，不对，是汊河边上，吧，因为中山大道那边太堵，太堵，得绕路走哈。", "clean": "车在梓州路这边，因为檖梓街那边封路，车停在汊河边上，因为中山大道那边太堵，得绕路走哈。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17051.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 716, "source_record_id": "assigned_records.jsonl:17051"}
{"oral": "诶，如果潖口那边潖口有啥好吃的，那我们就往汊口走，啊不，是往栖霞路走，我想想啊，要是那个，要是暎照路那边有那种，嗯，有名店的话，咱们直接打车去道那儿吧。", "clean": "如果汊口那边有啥好吃的，那我们就往栖霞路走。我想想，要是映照路那边有有名店的话，咱们直接打车去那儿吧。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17177.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 719, "source_record_id": "assigned_records.jsonl:17177"}
{"oral": "我刚到这儿，想去御道街御道街逛逛，听说葑门葑门那边有卖籼米，在金街那边是不是也有好吃的？你觉得翊君路道这个位置近不近？", "clean": "我刚到这儿，想去御道街逛逛，听说葑门那边有卖籼米，在金街那边是不是也有好吃的？你觉得翊君路这个位置近不近？", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17249.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 721, "source_record_id": "assigned_records.jsonl:17249"}
{"oral": "诶，对了，这家店在共和路还是南京路哈，我想想啊，是在汊口那边三层吧？", "clean": "这家店在共和路还是南京路，是在汊口那边3层吧？", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17319.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 723, "source_record_id": "assigned_records.jsonl:17319"}
{"oral": "师傅去麒麟路，麒麟路那个茑屋。哦对了，走崇礼路走崇礼路，别走潖口，潖口那道堵得厉害，对了得走汊河路汊河路再绕过去。", "clean": "师傅去麒麟路茑屋。走崇礼路，别走潖口，潖口那道堵得厉害，得走汊河路再绕过去。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17214.wav", "scene": "navigation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 720, "source_record_id": "assigned_records.jsonl:17214"}
{"oral": "简单来说，你从靖康路那边进来，过了濉江，嗯，就在汊口路拐弯，诶，我想想啊，你直接走崇礼路，吧，往崀山那个方向开，就是，应该是二号门进。", "clean": "简单来说，你从靖康路那边进来，过了濉江，在汊口路拐弯，你直接走崇礼路，往崀山那个方向开，应该是2号门进。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17260.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 722, "source_record_id": "assigned_records.jsonl:17260"}
{"oral": "咱们下服务区下服务区吧，我得去趟梓州路，哦不是，是靖安路，你开到开到那个珞珈山附近再停，得赶紧走，不然朝阳门内大街那边得堵死。", "clean": "咱们下服务区吧，我得去趟靖安路，你开到珞珈山附近再停，得赶紧走，不然朝阳门内大街那边得堵死。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17322.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 725, "source_record_id": "assigned_records.jsonl:17322"}
{"oral": "快递员，你走崇光路崇光路，转到寰宇大到，到了麓山大道麓山大道左转，在皛曄苑门口停，你再看下位置。", "clean": "快递员，你走崇光路，转到寰宇大道，到了麓山大道左转，在皛曄苑门口停，你再看下位置。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17510.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 728, "source_record_id": "assigned_records.jsonl:17510"}
{"oral": "要是那个，走霄云路那边能直接到的话，那我们就能，额，换个说，如果能直接走潏江路过去的话，是不是快一点？", "clean": "如果能直接走潏江路过去的话，是不是快一点？", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17540.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 729, "source_record_id": "assigned_records.jsonl:17540"}
{"oral": "诶，帮我指下登机口怎么走，就是我得赶快去，因为我之前在潏江那边走错了，还绕到承天大道跟崀山那边去，怎么说呢，感觉像在买籼米一样乱转，现在得赶紧去栖霞路那个口吧。", "clean": "帮我指下登机口怎么走，我得赶快去，因为我之前在潏江那边走错了，还绕到承天大道跟崀山那边去，感觉像在买籼米一样乱转，现在得赶紧去栖霞路那个口吧。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17501.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 727, "source_record_id": "assigned_records.jsonl:17501"}
{"oral": "我现在在璟南路璟南路这边，不对，是鲁迅路，鲁迅路这边出站口等你，你觉得你觉得怎么走快？", "clean": "我现在在鲁迅路这边出站口等你，你觉得怎么走快？", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17542.wav", "scene": "navigation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 730, "source_record_id": "assigned_records.jsonl:17542"}
{"oral": "诶，去积水潭得坐哪路，是坐动车转芍药居吗？我想想，得在钟楼下车，体育西路那边太远，还是走这一路吧。", "clean": "去积水潭得坐哪路，是坐动车转芍药居吗？得在钟楼下车，体育西路那边太远，还是走这一路吧。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17794.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 732, "source_record_id": "assigned_records.jsonl:17794"}
{"oral": "先是坐火车到站，然后沿着人行道走，诶我想用打车软件叫车结果信号不好，感觉像在南锣鼓巷那样绕圈，最后绕到内环线那边才找着，一共走了大概两千三百步，快下午三点半了才到这栋楼，就是没找着科室。", "clean": "先是坐火车到站，然后沿着人行道走，我想用打车软件叫车结果信号不好，感觉像在南锣鼓巷那样绕圈，最后绕到内环线那边才找着，一共走了大概2300步，快15:30了才到这栋楼，没找着科室。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17733.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 731, "source_record_id": "assigned_records.jsonl:17733"}
{"oral": "师傅，去寰宇大到，呃，我也不太确定，但是感觉走滎阳那边快点，对，过个环球路，别走建国门外大街，再，啊不，走崟山那条路，到吧。", "clean": "师傅，去寰宇大道，我也不太确定，但是感觉走荥阳那边快点，对，过个环球路，别走建国门外大街，走玉山那条路到吧。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17449.wav", "scene": "navigation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 726, "source_record_id": "assigned_records.jsonl:17449"}
{"oral": "诶，对了，周边周边有什么好吃的嘛？走这条小路走这条小路能飞过去春熙路吗？嗯，员村那边有自行车道没？", "clean": "周边有什么好吃的吗？走这条小路能到春熙路吗？员村那边有自行车道没？", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17803.wav", "scene": "navigation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 733, "source_record_id": "assigned_records.jsonl:17803"}
{"oral": "诶，直接往山顶走，就是往山顶走，因为武汉站那边堵，走体育西路，不对，走体育西道，嗯，在大街...在大街这边掉头再掉头哈。", "clean": "直接往山顶走，因为武汉站那边堵，走体育西道，在大街这边掉头哈。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17807.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 734, "source_record_id": "assigned_records.jsonl:17807"}
{"oral": "诶，先走海淀路，然后骑蓝单车钻支路，怎么说呢，最后走北三环再到郊区吧，开私家车得堵死呀。", "clean": "先走海淀路，然后骑蓝单车钻支路，最后走北三环再到郊区吧，开私家车得堵死呀。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17912.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 739, "source_record_id": "assigned_records.jsonl:17912"}
{"oral": "不是在江汉路，诶，是在友谊大街封闭路后面。", "clean": "是在友谊大街封闭路后面。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17926.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 741, "source_record_id": "assigned_records.jsonl:17926"}
{"oral": "诶，对了，你到解放碑路，解放碑路往东偏北走，嗯，过了密云路，密云路这边有个分隔带，就是得绕开人力车吧。", "clean": "你到解放碑路往东偏北走，过了密云路，密云路这边有个分隔带，得绕开人力车吧。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17849.wav", "scene": "navigation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 736, "source_record_id": "assigned_records.jsonl:17849"}
{"oral": "哦对了，你走通州路，走通州路，在下一个路口转弯，转弯去经纬路，不对，是滨江路，你…那边…要是走滨江路，滨江路就没那么堵。", "clean": "你走通州路，在下一个路口转弯去滨江路，走滨江路就没那么堵。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17917.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 740, "source_record_id": "assigned_records.jsonl:17917"}
{"oral": "师傅，去橘子洲，走解放碑那条路，不对，是走底道走到底，再走一条近路快点，你觉得行不？", "clean": "师傅，去橘子洲，走底道走到底，再走一条近路快点，你觉得行不？", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17976.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 742, "source_record_id": "assigned_records.jsonl:17976"}
{"oral": "老王，诶，你得走这条路，一直往前走一直往前走，到环岛路拐弯，记得记得打车到陵园苑，别走主路，走立交桥那边快，怎么说呢，大概能省十五分钟吧，只要在在立交桥右转就到了哈。", "clean": "老王，你得走这条路，一直往前走，到环岛路拐弯，记得打车到陵园苑，别走主路，走立交桥那边快，大概能省15分钟吧，只要在立交桥右转就到了哈。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17857.wav", "scene": "navigation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 737, "source_record_id": "assigned_records.jsonl:17857"}
{"oral": "你好，我想问问登机口怎么走，往东北方向走，往东北方向走能到吗，还是得走那条道，走那条道，我想想，我们要去体育西路，哦不是，是去机场北区，如果走那边的话，得乘公交，啊不，是得坐接驳车，怎么说，得坐摆渡车才能到吧？", "clean": "你好，我想问问登机口怎么走，往东北方向走能到吗，还是得走那条道？我们要去机场北区，如果走那边的话，得坐摆渡车才能到吧？", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl17816.wav", "scene": "navigation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 735, "source_record_id": "assigned_records.jsonl:17816"}
{"oral": "诶，我车在体育西路，就就就是，我想想啊，在那个高新园那边，你接单过来吧。", "clean": "我车在体育西路，在高新园那边，你接单过来吧。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl18108.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 747, "source_record_id": "assigned_records.jsonl:18108"}
{"oral": "诶，重庆北往南偏西，往南偏西走，在辅路，啊不对，龙冬大道辅路，就是中关村那边，哈，你觉得能到吗？", "clean": "重庆北往南偏西走，在龙冬大道辅路，中关村那边，你觉得能到吗？", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl18098.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 746, "source_record_id": "assigned_records.jsonl:18098"}
{"oral": "哦对了，我这会儿在二号出站口这边，就是那个天桥下面哈，你赶紧过来，我刚才想约车结果发现堵死，直接开车去接你，话说中环路现在快瘫痪了，我想想啊，你得从南面走，不对，得从北面出站，不然得绕好久，你赶紧看一眼屏幕，马上乘机时间紧，快点啊！", "clean": "我这会儿在2号出站口这边，天桥下面，你赶紧过来。我刚才想约车结果发现堵死，直接开车去接你。中环路现在快瘫痪了，你得从北面出站，不然得绕好久。你赶紧看一眼屏幕，马上乘机时间紧，快点啊！", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl18001.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 743, "source_record_id": "assigned_records.jsonl:18001"}
{"oral": "诶，太离谱了，我从黄埔大道走，黄埔大道走过来，就是经过南滨路，啊不对，是单行道，怎么说呢，还得搭轮渡到南滨路，走到冬门那边吧。", "clean": "太离谱了，我从黄埔大道走过来，经过单行道，还得搭轮渡到南滨路，走到东门那边吧。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl18113.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 748, "source_record_id": "assigned_records.jsonl:18113"}
{"oral": "先是公交道五一广场，然后走体育西路，诶，再走成雅高速，最后得在体育西路找那栋楼，我想想啊，得是三号楼吧。", "clean": "先是公交到五一广场，然后走体育西路，再走成雅高速，最后在体育西路找3号楼。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl18123.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 749, "source_record_id": "assigned_records.jsonl:18123"}
{"oral": "大爷，呼家楼那边...工业园区...工业园区怎么走？对了，右拐一百米，右拐一百米到汉口站，我觉得行。", "clean": "大爷，呼家楼那边工业园区怎么走？右拐100米到汉口站，我觉得行。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl18129.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 750, "source_record_id": "assigned_records.jsonl:18129"}
{"oral": "说白了，我想想啊，出口应该是直走直走，从工业路那边出去，诶，应该是从楚天大苑，不对，楚天大道走，然后坐摆渡车，摆渡车到开放路，怎么说呢，再直走三公里吧。", "clean": "说白了，出口应该是直走，从工业路那边出去，应该是从楚天大道走，然后坐摆渡车到开放路，再直走3公里吧。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl18070.wav", "scene": "navigation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 744, "source_record_id": "assigned_records.jsonl:18070"}
{"oral": "嗯，我刚从西安南站过来，就在陵园路这边，我想想啊，这附近…怎么说呢，有什么好吃的餐厅吧？就那种在观音桥附近得口味好的呀。", "clean": "我刚从西安南站过来，就在陵园路这边，这附近有什么好吃的餐厅吧？就那种在观音桥附近口味好的。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl18142.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 751, "source_record_id": "assigned_records.jsonl:18142"}
{"oral": "他说体育西路右转一百米到，诶，我看南锣鼓巷那边得徒步走长安大街才行吧。", "clean": "他说体育西路右转100米到，我看南锣鼓巷那边得徒步走长安大街才行吧。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl18308.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 755, "source_record_id": "assigned_records.jsonl:18308"}
{"oral": "诶，先是骑车骑车到成龙路，然后走走两公里，最后走脚程走脚程到那道车站吧，怎么说呢，接单快点。", "clean": "先骑车到成龙路，然后走2公里，最后步行到车站，接单快点。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl18290.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 754, "source_record_id": "assigned_records.jsonl:18290"}
{"oral": "Play the latest album by Taylor Swift on Spotify.", "clean": "Play the latest album by Taylor Swift on Spotify.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18599.wav", "scene": "passthrough", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 757, "source_record_id": "assigned_records.jsonl:18599"}
{"oral": "他说在东四那个出口，不对，不是，哎呀那个河西大到...对了，我这车停在武侯祠那边，您帮我看看是不是在观音桥路那块儿，我得赶紧叫辆车走，我都找了二十分钟了，在那新园子里找不到", "clean": "我这车停在武侯祠那边，您帮我看看是不是在观音桥路那块儿，我得赶紧叫辆车走，我都找了20分钟了，在那新园子里找不到", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl18204.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 753, "source_record_id": "assigned_records.jsonl:18204"}
{"oral": "老王，环湖路那边...堵得死死，往往往回走吧，走高新园那边，比体育西路快，东单那边...也堵，快四点啦。", "clean": "老王，环湖路那边堵得死死的，往回走吧，走高新园那边，比体育西路快，东单那边也堵，快16:00了。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl18336.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 756, "source_record_id": "assigned_records.jsonl:18336"}
{"oral": "Could you please update the project timeline in the shared spreadsheet and notify the stakeholders by 5:00pm today?", "clean": "Could you please update the project timeline in the shared spreadsheet and notify the stakeholders by 5:00pm today?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18605.wav", "scene": "passthrough", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 758, "source_record_id": "assigned_records.jsonl:18605"}
{"oral": "你刚才说往东走五百米能快，我觉得往东走五百米不行，那边现在是封闭路，封闭路得绕好远。我看地图说坐客车，哦不是，乘轻轨最稳当，乘轻轨直接能到郊区，或者咱们走快捷通道，走快捷通道避开市区堵车，比坐客车快多了，你快看一眼现在几点，是不是得赶紧出发了。", "clean": "你刚才说往东走500米能快，我觉得往东走500米不行，那边现在是封闭路，封闭路得绕好远。我看地图说乘轻轨最稳当，乘轻轨直接能到郊区，或者咱们走快捷通道，走快捷通道避开市区堵车，比坐客车快多了，你快看一眼现在几点，是不是得赶紧出发了。", "usage_text_words": null, "audio": "audio/navigation__assigned_recordsjsonl18156.wav", "scene": "navigation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 752, "source_record_id": "assigned_records.jsonl:18156"}
{"oral": "Could you please update the Jira ticket for the authentication bug and link it to the latest pull request?", "clean": "Could you please update the Jira ticket for the authentication bug and link it to the latest pull request?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18613.wav", "scene": "passthrough", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 760, "source_record_id": "assigned_records.jsonl:18613"}
{"oral": "Could you please update the deployment script to use the latest Docker image and set the memory limit to 4GB?", "clean": "Could you please update the deployment script to use the latest Docker image and set the memory limit to 4GB?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18608.wav", "scene": "passthrough", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 759, "source_record_id": "assigned_records.jsonl:18608"}
{"oral": "I'm thinking of grabbing some tacos for dinner around 7pm. Do you want to join me, or are you too busy?", "clean": "I'm thinking of grabbing some tacos for dinner around 7pm. Do you want to join me, or are you too busy?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18614.wav", "scene": "passthrough", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 761, "source_record_id": "assigned_records.jsonl:18614"}
{"oral": "Set a reminder for tomorrow at 10am to call the insurance company regarding the claim for the water damage.", "clean": "Set a reminder for tomorrow at 10am to call the insurance company regarding the claim for the water damage.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18616.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 762, "source_record_id": "assigned_records.jsonl:18616"}
{"oral": "The quarterly growth rate increased by 12% compared to the last fiscal year, primarily driven by our expansion into the Southeast Asian market.", "clean": "The quarterly growth rate increased by 12% compared to the last fiscal year, primarily driven by our expansion into the Southeast Asian market.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18625.wav", "scene": "passthrough", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 763, "source_record_id": "assigned_records.jsonl:18625"}
{"oral": "Play some lo-fi hip hop on Spotify and dim the living room lights to 30%.", "clean": "Play some lo-fi hip hop on Spotify and dim the living room lights to 30%.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18630.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 764, "source_record_id": "assigned_records.jsonl:18630"}
{"oral": "Set a timer for 15 minutes and dim the living room lights to 30%.", "clean": "Set a timer for 15 minutes and dim the living room lights to 30%.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18637.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 766, "source_record_id": "assigned_records.jsonl:18637"}
{"oral": "I'm feeling a bit overwhelmed with work lately, I think I need a short vacation this weekend to recharge.", "clean": "I'm feeling a bit overwhelmed with work lately, I think I need a short vacation this weekend to recharge.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18638.wav", "scene": "passthrough", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 767, "source_record_id": "assigned_records.jsonl:18638"}
{"oral": "My order number is 88492017, and I'd like to check if the shipping status has been updated to 'delivered'.", "clean": "My order number is 88492017, and I'd like to check if the shipping status has been updated to 'delivered'.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18635.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 765, "source_record_id": "assigned_records.jsonl:18635"}
{"oral": "The current latency in the production environment is spiking to 500ms, we need to check the Redis cache hit rate immediately.", "clean": "The current latency in the production environment is spiking to 500ms, we need to check the Redis cache hit rate immediately.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18639.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 768, "source_record_id": "assigned_records.jsonl:18639"}
{"oral": "I'll be about 15 minutes late because of the traffic on I-95, just start the meeting without me.", "clean": "I'll be about 15 minutes late because of the traffic on I-95, just start the meeting without me.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18641.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 769, "source_record_id": "assigned_records.jsonl:18641"}
{"oral": "Could you please check the status of order 8829104? It was supposed to arrive two days ago.", "clean": "Could you please check the status of order 8829104? It was supposed to arrive two days ago.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18647.wav", "scene": "passthrough", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 770, "source_record_id": "assigned_records.jsonl:18647"}
{"oral": "Please update the production environment to version 2.4.1 and restart the Nginx service.", "clean": "Please update the production environment to version 2.4.1 and restart the Nginx service.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18650.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 772, "source_record_id": "assigned_records.jsonl:18650"}
{"oral": "The current latency of the API endpoint is approximately 250ms, which exceeds our SLA of 100ms.", "clean": "The current latency of the API endpoint is approximately 250ms, which exceeds our SLA of 100ms.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18648.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 771, "source_record_id": "assigned_records.jsonl:18648"}
{"oral": "The attention mechanism in the Transformer architecture allows the model to focus on different parts of the input sequence.", "clean": "The attention mechanism in the Transformer architecture allows the model to focus on different parts of the input sequence.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18652.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 774, "source_record_id": "assigned_records.jsonl:18652"}
{"oral": "Set a reminder for 6:00pm tomorrow to call the dentist and reschedule my appointment for next Tuesday.", "clean": "Set a reminder for 6:00pm tomorrow to call the dentist and reschedule my appointment for next Tuesday.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18659.wav", "scene": "passthrough", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 775, "source_record_id": "assigned_records.jsonl:18659"}
{"oral": "My order number is 88293410, but the tracking status hasn't updated for 3 days.", "clean": "My order number is 88293410, but the tracking status hasn't updated for 3 days.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18651.wav", "scene": "passthrough", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 773, "source_record_id": "assigned_records.jsonl:18651"}
{"oral": "I need to schedule a meeting with the design team for next Tuesday at 2:00pm to review the Figma prototypes.", "clean": "I need to schedule a meeting with the design team for next Tuesday at 2:00pm to review the Figma prototypes.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18668.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 777, "source_record_id": "assigned_records.jsonl:18668"}
{"oral": "The current implementation of the transformer model is causing a memory leak during the backward pass of the training loop.", "clean": "The current implementation of the transformer model is causing a memory leak during the backward pass of the training loop.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18664.wav", "scene": "passthrough", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 776, "source_record_id": "assigned_records.jsonl:18664"}
{"oral": "Could you please update the status of order 88291 and check if the shipping address was corrected?", "clean": "Could you please update the status of order 88291 and check if the shipping address was corrected?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18670.wav", "scene": "passthrough", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 778, "source_record_id": "assigned_records.jsonl:18670"}
{"oral": "Turn off the living room lights and set the thermostat to 22 degrees.", "clean": "Turn off the living room lights and set the thermostat to 22 degrees.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18696.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 782, "source_record_id": "assigned_records.jsonl:18696"}
{"oral": "Could you explain the difference between the Transformer architecture and traditional RNNs in terms of parallelization?", "clean": "Could you explain the difference between the Transformer architecture and traditional RNNs in terms of parallelization?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18681.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 780, "source_record_id": "assigned_records.jsonl:18681"}
{"oral": "The current model's F1 score has improved by 2.5% after applying the new regularization technique described in the 2023 paper.", "clean": "The current model's F1 score has improved by 2.5% after applying the new regularization technique described in the 2023 paper.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18672.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 779, "source_record_id": "assigned_records.jsonl:18672"}
{"oral": "The p-value of the experimental group is 0.03, which indicates a statistically significant result at the 5% level.", "clean": "The p-value of the experimental group is 0.03, which indicates a statistically significant result at the 5% level.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18684.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 781, "source_record_id": "assigned_records.jsonl:18684"}
{"oral": "Regarding the Q3 financial report, we need to analyze the correlation between user acquisition costs and the lifetime value of the customer.", "clean": "Regarding the Q3 financial report, we need to analyze the correlation between user acquisition costs and the lifetime value of the customer.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18697.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 783, "source_record_id": "assigned_records.jsonl:18697"}
{"oral": "Set a reminder for 8:00pm tonight to take the laundry out of the washing machine.", "clean": "Set a reminder for 8:00pm tonight to take the laundry out of the washing machine.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18721.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 786, "source_record_id": "assigned_records.jsonl:18721"}
{"oral": "Can you check the status of order number 8829401 and tell me if it has shipped yet?", "clean": "Can you check the status of order number 8829401 and tell me if it has shipped yet?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18711.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 785, "source_record_id": "assigned_records.jsonl:18711"}
{"oral": "The latest benchmark results show that the Transformer-based model outperforms the previous baseline by 12% in terms of F1 score.", "clean": "The latest benchmark results show that the Transformer-based model outperforms the previous baseline by 12% in terms of F1 score.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18705.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 784, "source_record_id": "assigned_records.jsonl:18705"}
{"oral": "Remind me to buy milk and pick up the dry cleaning tomorrow at 9:00am.", "clean": "Remind me to buy milk and pick up the dry cleaning tomorrow at 9:00am.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18732.wav", "scene": "passthrough", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 787, "source_record_id": "assigned_records.jsonl:18732"}
{"oral": "How do I configure the environment variables for a React project using Vite?", "clean": "How do I configure the environment variables for a React project using Vite?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18735.wav", "scene": "passthrough", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 788, "source_record_id": "assigned_records.jsonl:18735"}
{"oral": "Navigate to the nearest Starbucks on Broadway and avoid tolls if possible.", "clean": "Navigate to the nearest Starbucks on Broadway and avoid tolls if possible.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18747.wav", "scene": "passthrough", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 791, "source_record_id": "assigned_records.jsonl:18747"}
{"oral": "Could you please update the project timeline for the Q3 deliverables and send it to the stakeholders by 5:00pm tomorrow?", "clean": "Could you please update the project timeline for the Q3 deliverables and send it to the stakeholders by 5:00pm tomorrow?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18742.wav", "scene": "passthrough", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 790, "source_record_id": "assigned_records.jsonl:18742"}
{"oral": "Who won the 2024 Academy Award for Best Picture?", "clean": "Who won the 2024 Academy Award for Best Picture?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18752.wav", "scene": "passthrough", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 792, "source_record_id": "assigned_records.jsonl:18752"}
{"oral": "Please update the Jira ticket to 'In Progress' and assign the bug fix to the backend team.", "clean": "Please update the Jira ticket to 'In Progress' and assign the bug fix to the backend team.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18736.wav", "scene": "passthrough", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 789, "source_record_id": "assigned_records.jsonl:18736"}
{"oral": "We should pivot the current sprint to prioritize the security patches before we release the beta version to the public.", "clean": "We should pivot the current sprint to prioritize the security patches before we release the beta version to the public.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18760.wav", "scene": "passthrough", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 794, "source_record_id": "assigned_records.jsonl:18760"}
{"oral": "I need to schedule a 1:1 sync with the product manager to discuss the Q3 roadmap and resource allocation.", "clean": "I need to schedule a 1:1 sync with the product manager to discuss the Q3 roadmap and resource allocation.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18757.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 793, "source_record_id": "assigned_records.jsonl:18757"}
{"oral": "Could you please update the Jira ticket for the API integration issue and assign it to the backend team?", "clean": "Could you please update the Jira ticket for the API integration issue and assign it to the backend team?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18782.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 796, "source_record_id": "assigned_records.jsonl:18782"}
{"oral": "Please update the Kubernetes cluster to version 1.28 and restart the ingress controller.", "clean": "Please update the Kubernetes cluster to version 1.28 and restart the ingress controller.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18775.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 795, "source_record_id": "assigned_records.jsonl:18775"}
{"oral": "Can you remind me to submit the quarterly budget report by 5:00pm this Friday?", "clean": "Can you remind me to submit the quarterly budget report by 5:00pm this Friday?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18796.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 799, "source_record_id": "assigned_records.jsonl:18796"}
{"oral": "Remind me to pick up the dry cleaning and buy some almond milk at 6:30pm tonight.", "clean": "Remind me to pick up the dry cleaning and buy some almond milk at 6:30pm tonight.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18814.wav", "scene": "passthrough", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 801, "source_record_id": "assigned_records.jsonl:18814"}
{"oral": "I need to draft a summary of today's sprint planning meeting, focusing on the three main blockers identified by the frontend team.", "clean": "I need to draft a summary of today's sprint planning meeting, focusing on the three main blockers identified by the frontend team.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18811.wav", "scene": "passthrough", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 800, "source_record_id": "assigned_records.jsonl:18811"}
{"oral": "I think we should prioritize the API integration before moving on to the frontend UI polish.", "clean": "I think we should prioritize the API integration before moving on to the frontend UI polish.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18792.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 798, "source_record_id": "assigned_records.jsonl:18792"}
{"oral": "I'm looking for a high-resolution PDF of the 2023 Transformer architecture survey paper, specifically focusing on the attention mechanism metrics.", "clean": "I'm looking for a high-resolution PDF of the 2023 Transformer architecture survey paper, specifically focusing on the attention mechanism metrics.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18783.wav", "scene": "passthrough", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 797, "source_record_id": "assigned_records.jsonl:18783"}
{"oral": "I'm sorry to hear that your package hasn't arrived yet. Could you please provide your order number so I can track it?", "clean": "I'm sorry to hear that your package hasn't arrived yet. Could you please provide your order number so I can track it?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18815.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 802, "source_record_id": "assigned_records.jsonl:18815"}
{"oral": "Could you please provide the latest project update and let me know if there are any blockers for the Q3 release?", "clean": "Could you please provide the latest project update and let me know if there are any blockers for the Q3 release?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18828.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 803, "source_record_id": "assigned_records.jsonl:18828"}
{"oral": "I need to remember to pick up the dry cleaning and buy some almond milk on my way home tonight.", "clean": "I need to remember to pick up the dry cleaning and buy some almond milk on my way home tonight.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18831.wav", "scene": "passthrough", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 806, "source_record_id": "assigned_records.jsonl:18831"}
{"oral": "Hey, are you free this Saturday? A few of us are planning to go hiking at 8:00am.", "clean": "Hey, are you free this Saturday? A few of us are planning to go hiking at 8:00am.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18836.wav", "scene": "passthrough", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 808, "source_record_id": "assigned_records.jsonl:18836"}
{"oral": "The current model achieves an F1 score of 0.89 on the validation set, but we are seeing significant overfitting on the training data.", "clean": "The current model achieves an F1 score of 0.89 on the validation set, but we are seeing significant overfitting on the training data.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18829.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 804, "source_record_id": "assigned_records.jsonl:18829"}
{"oral": "Please run the docker-compose up command and check if the redis container is running on port 6379.", "clean": "Please run the docker-compose up command and check if the redis container is running on port 6379.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18830.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 805, "source_record_id": "assigned_records.jsonl:18830"}
{"oral": "Could you please update the project timeline in Jira and notify the stakeholders about the delay in the API integration?", "clean": "Could you please update the project timeline in Jira and notify the stakeholders about the delay in the API integration?", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18840.wav", "scene": "passthrough", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 809, "source_record_id": "assigned_records.jsonl:18840"}
{"oral": "I think we should pivot our strategy for Q3 to focus more on user retention rather than aggressive acquisition.", "clean": "I think we should pivot our strategy for Q3 to focus more on user retention rather than aggressive acquisition.", "usage_text_words": null, "audio": "audio/passthrough__assigned_recordsjsonl18832.wav", "scene": "passthrough", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 807, "source_record_id": "assigned_records.jsonl:18832"}
{"oral": "Hey, so, quick thing. People always think that just dumping everything into Azure Blob is the fastest way to cache, but that is actually a huge misconception because of the latency. I want to run a real benchmark. I have got some configs in slash dot kube slash config and I am looking at the trivy image scan logs. Uh, let us compare the performance between using Amazon RDS... oh wait, no, I mean Amazon ElastiCache... and a local filesystem cache at slash var lib docker slash. Can you help me script a test for this? I need it to run for like ten... uh, let us say fifteen iterations.", "clean": "Hey, quick thing. People always think that just dumping everything into Azure Blob is the fastest way to cache, but that is actually a huge misconception because of the latency. I want to run a real benchmark. I have got some configs in /.kube/config and I am looking at the Trivy image scan logs. Let us compare the performance between using Amazon ElastiCache and a local filesystem cache at /var/lib/docker/. Can you help me script a test for this? I need it to run for 15 iterations.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl19011.wav", "scene": "tech", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 811, "source_record_id": "assigned_records.jsonl:19011"}
{"oral": "Hey! So I got this Express app and I need to add some authentication middleware. Currently I'm using Azure Cosmos DB... oh no, I mean Azure SQL Database for user storage. Anyway, I just ran docker-compose up to test locally and it's fine, but I'm seeing some weird cluster-dns issues in the dev environment. I tried a kubectl rollout undo but that didn't help, so I'm gonna try an argocd app sync. Can you help me write the middleware logic?", "clean": "Hey! So I got this Express app and I need to add some authentication middleware. Currently I'm using Azure SQL Database for user storage. Anyway, I just ran docker-compose up to test locally and it's fine, but I'm seeing some weird cluster-dns issues in the dev environment. I tried a kubectl rollout undo but that didn't help, so I'm gonna try an ArgoCD app sync. Can you help me write the middleware logic?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl19256.wav", "scene": "tech", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 813, "source_record_id": "assigned_records.jsonl:19256"}
{"oral": "Actually, people think I need to run terraform state pull to check the current schema, but that's not it. I'm actually seeing a 401 Unauthorized error when accessing the config in slash etc slash kubernetes slash or the one in tilde dot kube slash config. I already ran a trivy scan and everything is clean. Can you write the migration script for the database schema change?", "clean": "People think I need to run terraform state pull to check the current schema, but that's not it. I'm seeing a 401 Unauthorized error when accessing the config in /etc/kubernetes/ or the one in ~/.kube/config. I already ran a Trivy scan and everything is clean. Can you write the migration script for the database schema change?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl19580.wav", "scene": "tech", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 817, "source_record_id": "assigned_records.jsonl:19580"}
{"oral": "How should we configure the alerting for our queue depth? Actually I think we should trigger a warning when the concurrency limit hits eighty percent. We need to check if eksctl managed the node groups correctly. I will update the ansible playbook, update the ansible playbook to include the monitoring agent. Then we can monitor the RabbitMQ cluster, the RabbitMQ cluster specifically for unacknowledged messages. Also check the Amazon Aurora instance, wait, not the Aurora instance, the Aurora cluster for slow queries. Let us set the threshold to five hundred milliseconds.", "clean": "How should we configure the alerting for our queue depth? I think we should trigger a warning when the concurrency limit hits 80%. We need to check if eksctl managed the node groups correctly. I will update the Ansible playbook to include the monitoring agent. Then we can monitor the RabbitMQ cluster specifically for unacknowledged messages. Also check the Amazon Aurora cluster for slow queries. Let us set the threshold to 500ms.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl19568.wav", "scene": "tech", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 816, "source_record_id": "assigned_records.jsonl:19568"}
{"oral": "Hey, so my lead just told me, and I quote, the system is throwing a five zero four Gateway Timeout and we need to check the ingress. But, uh, looking at the logs, it seems... wait, let me rephrase... it looks more like a backend hang. I already ran aws e-k-s update kubeconfig to get into the cluster, and everything seems fine there. I suspect maybe RabbitMQ is bottlenecking or... um... maybe the connection to DynamoDB is lagging. Or, actually, it could be Amazon Aurora hitting some connection limit. Can you help me analyze if this timeout is coming from the load balancer or deeper in the stack?", "clean": "Hey, so my lead just told me, and I quote, the system is throwing a 504 Gateway Timeout and we need to check the ingress. But looking at the logs, it looks more like a backend hang. I already ran aws eks update-kubeconfig to get into the cluster, and everything seems fine there. I suspect maybe RabbitMQ is bottlenecking or maybe the connection to DynamoDB is lagging. Or, actually, it could be Amazon Aurora hitting some connection limit. Can you help me analyze if this timeout is coming from the load balancer or deeper in the stack?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl19317.wav", "scene": "tech", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 814, "source_record_id": "assigned_records.jsonl:19317"}
{"oral": "Hey, can you write me a migration script to add a new column for event tracking to our main table? Uh, so we're moving some logic from Apache Kafka into the DB to reduce latency. I tried running gcloud alpha... oh wait, not alpha, gcloud beta to check some configs, and I noticed we need this change before we push to /etc/kubernetes/manifests/ for the core components. We also gotta make sure the enable admission webhook... uh, enable admission webhook flag is set in ArgoCD, wait, no, it's in the cluster config... oh no, it's in ArgoCD, sorry. Anyway, just need that SQL script for the schema update, thanks.", "clean": "Hey, can you write me a migration script to add a new column for event tracking to our main table? So we're moving some logic from Apache Kafka into the DB to reduce latency. I tried running gcloud beta to check some configs, and I noticed we need this change before we push to /etc/kubernetes/manifests/ for the core components. We also gotta make sure the enable admission webhook flag is set in ArgoCD. Anyway, just need that SQL script for the schema update, thanks.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl19202.wav", "scene": "tech", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 812, "source_record_id": "assigned_records.jsonl:19202"}
{"oral": "Hey. So, first off, just to be clear, we aren't just wrapping the shell commands in a subprocess call. That is not what I mean. I want a proper rewrite into a Python CLI tool. Okay, so... uh... the script basically handles the RabbitMQ setup. It needs to trigger pulumi up for the infra, and then it scans images using trivy. Actually, wait... it should... let me rephrase... it needs to check the files in forward slash etc forward slash kubernetes forward slash manifests forward slash and then run trivy against those specific images. Can you handle that conversion?", "clean": "Hey. So, first off, just to be clear, we aren't just wrapping the shell commands in a subprocess call. That is not what I mean. I want a proper rewrite into a Python CLI tool. The script handles the RabbitMQ setup. It needs to trigger pulumi up for the infra, and then it needs to check the files in /etc/kubernetes/manifests/ and then run trivy against those specific images. Can you handle that conversion?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl19696.wav", "scene": "tech", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 818, "source_record_id": "assigned_records.jsonl:19696"}
{"oral": "Hey, quick question. I think some people think Exit Code one hundred thirty seven means a generic application crash, but oh no, it's specifically an OOM kill, right? So I'm trying to fix this in my pipeline. I need to scale the worker-processes, scale the worker-processes to maybe four or eight. I'm running the ansible-playbook now, but it keeps failing. Wait, it's failing with Exit Code one hundred thirty seven... ah no, it's actually one hundred twenty seven. Anyway, I'm also seeing some throttled requests on DynamoDB, so I think I need to bump the write-capacity-units to five hundred... five hundred. And lastly, can you help me check if Grafana Loki is actually indexing the logs? Because, how to say, the dashboard is just blank.", "clean": "Hey, quick question. I think some people think Exit Code 137 means a generic application crash, but it's specifically an OOM kill, right? So I'm trying to fix this in my pipeline. I need to scale the worker-processes to maybe 4 or 8. I'm running the ansible-playbook now, but it keeps failing. It's failing with Exit Code 127. Anyway, I'm also seeing some throttled requests on DynamoDB, so I think I need to bump the write-capacity-units to 500. And lastly, can you help me check if Grafana Loki is actually indexing the logs? Because the dashboard is just blank.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl19454.wav", "scene": "tech", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 815, "source_record_id": "assigned_records.jsonl:19454"}
{"oral": "Honestly, I think we got a four hundred nine Conflict on the state lock. I was trying to push the az network vnet peer changes, but uh, maybe we should... wait, let me rephrase... if we just tweak the concurrency limit in the provider block, it might stop hitting MongoDB Atlas so hard. Also, we need to check if Grafana Loki is actually receiving the logs or if it is just... uh... dropping them.", "clean": "Honestly, I think we got a 409 Conflict on the state lock. I was trying to push the az network vnet peer changes, but if we just tweak the concurrency limit in the provider block, it might stop hitting MongoDB Atlas so hard. Also, we need to check if Grafana Loki is actually receiving the logs or if it is just dropping them.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl19782.wav", "scene": "tech", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 819, "source_record_id": "assigned_records.jsonl:19782"}
{"oral": "If we switch from the standard az network vnet setup to crossplane beta, then we could benchmark if increasing keepalive_timeout helps, but I need you to check the auth context in my dot kube slash config first.", "clean": "If we switch from the standard Azure network VNet setup to Crossplane beta, then we could benchmark if increasing keepalive_timeout helps, but I need you to check the auth context in my .kube/config first.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl20180.wav", "scene": "tech", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 822, "source_record_id": "assigned_records.jsonl:20180"}
{"oral": "Honestly, I am staring at this production log and it makes zero sense. I tried running flux get kustomization to check the state, but it just keeps throwing a weird timeout. I think it might be buffer size issue, maybe set it to five hundred twelve? Just, how should I put this, I noticed the disk is almost full at forward slash var lib docker forward slash and we are connecting to Amazon RDS and az iot hub simultaneously. Can you help me decode this error?", "clean": "Honestly, I am staring at this production log and it makes zero sense. I tried running flux get kustomization to check the state, but it just keeps throwing a weird timeout. I think it might be a buffer size issue, maybe set it to 512? I noticed the disk is almost full at /var/lib/docker/ and we are connecting to Amazon RDS and Azure IoT Hub simultaneously. Can you help me decode this error?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl19809.wav", "scene": "tech", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 820, "source_record_id": "assigned_records.jsonl:19809"}
{"oral": "Why is my pod stuck in crash loop back off again? Hmm, I think it might be the cluster port configuration. I used eksctl to spin up the nodes, but I might have messed up the networking. It should be connecting to Amazon RDS, oh no, I mean Azure Cosmos DB for the state store, and then passing messages through RabbitMQ. Let me check if port eight zero eight zero... no, eight thousand eighty is open.", "clean": "Why is my pod stuck in CrashLoopBackOff again? I think it might be the cluster port configuration. I used eksctl to spin up the nodes, but I might have messed up the networking. It should be connecting to Azure Cosmos DB for the state store, and then passing messages through RabbitMQ. Let me check if port 8080 is open.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl20290.wav", "scene": "tech", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 824, "source_record_id": "assigned_records.jsonl:20290"}
{"oral": "So I just finished cleaning up the environment. I had to run terraform state rm to remove some stale resources. Then I did a trivy scan on the images and everything looks good. I already ran terraform apply for the Apache Kafka cluster and the MongoDB Atlas instance. Actually the MongoDB Atlas... it's like... it is fully provisioned now. So, uh, I think I think we are ready. Just, how should I put it, if you could just... could you write some integration tests for the user profile API endpoint, right?", "clean": "So I just finished cleaning up the environment. I had to run terraform state rm to remove some stale resources. Then I did a Trivy scan on the images and everything looks good. I already ran terraform apply for the Apache Kafka cluster and the MongoDB Atlas instance. The MongoDB Atlas is fully provisioned now. So, I think we are ready. If you could just write some integration tests for the user profile API endpoint, right?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl20234.wav", "scene": "tech", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 823, "source_record_id": "assigned_records.jsonl:20234"}
{"oral": "I am setting up the pipeline and I am not totally sure but I think I need to check the az network vnet configuration first. Then I will probably git rebase my feature branch before checking the config in /etc/nginx/ to make sure it is correct. After that I will trigger an argocd app sync and maybe tweak the cpu-shares to five hundred twelve to avoid throttling.", "clean": "I am setting up the pipeline and I am not totally sure but I think I need to check the Azure Network VNet configuration first. Then I will probably git rebase my feature branch before checking the config in /etc/nginx/ to make sure it is correct. After that I will trigger an ArgoCD app sync and maybe tweak the cpu-shares to 512 to avoid throttling.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl20389.wav", "scene": "tech", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 826, "source_record_id": "assigned_records.jsonl:20389"}
{"oral": "Look, most people think we just need to tweak the /proc/sys/net/ settings for this latency issue. But actually, it's not a kernel parameter problem. No, it's not a kernel parameter problem. I've been checking Datadog and the spikes are coming from the logging layer. We need to change Terraform to mount /var/log/pods/ as a separate volume. We should... I mean... if we adjust the storage class for the logs, it should stabilize. Also, make sure the trivy image scan is running before the deploy. I want the trivy image scan to fail the build if high severity hits. Once that's done, just trigger the argocd app sync. Trigger the argocd app sync for all twenty four clusters.", "clean": "Look, most people think we just need to tweak the /proc/sys/net/ settings for this latency issue. But actually, it's not a kernel parameter problem. I've been checking Datadog and the spikes are coming from the logging layer. We need to change Terraform to mount /var/log/pods/ as a separate volume. If we adjust the storage class for the logs, it should stabilize. Also, make sure the Trivy image scan is running before the deploy. I want the Trivy image scan to fail the build if high severity hits. Once that's done, just trigger the ArgoCD app sync for all 24 clusters.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl19965.wav", "scene": "tech", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 821, "source_record_id": "assigned_records.jsonl:19965"}
{"oral": "Hey Alex, I need help optimizing a slow SQL query. I think I think the bottleneck is in the worker-processes configuration. We are currently running on MongoDB Atlas, but the latency is spiking. I checked the port-mapping and ran istioctl analyze... it showed some issues with the health-check-interval. It was set to ten seconds... oh no, it was fifteen seconds. Can you look at the query plan and tell me if we should adjust the worker-processes worker-processes to handle more concurrent loads?", "clean": "Hey Alex, I need help optimizing a slow SQL query. I think the bottleneck is in the worker-processes configuration. We are currently running on MongoDB Atlas, but the latency is spiking. I checked the port-mapping and ran istioctl analyze... it showed some issues with the health-check-interval. It was set to 15 seconds. Can you look at the query plan and tell me if we should adjust the worker-processes to handle more concurrent loads?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl20515.wav", "scene": "tech", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 827, "source_record_id": "assigned_records.jsonl:20515"}
{"oral": "Hey, so I thought this was a gcloud pubsub issue at first, but actually it is not. The leak is happening at the node level. I checked the pod status and it was showing ErrImagePull... oh wait, no, it was ImagePullBackOff. Anyway, I am looking at the config in etc kubernetes manifests slash, and I think max worker processes is set way too high. I tried tuning some kernel parameters in proc sys net ipv4 slash but it did not help. I need to figure out why memory keeps climbing in this xing environment, so let me know if you see any pattern in the logs when it reaches two gigabytes of usage. I just want to make sure it does not crash again in the production garden.", "clean": "Hey, so I thought this was a gcloud Pub/Sub issue at first, but actually it is not. The leak is happening at the node level. I checked the pod status and it was showing ImagePullBackOff. Anyway, I am looking at the config in /etc/kubernetes/manifests/, and I think max worker processes is set way too high. I tried tuning some kernel parameters in /proc/sys/net/ipv4/ but it did not help. I need to figure out why memory keeps climbing in this xing environment, so let me know if you see any pattern in the logs when it reaches 2GB of usage. I just want to make sure it does not crash again in the production garden.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl20346.wav", "scene": "tech", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 825, "source_record_id": "assigned_records.jsonl:20346"}
{"oral": "So my lead told me we need to move the stateful set to cluster mode, cluster mode to handle the scaling better. I think I think this is overkill for our current load. I checked the nodes via k9s and noticed the disk pressure is mostly in var lib docker. I tried running aws eks update kubeconfig to sync everything but I still cannot hit the pod. Also we need to check gcloud iam permissions for the service account. Can you write me a migration script to update the schema? Make sure it handles the three hundred twenty two tables without locking the database.", "clean": "So my lead told me we need to move the stateful set to cluster mode to handle the scaling better. I think this is overkill for our current load. I checked the nodes via k9s and noticed the disk pressure is mostly in /var/lib/docker. I tried running aws eks update-kubeconfig to sync everything but I still cannot hit the pod. Also we need to check gcloud IAM permissions for the service account. Can you write me a migration script to update the schema? Make sure it handles the 322 tables without locking the database.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl20520.wav", "scene": "tech", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 828, "source_record_id": "assigned_records.jsonl:20520"}
{"oral": "Honestly, this function is a total mess. I need you to refactor it to handle the integration between Kong Gateway and HashiCorp Vault more cleanly, and then split the messaging logic between RabbitMQ and Apache Kafka. Also, make sure it correctly parses the path to dot slash kube slash config. Just keep it modular.", "clean": "Honestly, this function is a total mess. I need you to refactor it to handle the integration between Kong Gateway and HashiCorp Vault more cleanly, and then split the messaging logic between RabbitMQ and Apache Kafka. Also, make sure it correctly parses the path to ./kube/config. Just keep it modular.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl20786.wav", "scene": "tech", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 831, "source_record_id": "assigned_records.jsonl:20786"}
{"oral": "Honestly, I need you to scan the source and generate full API docs from the comments. I keep getting this... uh... Invalid Token error when I try to pull from slash etc slash kubernetes slash manifests slash. Also, could you... wait, actually, just look at the CI pipeline. Check if the helm upgrade dash install command is failing. I tried to run flux get kustomization but... let me rephrase... if you could just check the sync status first, that would be better. Oh, and I probably need to do a git rebase dash i for those last three commits. Just... skip the a header for now.", "clean": "Honestly, I need you to scan the source and generate full API docs from the comments. I keep getting this Invalid Token error when I try to pull from /etc/kubernetes/manifests/. Also, just look at the CI pipeline. Check if the helm upgrade --install command is failing. If you could just check the sync status first, that would be better. Oh, and I probably need to do a git rebase -i for those last 3 commits. Just skip the a header for now.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl20728.wav", "scene": "tech", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 829, "source_record_id": "assigned_records.jsonl:20728"}
{"oral": "Hey, I need to benchmark these two caching strategies. It's more like comparing throughput efficiency than just latency, you know? I'm seeing some weird spikes in Grafana Loki, and uh, I keep hitting a four hundred nine Conflict when I scale up worker-processes. Oh, wait, make that eight worker-processes. Also, the trivy scan is failing with Exit Code one hundred thirty seven, so it's probably OOM. Can you help me figure out which strategy handles this better?", "clean": "Hey, I need to benchmark these two caching strategies. It's more comparing throughput efficiency than just latency. I'm seeing some weird spikes in Grafana Loki, and I keep hitting a 409 Conflict when I scale up 8 worker-processes. Also, the Trivy scan is failing with Exit Code 137, so it's probably OOM. Can you help me figure out which strategy handles this better?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl20788.wav", "scene": "tech", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 832, "source_record_id": "assigned_records.jsonl:20788"}
{"oral": "Hey, can you help me add some auth middleware to my Express app? I need it to check tokens and... uh... maybe log the requests to var slash log slash pods slash... oh wait, actually just set the log level to debug for now. Also, make sure the idle timeout is set to sixty seconds... no, sorry, ninety seconds. I just did an az login and tried an Argo CD app sync but... wait, let me rephrase that. I just ran az login and then triggered an Argo CD app sync, so can you show me where to plug in the middleware in the main app file?", "clean": "Hey, can you help me add some auth middleware to my Express app? I need it to check tokens and just set the log level to debug for now. Also, make sure the idle timeout is set to 90 seconds. I just ran az login and then triggered an Argo CD app sync, so can you show me where to plug in the middleware in the main app file?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl20740.wav", "scene": "tech", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 830, "source_record_id": "assigned_records.jsonl:20740"}
{"oral": "So I've been looking at our setup on Cloudflare and some of the Google Cloud Spanner instances. I noticed we had some ImagePullBackOff errors on a few pods... oh wait, not ImagePullBackOff, it was CrashLoopBackOff. Anyway, I tried to trigger an ArgoCD app sync, an ArgoCD app sync to see if it clears up, but the memory usage just keeps climbing. I think, how should I put it, I think the memory leak leak is happening in the connection pool logic.", "clean": "So I've been looking at our setup on Cloudflare and some of the Google Cloud Spanner instances. I noticed we had some CrashLoopBackOff errors on a few pods. Anyway, I tried to trigger an ArgoCD app sync to see if it clears up, but the memory usage just keeps climbing. I think the memory leak is happening in the connection pool logic.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl20842.wav", "scene": "tech", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 833, "source_record_id": "assigned_records.jsonl:20842"}
{"oral": "Hey, can you write some integration tests for this endpoint? I need it to be more like a full end to end flow than just simple mocking. Just, uh, make sure it covers the Amazon Aurora connection and gcloud pubsub triggers. Also, we need to simulate a five hundred Internal Server Error when idle timeout hits. Oh and let's check if Datadog picks up the trace, right?", "clean": "Hey, can you write some integration tests for this endpoint? I need it to be more like a full end-to-end flow than just simple mocking. Just make sure it covers the Amazon Aurora connection and gcloud Pub/Sub triggers. Also, we need to simulate a 500 Internal Server Error when idle timeout hits. Oh and let's check if Datadog picks up the trace, right?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl20867.wav", "scene": "tech", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 834, "source_record_id": "assigned_records.jsonl:20867"}
{"oral": "Hey, I got this slow query on Amazon Aurora that is just killing performance. I'm not totally sure but I think it might be related to some cluster-visibility issues, uh, maybe combined with how we configured the Cloudflare Tunnel. Wait, I checked the logs and it is not the tunnel, oh no, it is actually the database indexing. Anyway, I already ran istioctl analyze and scanned everything with trivy, but the latency is still over five hundred milliseconds. Can you help me optimize the join logic?", "clean": "Hey, I got this slow query on Amazon Aurora that is just killing performance. I'm not totally sure but I think it might be related to some cluster-visibility issues, maybe combined with how we configured the Cloudflare Tunnel. I checked the logs and it is actually the database indexing. Anyway, I already ran istioctl analyze and scanned everything with Trivy, but the latency is still over 500 milliseconds. Can you help me optimize the join logic?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl21286.wav", "scene": "tech", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 837, "source_record_id": "assigned_records.jsonl:21286"}
{"oral": "Hey can you help me port this bash script to a Python CLI tool. I need it to parse /var/log/syslog and then... uh... maybe instead of just grep... let's say it should filter for five hundred Internal Server Error. Also need to add a configurable buffer size... say like four thousand ninety six bytes. Plus it should... wait... actually... if we could integrate it with k9s or... how should I put it... basically just trigger an ArgoCD app sync when it finds an error. Hope this makes sense and let me know if you need more context to get it right.", "clean": "Hey, can you help me port this bash script to a Python CLI tool? I need it to parse /var/log/syslog and then it should filter for 500 Internal Server Error. Also need to add a configurable buffer size, say 4096 bytes. Plus it should trigger an ArgoCD app sync when it finds an error. Hope this makes sense and let me know if you need more context to get it right.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl21097.wav", "scene": "tech", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 835, "source_record_id": "assigned_records.jsonl:21097"}
{"oral": "Actually, use curl dash d... oh no, curl dash k for docs.", "clean": "Use curl -k for docs.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl21456.wav", "scene": "tech", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 843, "source_record_id": "assigned_records.jsonl:21456"}
{"oral": "I checked helm list and saw the pod crashing. I tried wget dash dash spider... oh wait, wget dash dash header to check the endpoint, but it failed. I might need to run terraform destroy... no, terraform apply to reset everything. I'll update the config using aws ssm put parameter. Does that sound right to you, or should I check the logs first in the new cluster?", "clean": "I checked helm list and saw the pod crashing. I tried wget --header to check the endpoint, but it failed. I might need to run terraform apply to reset everything. I'll update the config using aws ssm put-parameter. Does that sound right to you, or should I check the logs first in the new cluster?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl21374.wav", "scene": "tech", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 840, "source_record_id": "assigned_records.jsonl:21374"}
{"oral": "Hey, I want you to scan my source code and generate API documentation from comments. I am not totally sure but I think... I think we need to adjust heap-size and buffer-size, maybe set heap-size to two gigabytes. We are using Amazon Aurora for database, and... wait... if we just use... how should I put this... if we use pulumi up to deploy, it might trigger a restart. Also check the az network vnet configuration. Make sure the config is correct so it can reach the server. I think we need to check check the network routes again.", "clean": "Hey, I want you to scan my source code and generate API documentation from comments. I am not totally sure but I think we need to adjust heap-size and buffer-size, maybe set heap-size to 2GB. We are using Amazon Aurora for database, and if we use pulumi up to deploy, it might trigger a restart. Also check the Azure network VNet configuration. Make sure the config is correct so it can reach the server. I think we need to check the network routes again.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl21336.wav", "scene": "tech", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 838, "source_record_id": "assigned_records.jsonl:21336"}
{"oral": "Hey Claude, uh, I got this messy script for gcloud container clusters and it needs a serious cleanup. It is... wait, actually... let me rephrase, it is basically a disaster. Like, it tries to check the liveness-probe by digging into /etc/kubernetes/ and then it just fails with ImagePullBackOff... just totally randomly, you know? I want to stop doing docker exec -it on every single node, so, uh, can you refactor this into some clean Python logic, maybe? I think it would be better, right?", "clean": "Hey Claude, I got this messy script for gcloud container clusters and it needs a serious cleanup. It is a disaster. It tries to check the liveness-probe by digging into /etc/kubernetes/ and then it just fails with ImagePullBackOff totally randomly. I want to stop doing docker exec -it on every single node, so can you refactor this into some clean Python logic, maybe? I think it would be better, right?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl21271.wav", "scene": "tech", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 836, "source_record_id": "assigned_records.jsonl:21271"}
{"oral": "Hey, can you refactor this function? It's more like a messy terraform plan than clean code. Um, it's like... just a giant list of az group create... oh no, az webapp deploy calls. I mean... it's more like a shell script with ssh dash i and docker ps, just... way too cluttered.", "clean": "Hey, can you refactor this function? It's more like a messy Terraform plan than clean code. It's just a giant list of az webapp deploy calls. It's more like a shell script with ssh -i and docker ps, just way too cluttered.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl21429.wav", "scene": "tech", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 842, "source_record_id": "assigned_records.jsonl:21429"}
{"oral": "Hey, can you help me write some integration tests for the new endpoint? I'm not totally sure but I think we need to mock the Apache Kafka producer, mock the Apache Kafka producer to avoid hitting the actual broker. Also, we should check if the message reaches gcloud pubsub. Wait, oh no, it's not pubsub, it's Amazon RDS for the final state check. And I want to make sure the pipeline includes a trivy image scan, includes a trivy image scan before we deploy through Cloudflare Tunnel. Can you use port eight thousand eight hundred eighty eight for the local setup?", "clean": "Hey, can you help me write some integration tests for the new endpoint? I'm not totally sure but I think we need to mock the Apache Kafka producer to avoid hitting the actual broker. Also, we should check if the message reaches Amazon RDS for the final state check. And I want to make sure the pipeline includes a Trivy image scan before we deploy through Cloudflare Tunnel. Can you use port 8888 for the local setup?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl21337.wav", "scene": "tech", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 839, "source_record_id": "assigned_records.jsonl:21337"}
{"oral": "Uh, I need to sync the monitoring rules. It is more like... uh... debugging the infrastructure flow than just writing a config. So, let us start by running helm list to see what is deployed, then maybe a curl dash v to check the health endpoint. Oh, I should probably use ssh dash P... no, wait, not P, it is ssh dash N for the tunnel. After that, I will run terraform output to grab the load balancer I P and then, uh, terraform apply to push the changes. I think we need to set the threshold to eighty percent... oh no, not eighty, it is ninety five percent for the critical alert. Does that sound right?", "clean": "I need to sync the monitoring rules. It is more like debugging the infrastructure flow than just writing a config. So, let us start by running helm list to see what is deployed, then maybe a curl -v to check the health endpoint. I should probably use ssh -N for the tunnel. After that, I will run terraform output to grab the load balancer IP and then terraform apply to push the changes. I think we need to set the threshold to 95% for the critical alert. Does that sound right?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl21388.wav", "scene": "tech", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 841, "source_record_id": "assigned_records.jsonl:21388"}
{"oral": "Hmm it's more like a leak during helm upgrade dash dash install than a fresh helm install. The id id from jq dot items bracket zero bracket dot id is climbing, climbing way too fast.", "clean": "It's more like a leak during helm upgrade --install than a fresh helm install. The id from jq .items[0].id is climbing way too fast.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl21858.wav", "scene": "tech", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 846, "source_record_id": "assigned_records.jsonl:21858"}
{"oral": "Actually it is not just adding a column but splitting the table, so please write a migration script where I can check logs using ssh minus v or verify checksums with openssl dgst dash sha two hundred fifty six after I run docker volume ls to confirm the data paths.", "clean": "Actually it is not just adding a column but splitting the table, so please write a migration script where I can check logs using ssh -v or verify checksums with openssl dgst -sha256 after I run docker volume ls to confirm the data paths.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl21716.wav", "scene": "tech", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 845, "source_record_id": "assigned_records.jsonl:21716"}
{"oral": "Wait, do I need to manually handle the state for this database schema change? Actually, I think it's better if we just automate it. So, uh, I already did the terraform init and ran aws iam create user to set up permissions. Now I need a migration script because I am about to do a git merge into main. After that, I will probably run helm upgrade dash dash values for the config, and then maybe helm upgrade dash dash atomic to make sure it rolls back if it fails. Can you write that script for me, please?", "clean": "Do I need to manually handle the state for this database schema change? I think it's better if we just automate it. I already did the terraform init and ran aws iam create user to set up permissions. Now I need a migration script because I am about to do a git merge into main. After that, I will probably run helm upgrade --values for the config, and then maybe helm upgrade --atomic to make sure it rolls back if it fails. Can you write that script for me, please?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl21572.wav", "scene": "tech", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 844, "source_record_id": "assigned_records.jsonl:21572"}
{"oral": "Wait, I thought terraform output would show the error, but it actually just shows state. Actually, actually, I need to check docker volume ls first, then maybe kubectl get pods, huh. Let me try git rebase dash i to clean up, or maybe curl dash u with user credentials, user credentials, maybe that works, right?", "clean": "I thought terraform output would show the error, but it actually just shows state. I need to check docker volume ls first, then maybe kubectl get pods. Let me try git rebase -i to clean up, or maybe curl -u with user credentials, maybe that works, right?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl22020.wav", "scene": "tech", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 849, "source_record_id": "assigned_records.jsonl:22020"}
{"oral": "My lead said just run kubectl delete namespace... uh no, kubectl get namespace first, then... I mean... just use az vm create and ssh-keygen dash t ed two five five one nine, plus openssl req dash new and openssl pkey, you know? Now write tests for that endpoint, okay?", "clean": "My lead said just run kubectl get namespace first, then just use az vm create and ssh-keygen -t ed25519, plus openssl req -new and openssl pkey. Now write tests for that endpoint, okay?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl22054.wav", "scene": "tech", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 850, "source_record_id": "assigned_records.jsonl:22054"}
{"oral": "I want to start by az group create, az group create a new resource group. Then use openssl req dash new for certificates, oh no, use az keyvault secret set instead. Then I'll check kubectl get events, kubectl get events to see if it's running. Does that sound like a good plan?", "clean": "I want to start by az group create a new resource group. Then use az keyvault secret set instead. Then I'll check kubectl get events to see if it's running. Does that sound like a good plan?", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl22161.wav", "scene": "tech", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 851, "source_record_id": "assigned_records.jsonl:22161"}
{"oral": "Honestly, I think we have a memory leak... a memory leak in the pod. I tried docker build, then... how should I put it... I ran helm upgrade dash dash atomic. Wait, I meant helm upgrade dash dash install, no, it was atomic. I used aws ssm put parameter to update config, but the pods keep crashing. I tried ssh dash L to tunnel in... tunnel in and ran aws rds describe dash db dash instances to check connection. I think the memory usage is... is getting higher and higher in the new system.", "clean": "Honestly, I think we have a memory leak in the pod. I tried docker build, then I ran helm upgrade --atomic. I used aws ssm put-parameter to update config, but the pods keep crashing. I tried ssh -L to tunnel in and ran aws rds describe-db-instances to check connection. I think the memory usage is getting higher and higher in the new system.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl21899.wav", "scene": "tech", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 847, "source_record_id": "assigned_records.jsonl:21899"}
{"oral": "If terraform validate passes, then we could check helm rollout history or kubectl get events, maybe run helm template and openssl x509 dash in to verify certs.", "clean": "If terraform validate passes, then we could check helm rollout history or kubectl get events, maybe run helm template and openssl x509 -in to verify certs.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl22269.wav", "scene": "tech", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 854, "source_record_id": "assigned_records.jsonl:22269"}
{"oral": "Hey, can you help me add authentication middleware to my Express app? I need it to parse the token from a response using jq minus r... wait, no, using jq map dot name. I was trying to run terraform validate and then git stash pop to get my changes back... I mean... if we could just... maybe integrate it so I can check the bucket with aws s3 ls. I need to run the run the script and... and the script should check the user roles... oh not roles, the user permissions for the five hundred twelve endpoints.", "clean": "Hey, can you help me add authentication middleware to my Express app? I need it to parse the token from a response using jq map .name. I was trying to run terraform validate and then git stash pop to get my changes back, if we could just integrate it so I can check the bucket with aws s3 ls. I need to run the script and the script should check the user permissions for the 512 endpoints.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl21981.wav", "scene": "tech", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 848, "source_record_id": "assigned_records.jsonl:21981"}
{"oral": "Hey, if this bash script still uses docker push... oh wait, not docker pull, then we could rewrite it as Python CLI tool. Like, handle terraform plan, openssl x five hundred nine, and wget dash c, plus that openssl s underscore client dash connect part too.", "clean": "Hey, if this bash script still uses docker push, then we could rewrite it as a Python CLI tool. Handle terraform plan, openssl x509, and wget -c, plus that openssl s_client -connect part too.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl22248.wav", "scene": "tech", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 853, "source_record_id": "assigned_records.jsonl:22248"}
{"oral": "First I ran terraform workspace select, terraform workspace select to switch env. Then I tried terraform import, wait, I mean I used terraform import to bring in state. Finally I did helm upgrade dash dash install and curl dash k, curl dash k to check, then ssh dash R for tunnel, but memory keep climbing.", "clean": "1. I ran terraform workspace select to switch env.\n2. I used terraform import to bring in state.\n3. I did helm upgrade --install, curl -k to check, then ssh -R for tunnel, but memory keep climbing.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl22275.wav", "scene": "tech", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 855, "source_record_id": "assigned_records.jsonl:22275"}
{"oral": "Hey, this SQL query performance is just... uh, it's more like a slow crawl than a quick fetch. I already checked gcloud sql instances list and terraform output, but nothing. Can you optimize it? It feels more like trying to run kubectl describe node on a thousand pods than a simple select.", "clean": "Hey, this SQL query performance is just more like a slow crawl than a quick fetch. I already checked gcloud sql instances list and terraform output, but nothing. Can you optimize it? It feels more like trying to run kubectl describe node on a thousand pods than a simple select.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl22298.wav", "scene": "tech", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 856, "source_record_id": "assigned_records.jsonl:22298"}
{"oral": "Need migration script for user table. Need migration script because I gotta git commit dash m then git push origin. Also check gcloud compute instances list and curl dash L the endpoint, then aws secrets manager get secret value for keys.", "clean": "Need migration script for user table. Need migration script because I gotta git commit -m then git push origin. Also check gcloud compute instances list and curl -L the endpoint, then aws secrets manager get secret value for keys.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl22310.wav", "scene": "tech", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 857, "source_record_id": "assigned_records.jsonl:22310"}
{"oral": "Can you help me clean up this function? Actually, actually I think it needs a full refactor. This block is a mess, it mixes git cherry pick and git rebase dash i logic with cloud commands like az appservice plan create. It even has these raw openssl x five hundred nine and openssl dgst dash s h a two hundred fifty six calls. The logic... the logic is all tangled. I want you to separate the git operations and these openssl commands into different modules. This function is too long, too long for one file.", "clean": "Can you help me clean up this function? I think it needs a full refactor. This block is a mess; it mixes git cherry-pick and git rebase -i logic with cloud commands like az appservice plan create. It even has these raw openssl x509 and openssl dgst -sha256 calls. The logic is all tangled. I want you to separate the git operations and these openssl commands into different modules. This function is too long for one file.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl22167.wav", "scene": "tech", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 852, "source_record_id": "assigned_records.jsonl:22167"}
{"oral": "Maya, what score score is it for the Lakers game? Wait, not Lakers, oh no, the Celtics game. Tell me tell me the score and wake me up at seven tomorrow morning. Also, turn off all lights in living room and open Pin-t-rest for Sophia.", "clean": "Maya, what score is it for the Celtics game? Tell me the score and wake me up at 7 tomorrow morning. Also, turn off all lights in the living room and open Pinterest for Sophia.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24381.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 861, "source_record_id": "assigned_records.jsonl:24381"}
{"oral": "Hey, check flight status for flight one hundred twenty. I thought it was delayed three and a half hours, oh no, it's just a short delay. Check check the gate. I was looking at... I mean, I was browsing Pinterest and Duolingo and saw some ad for eighty-eight cents, but that's not it. The flight... the flight is departing from terminal two, no, terminal three. Wait, it's departing from terminal three. Is it on time or on time for the gate change? I need to know if it's still at the same gate.", "clean": "Hey, check flight status for flight 120. I thought it was delayed 3.5 hours, no, it's just a short delay. Check the gate. I was browsing Pinterest and Duolingo and saw some ad for $0.88, but that's not it. The flight is departing from terminal 3. Is it on time for the gate change? I need to know if it's still at the same gate.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24355.wav", "scene": "voice_search", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 860, "source_record_id": "assigned_records.jsonl:24355"}
{"oral": "Hey, find that podcast with... uh... the one about Dominic and Isabella. Also, set alarm for seven fifteen tomorrow, and then... wait, no... actually, what's definition of serendipity? Plus, call my brother on speaker.", "clean": "Hey, find that podcast with the one about Dominic and Isabella. Also, set alarm for 7:15 tomorrow, and then what's the definition of serendipity? Plus, call my brother on speaker.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24543.wav", "scene": "voice_search", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 863, "source_record_id": "assigned_records.jsonl:24543"}
{"oral": "Can you help me port this bash script into a Python CLI tool? Actually I want I want to convert it. It uses openssl x five zero nine and wget dash O to fetch files, then it uses jq quote map dot name quote for parsing, and it needs to run terraform fmt and ssh dash o for config. Wait, I mean it should handle ssh dash o options, it should handle options in Python. I think we can use a library for that, but it must be fast, fast enough for production, so we need to make sure it's efficient in the new version.", "clean": "Can you help me convert this bash script into a Python CLI tool? It uses openssl x509 and wget -O to fetch files, then it uses jq \"map(.name)\" for parsing, and it needs to run terraform fmt and ssh -o for config. It should handle options in Python. I think we can use a library for that, but it must be fast enough for production, so we need to make sure it's efficient in the new version.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl22373.wav", "scene": "tech", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 858, "source_record_id": "assigned_records.jsonl:22373"}
{"oral": "Wait, people think we need kubectl delete namespace, no, not that, we just need to update the schema. I mean, we need to update update the schema. I think... if we... how to put it... if we use aws eks update-kubeconfig first... wait, no, gcloud compute instances list first. Just give me migration script for the users table, users table, for that new column. I need it to be... to be compatible with the current version. Oh, I mean... the version two point zero. I'm working in this new new environment.", "clean": "People think we need kubectl delete namespace, but we just need to update the schema. If we use gcloud compute instances list first, just give me the migration script for the users table for that new column. I need it to be compatible with version 2.0. I'm working in this new environment.", "usage_text_words": null, "audio": "audio/tech__assigned_recordsjsonl22380.wav", "scene": "tech", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 859, "source_record_id": "assigned_records.jsonl:22380"}
{"oral": "Hey kitchen speaker, if eighty eight percent of a cup is too much... then we could use... use half a cup. Wait, actually, call my brother on speaker. Call my brother on speaker to ask him. Also, set an alarm for seven forty five and... open You tube on the kitchen speakar.", "clean": "Hey kitchen speaker, if 88% of a cup is too much, then we could use half a cup. Call my brother on speaker to ask him. Also, set an alarm for 7:45 and open YouTube on the kitchen speaker.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24442.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 862, "source_record_id": "assigned_records.jsonl:24442"}
{"oral": "Actually, uh, set timer for ten minutes. Oh, and just... search for best pizza place nearby, maybe order pepperoni pizza from Dominoes, ya. I think I'll check TikTok, just... how should I put it... just for reviews, ba.", "clean": "Set timer for 10 minutes. Search for best pizza place nearby, maybe order pepperoni pizza from Domino's. I think I'll check TikTok for reviews.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24565.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 865, "source_record_id": "assigned_records.jsonl:24565"}
{"oral": "Hey, check flight status for... check flight status for my trip. Also, find nearest Whole Foods... oh no, nearest airport parking. Wait, actually, schedule a meeting with Sarah for tomorrow at two P M. And check if Duolingo... Duolingo app updated. Oh, and... open Pinterest... no, open Pinterest at eleven forty five P M.", "clean": "Hey, check flight status for my trip. Also, find nearest airport parking. Schedule a meeting with Sarah for tomorrow at 2pm. And check if the Duolingo app updated. Oh, and open Pinterest at 11:45pm.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24549.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 864, "source_record_id": "assigned_records.jsonl:24549"}
{"oral": "Hey, wait, I thought one cup was eight ounces but eh, that's only for liquid. How many grams in one cup of flour? Also, read my unread emails from today on Discord, Coursera, and LinkedIn.", "clean": "Hey, wait, I thought 1 cup was 8 ounces but that's only for liquid. How many grams in 1 cup of flour? Also, read my unread emails from today on Discord, Coursera, and LinkedIn.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24631.wav", "scene": "voice_search", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 867, "source_record_id": "assigned_records.jsonl:24631"}
{"oral": "Play my workout playlist for 15 minutes, also set smart thermostat, and then set an alarm for 7:15 tomorrow, plus find closest gas station and read my last 3 unread emails", "clean": "Play my workout playlist for 15 minutes, also set smart thermostat, and then set an alarm for 7:15 tomorrow, plus find closest gas station and read my last 3 unread emails", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24661.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 869, "source_record_id": "assigned_records.jsonl:24661"}
{"oral": "Hey, I'm not totally sure but I think... I think it's a song Isabella and Penelope found on Pinterest. It's... wait, it's the one Julian likes. Let me... let me hum it. Hmm, it sounds like... well, if I can remember... it's like this. Does it sound like something from city library playlist?", "clean": "Hey, I'm not totally sure but I think it's a song Isabella and Penelope found on Pinterest. It's the one Julian likes. Let me hum it. It sounds like this. Does it sound like something from the city library playlist?", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24598.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 866, "source_record_id": "assigned_records.jsonl:24598"}
{"oral": "Hey, my friend said it's gonna rain all weekend, but I think it'll be fine. Um, what's weather forecast for Saturday and Sunday? Also, find closest twenty four hour gym. Wait, I need to check TikTok... oh no, Airbnb for my trip. What's on my calendar for next Tuesday? Like, ninety seconds from now, tell me if I have any meetings.", "clean": "Hey, my friend said it's gonna rain all weekend, but I think it'll be fine. What's the weather forecast for Saturday and Sunday? Also, find the closest 24-hour gym. I need to check Airbnb for my trip. What's on my calendar for next Tuesday? 90 seconds from now, tell me if I have any meetings.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24633.wav", "scene": "voice_search", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 868, "source_record_id": "assigned_records.jsonl:24633"}
{"oral": "Hey Siri, tell me scores for the game. Give me give me the score. Wait, for the Snapchat league... no, DoorDash league. Did they hit eleven thousand points? Eleven thousand points? Also find closest gym, then open Duolingo.", "clean": "Hey Siri, tell me scores for the game. Give me the score for the DoorDash league. Did they hit 11,000 points? Also find closest gym, then open Duolingo.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24705.wav", "scene": "voice_search", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 870, "source_record_id": "assigned_records.jsonl:24705"}
{"oral": "Honestly, uh, search for best pizza place nearby. Oh wait, also what's on my calendar for next Tuesday? And, just... book ride to airport for tomorrow at five. Oh, and send it to Xavier and Sophia.", "clean": "Honestly, search for the best pizza place nearby. Also, what's on my calendar for next Tuesday? And just book a ride to the airport for tomorrow at 5. Oh, and send it to Xavier and Sophia.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24756.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 872, "source_record_id": "assigned_records.jsonl:24756"}
{"oral": "Wow! Quick, tell me score score of the game. Also... check if I got... check if I got any new messages on Discord, and read my last unread email. Then open Coursera, open Coursera and Twitch.", "clean": "Wow! Quick, tell me the score of the game. Also, check if I got any new messages on Discord, and read my last unread email. Then open Coursera and Twitch.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24840.wav", "scene": "voice_search", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 874, "source_record_id": "assigned_records.jsonl:24840"}
{"oral": "Hey, uh, can you find this song? Hmm, it goes like... wait, let me... it sounds like Maya singing, plus also check if Oliver or Olivia shared it on Discord, and then just... just search Coursera for that music course, yeah.", "clean": "Hey, can you find this song? It sounds like Maya singing, plus also check if Oliver or Olivia shared it on Discord, and then just search Coursera for that music course, yeah.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24745.wav", "scene": "voice_search", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 871, "source_record_id": "assigned_records.jsonl:24745"}
{"oral": "Set timer for ten minutes. If timer goes off... then we could check Waze, check Waze for traffic. Also add organic eggs to my grocery list. Remind me to finish Coursera course... no, finish Udemy course tomorrow. Find nearest DMV and close window blinds, close window blinds.", "clean": "Set timer for 10 minutes. If timer goes off, then we could check Waze for traffic. Also add organic eggs to my grocery list. Remind me to finish Udemy course tomorrow. Find nearest DMV and close window blinds.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24816.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 873, "source_record_id": "assigned_records.jsonl:24816"}
{"oral": "Julian said just... uh... turn off all the lights in living room. I think he's crazy. Add... uh... LinkedIn... no, Pinterest... wait wrong, it's Duolingo to shopping list. Oh wait, I meant milk!", "clean": "Julian said just turn off all the lights in the living room. I think he's crazy. Add milk to shopping list.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24911.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 876, "source_record_id": "assigned_records.jsonl:24911"}
{"oral": "Sebastian, wait, what time... uh, just start a workout for thirty minutes. And check traffic on the way to airport. Oh, and check my Zoom, wait, no, check Tinder. Is it... is it raining? No, it's fine. Actually, what's traffic like on the way to airport? Oh, wait, I already asked that. Just do it.", "clean": "Sebastian, what time... just start a workout for 30 minutes. And check traffic on the way to the airport. Oh, and check Tinder. Is it raining? No, it's fine. What's traffic like on the way to the airport? Oh, wait, I already asked that. Just do it.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl24887.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 875, "source_record_id": "assigned_records.jsonl:24887"}
{"oral": "Hey, what's weather like this weekend? Oh, wait, I can just check. Actually, uh, check weather for Saturday and Sunday. And also, find nearest twenty four hour gym... no, wait, local pharmacy. And tell Isabella how long does it take to get to airport. Oh, and check if Tinder notification is on.", "clean": "Hey, what's the weather like this weekend? Check weather for Saturday and Sunday. And also, find nearest 24-hour local pharmacy. And tell Isabella how long does it take to get to the airport. Oh, and check if Tinder notification is on.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25031.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 877, "source_record_id": "assigned_records.jsonl:25031"}
{"oral": "Hey, uh, find me that YouTube podcast about vacuum robot... vacuum robot technology, and, well, I think think it was episode twenty two. Actually, it was... maybe episode twenty five? Do you think it's worth a listen, huh?", "clean": "Hey, find me that YouTube podcast about vacuum robot technology, and I think it was episode 25. Do you think it's worth a listen?", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25035.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 878, "source_record_id": "assigned_records.jsonl:25035"}
{"oral": "Oh! Set alarm for six thirty. Six thirty. Then set one for seven. Then set one for seven. Remind Julian to check Spotify. Also remind Benjamin... remind Benjamin to navigate to nearest airport. And tell me how many calories... how many calories in avocado.", "clean": "Set alarm for 6:30. Then set one for 7. Remind Julian to check Spotify. Also remind Benjamin to navigate to nearest airport. And tell me how many calories in avocado.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25121.wav", "scene": "voice_search", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 879, "source_record_id": "assigned_records.jsonl:25121"}
{"oral": "Hey! Show me photos from my trip. I think I took... I took two hundred and ten... no, two hundred and twenty... wait wrong, is two hundred and twelve! The ones from... from Uber ride. Also check front door lock. Oh, what's latest score of Lakers game? And set alarm for eleven forty-five PM.", "clean": "Hey! Show me photos from my trip. I think I took 212! The ones from the Uber ride. Also check front door lock. Oh, what's the latest score of the Lakers game? And set alarm for 11:45pm.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25174.wav", "scene": "voice_search", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 880, "source_record_id": "assigned_records.jsonl:25174"}
{"oral": "Honestly... uh... set timer for... wait... set timer for three and a half hours. And... um... play some relaxing jazz in living room. Oh, also... remind me to check Discord... actually, remind me to check Coursera and that two hundred dollars thing.", "clean": "Honestly, set timer for 3.5 hours. And play some relaxing jazz in the living room. Oh, also, remind me to check Coursera and that $200 thing.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25253.wav", "scene": "voice_search", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 882, "source_record_id": "assigned_records.jsonl:25253"}
{"oral": "Set alarm for six... uh six thirty. And if I wake up early then we could find a recipe for gluten-free lasagna. Also set one for seven... oh no, seven fifteen. What's definition of serendipity? Remind me to do Duolingo at eight. Last one, find closest gym... no, closest movie theater for tomorrow.", "clean": "Set alarm for 6:30. And if I wake up early then we could find a recipe for gluten-free lasagna. Also set one for 7:15. What's the definition of serendipity? Remind me to do Duolingo at 8. Last one, find closest movie theater for tomorrow.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25178.wav", "scene": "voice_search", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 881, "source_record_id": "assigned_records.jsonl:25178"}
{"oral": "Find podcast episode about Sophia... Sophia... about Pinterest... no, not Pinterest, about half a million... half a million calories... oh no, how many calories are in a medium banana, and turn off all the lights in the living room.", "clean": "Find podcast episode about Sophia about how many calories are in a medium banana, and turn off all the lights in the living room.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25324.wav", "scene": "voice_search", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 884, "source_record_id": "assigned_records.jsonl:25324"}
{"oral": "Actually... check flight status check flight status for me. Is it still on time? Also find closest gym near city hall. I need to check Coursera Coursera course later. Is it twenty-two degrees outside? And delete Tinder.", "clean": "Check flight status for me. Is it still on time? Also find the closest gym near City Hall. I need to check Coursera course later. Is it 22 degrees outside? And delete Tinder.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25290.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 883, "source_record_id": "assigned_records.jsonl:25290"}
{"oral": "Hey, find me Xavier's pasta recipe on YouTube, and then check if Beatrice sent one, plus... uh, see if hallway dimmer light is on. Oh, and how much one hundred twenty dollars in euros?", "clean": "Hey, find me Xavier's pasta recipe on YouTube, and then check if Beatrice sent one, plus see if the hallway dimmer light is on. Oh, and how much is $120 in euros?", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25406.wav", "scene": "voice_search", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 888, "source_record_id": "assigned_records.jsonl:25406"}
{"oral": "Actually, can you search Spotify, search Spotify for that episode from eleven forty five PM? Oh and set an alarm for seven fifteen tomorrow morning. Also search for best pizza place nearby on DoorDash... DoorDash, but make sure it's open now. Wait, check if it's open now.", "clean": "Can you search Spotify for that episode from 11:45pm? Oh and set an alarm for 7:15 tomorrow morning. Also search for best pizza place nearby on DoorDash, but check if it's open now.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25336.wav", "scene": "voice_search", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 886, "source_record_id": "assigned_records.jsonl:25336"}
{"oral": "Hey Serenity, uh, turn off lights. All lights. Wait, search for nearest charging station... no, closest movie theater... wait, wrong, is nearest charging station! I need to check LinkedIn... check LinkedIn, and Pinterest, just for a bit, ya.", "clean": "Hey Serenity, turn off the lights. All lights. Search for the nearest charging station! I need to check LinkedIn and Pinterest, just for a bit, ya.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25330.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 885, "source_record_id": "assigned_records.jsonl:25330"}
{"oral": "Find that podcast where he said call my boss on speaker and then mentioned Tinder. He said eighty-five percent eighty-five percent of people use it wrong. I think he think... he meant... basically he's wrong about Pinterest. Book a ride to airport for eight AM.", "clean": "Find that podcast where he said call my boss on speaker and then mentioned Tinder. He said 85% of people use it wrong. I think he meant he's wrong about Pinterest. Book a ride to the airport for 8am.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25453.wav", "scene": "voice_search", "language": "en", "voice_id": "en_male_michael_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 889, "source_record_id": "assigned_records.jsonl:25453"}
{"oral": "Send text to Julian. Tell him it's negative five... oh no, negative ten degrees outside, so turn on smart plug. Also, schedule a meeting with Sarah for tomorrow at two p.m. and then remind him about Airbnb... plus, check if he got the mail in the new garden.", "clean": "Send text to Julian. Tell him it's -10 degrees outside, so turn on smart plug. Also, schedule a meeting with Sarah for tomorrow at 2pm and then remind him about Airbnb... plus, check if he got the mail in the new garden.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25366.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 887, "source_record_id": "assigned_records.jsonl:25366"}
{"oral": "Hey, check flight status for my sister, uh, is flight 204 on time? Actually, just tell me if flight 204 is delayed.", "clean": "Hey, check flight status for my sister, tell me if flight 204 is delayed.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25635.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 894, "source_record_id": "assigned_records.jsonl:25635"}
{"oral": "Eh, set timer for three hours, three hours, just... I mean, turn off TV... no, turn off the TV first and turn on sprinkler system, turn on sprinkler system since it's fifty percent dry, ya. Wait, Frank, Frank said... how should I put it... just make sure the water is on, ha.", "clean": "Set timer for 3 hours, turn off the TV first and turn on sprinkler system since it's 50% dry. Frank said just make sure the water is on.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25523.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 891, "source_record_id": "assigned_records.jsonl:25523"}
{"oral": "Hey, check flight status for Ian, Ian's flight. Also, check if... check if it's delayed by fifteen percent or more. And then find out who is Prime Minister of Canada, plus... see if switches are... wait, check if Ian's six months visa is still valid.", "clean": "Hey, check flight status for Ian's flight. Also, check if it's delayed by 15% or more. And then find out who is Prime Minister of Canada, plus check if Ian's 6 months visa is still valid.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25586.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 892, "source_record_id": "assigned_records.jsonl:25586"}
{"oral": "Honestly, uh, turn off all lights. Also, play some relaxing jazz in living room. Oh, and check if Sophia... wait, Sophia... no, Discord... wait wrong, is Duolingo sent me notification? I think... I mean... let me see... check if I spent one hundred ten dollars and fifty cents... no, one hundred twenty dollars and ten cents... wait wrong, is one hundred twenty dollars and fifty cents, huh. Just do it, ya. Just... just shut everything down, ba.", "clean": "Honestly, turn off all lights. Also, play some relaxing jazz in the living room. Oh, and check if Duolingo sent me a notification? Let me see... check if I spent $120.50. Just do it, ya. Just shut everything down, ba.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25498.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 890, "source_record_id": "assigned_records.jsonl:25498"}
{"oral": "Hey Alexa, play some eighties pop... play play some eighties pop, oh no, play a random song, random song from library, and set timer... set timer for ten minutes.", "clean": "Hey Alexa, play a random song from library, and set timer for 10 minutes.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25861.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 897, "source_record_id": "assigned_records.jsonl:25861"}
{"oral": "Hey, my brother said make it warmer in here, but uh, I think we just need to read weather forecast for weekend. Check score for the Lakers game. They got eighty two... oh no, eighty four points. Just turn on living room lights and lock front door, deadbolt it.", "clean": "Hey, my brother said make it warmer in here, but I think we just need to read the weather forecast for the weekend. Check the score for the Lakers game. They got 84 points. Just turn on the living room lights and lock the front door, deadbolt it.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25804.wav", "scene": "voice_search", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 896, "source_record_id": "assigned_records.jsonl:25804"}
{"oral": "Eh, play soft pop song, hmm, you know the one that sounds like... wait, actually, just search for the song I am humming now and then turn volume down. Plus, I want to know if sushi bar is open for one week, or like, ten million people go there, how say... just check it, ya.", "clean": "Play soft pop song, just search for the song I am humming now and then turn volume down. Plus, I want to know if the sushi bar is open for 1 week, or if 10 million people go there, just check it.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25641.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 895, "source_record_id": "assigned_records.jsonl:25641"}
{"oral": "Hey, tell me score for Son. Wait, s-s-son is playing... uh, one hundred yen... no, one hundred twenty... wait wrong, it is one hundred thirty five! This game... I missing because Microsoft Teams meeting. Who who who is president of Brazil? I forgot. Play latest episode of The Daily... and then, the score, please tell me!", "clean": "Hey, tell me the score for Son. Son is playing 135! I missed this game because of a Microsoft Teams meeting. Who is the president of Brazil? I forgot. Play the latest episode of The Daily... and then, the score, please tell me!", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25601.wav", "scene": "voice_search", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 893, "source_record_id": "assigned_records.jsonl:25601"}
{"oral": "Hey, uh, I'm not totally sure but I think Jessica... oh no, John, start a call with John, and then check weather for Saturday and Sunday, right?", "clean": "I'm not totally sure but I think start a call with John, and then check weather for Saturday and Sunday, right?", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl26002.wav", "scene": "voice_search", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 899, "source_record_id": "assigned_records.jsonl:26002"}
{"oral": "Set alarms every thir- thir- thirty minutes, also... wait, play my Workout playlist, and then... uh, check Binance in thir- thirty seconds.", "clean": "Set alarms every 30 minutes, play my Workout playlist, and check Binance in 30 seconds.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl26095.wav", "scene": "voice_search", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 903, "source_record_id": "assigned_records.jsonl:26095"}
{"oral": "Actually Quora says it is different, but what is current exchange rate for Yen? Set alarm for Sophia at six thirty, then one for Isabel at seven. Set thermostat to eco mode.", "clean": "Quora says it is different, but what is the current exchange rate for Yen? Set alarm for Sophia at 6:30, then one for Isabel at 7:00. Set thermostat to eco mode.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl26033.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 901, "source_record_id": "assigned_records.jsonl:26033"}
{"oral": "Actually, uh, can you find recipe for... recipe for chicken? Oh and just... just check what's on my schedule for Friday, and oh, turn off bedroom fan, yeah.", "clean": "Can you find a recipe for chicken? Check what's on my schedule for Friday, and turn off the bedroom fan.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl26094.wav", "scene": "voice_search", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 902, "source_record_id": "assigned_records.jsonl:26094"}
{"oral": "Hey, uh, resume playback for a sec. I'm not totally sure but I think my flight with Liam Airlines is at five fifteen P M. Could you check status? I just need to find parking spot nearby, maybe near Century Plaza? I mean, I think it's five fifteen P M, but let me double check. Just... uh, actually, just tell me if it's delayed, okay?", "clean": "Hey, resume playback for a sec. I'm not totally sure but I think my flight with Liam Airlines is at 5:15pm. Could you check status? I just need to find a parking spot nearby, maybe near Century Plaza? Just tell me if it's delayed, okay?", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl25999.wav", "scene": "voice_search", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 898, "source_record_id": "assigned_records.jsonl:25999"}
{"oral": "I been busy for ten minutes... ten minutes washing machine and two hours slow cooker. I need to call call my accountant... oh no, call my lawyer. Anyway, can you find this song? Hmm, it goes like... mmm hmmm mmm.", "clean": "I've been busy for 10 minutes washing machine and 2 hours slow cooker. I need to call my lawyer. Anyway, can you find this song? It goes mmm hmmm mmm.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl26032.wav", "scene": "voice_search", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 900, "source_record_id": "assigned_records.jsonl:26032"}
{"oral": "David said... David said call the Uber driver... no, call the dentist at two... two P M. Convert... convert ten days... wait... how many hours ten days is.", "clean": "David said call the dentist at 2pm. Convert how many hours 10 days is.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl26152.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 905, "source_record_id": "assigned_records.jsonl:26152"}
{"oral": "Actually, can you find that podcast episode from June twelfth... oh no, June fourteenth? And... check if... just tell me what's the status of my Amazon package, turn off oven, and also... uh... tell me what stock price of Apple is right now, then call my brother on speaker, thanks.", "clean": "Can you find that podcast episode from June 14? And tell me what's the status of my Amazon package, turn off the oven, and tell me what the stock price of Apple is right now, then call my brother on speaker, thanks.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl26116.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 904, "source_record_id": "assigned_records.jsonl:26116"}
{"oral": "Hey, tell me weather for this weekend, eh, I need know if I can... can take my wallet and go out. Also show me front porch camera and check security system, just to be sure. Oh, and what's on my calender for tomorrow? I think I orderd something from Amazon, check if it arrive on Saturday the twelfth.", "clean": "Hey, tell me the weather for this weekend, I need to know if I can take my wallet and go out. Also show me the front porch camera and check the security system, just to be sure. Oh, and what's on my calendar for tomorrow? I think I ordered something from Amazon, check if it arrives on Saturday the 12th.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl26199.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 906, "source_record_id": "assigned_records.jsonl:26199"}
{"oral": "明早我们在落家山...不对，珞珈山见，珞是王字旁一个各自的各，珈是王字旁一个加法的加。", "clean": "明早我们在珞珈山见。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2002.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 912, "source_record_id": "explanation.jsonl:2002"}
{"oral": "I need to calculate 15 percent of 85 dollars and check Postmates, oh and what's the temperature outside, also 180 centimeters and what's the definition of serendipity, send this to Mom.", "clean": "I need to calculate 15% of $85 and check Postmates, oh and what's the temperature outside, also 180cm and what's the definition of serendipity, send this to Mom.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl26482.wav", "scene": "voice_search", "language": "en", "voice_id": "en_female_nadia_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 909, "source_record_id": "assigned_records.jsonl:26482"}
{"oral": "Actually, uh can you find that podcast about wine cooler? No, wait. Play some eighty s pop hits first, then remind me to water plants at six P M. Oh, and set reminder for my anniversary. Just... check my WeChat for date, okay?", "clean": "Play some 80s pop hits first, then remind me to water plants at 6pm. Oh, and set a reminder for my anniversary. Just check my WeChat for the date, okay?", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl26254.wav", "scene": "voice_search", "language": "en", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 907, "source_record_id": "assigned_records.jsonl:26254"}
{"oral": "Honestly, can you... uh... show photos from my trip to Coursera... oh no, I mean Headspace? And uh... maybe... just... play the clock sound. Wendy told me to... wait... tell me a fun fact about octopuses. Can you do that in four seconds?", "clean": "Honestly, can you show photos from my trip to Headspace? And maybe just play the clock sound. Wendy told me to tell me a fun fact about octopuses. Can you do that in 4 seconds?", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl26624.wav", "scene": "voice_search", "language": "en", "voice_id": "en_male_russell_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 910, "source_record_id": "assigned_records.jsonl:26624"}
{"oral": "Hey, wait, I thought one cup was eight ounces but actually it depends on weight, huh. Anyway, how many grams in a cup of flour? Also, can you... can you play some smooth soul while I cook? And oh, set a reminder for January first at ten forty five P M to start a call with John about that Khan Academy course. I...I mean, just make sure it is on my calendar.", "clean": "I thought one cup was 8 ounces but actually it depends on weight, huh. Anyway, how many grams in a cup of flour? Also, can you play some smooth soul while I cook? And oh, set a reminder for January 1 at 10:45pm to start a call with John about that Khan Academy course. Just make sure it is on my calendar.", "usage_text_words": null, "audio": "audio/voice_search__assigned_recordsjsonl26420.wav", "scene": "voice_search", "language": "en", "voice_id": "en_female_skye_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 908, "source_record_id": "assigned_records.jsonl:26420"}
{"oral": "酒厂那个地址你写四川穷崃...不是，邛崃，邛是工人的工右边一个双耳旁，崃是山字旁一个来回的来。", "clean": "酒厂那个地址你写四川邛崃。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2003.wav", "scene": "explanation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 913, "source_record_id": "explanation.jsonl:2003"}
{"oral": "会议室留给乔溪...不对，荞溪，荞是荞麦的荞，还有纸安，不是，祉安，祉是福祉的祉。", "clean": "会议室留给荞溪和祉安。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2007.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 917, "source_record_id": "explanation.jsonl:2007"}
{"oral": "联系人一个叫陈燕...不对，岑砚，岑是岑寂的岑，砚是砚台的砚，另一个叫瑞宁，不是，芮宁，芮是草字头下面一个内外的内。", "clean": "联系人一个叫岑砚，另一个叫芮宁。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2005.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 915, "source_record_id": "explanation.jsonl:2005"}
{"oral": "花店订单写许棠...不对，栩棠，栩是栩栩如生的栩，棠是海棠的棠，再送给肖野...不是，骁野，骁是骁勇的骁。", "clean": "花店订单写栩棠，再送给骁野。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2006.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 916, "source_record_id": "explanation.jsonl:2006"}
{"oral": "家长会地点定在鲁湖书院...不对，麓湖书院，麓是山麓的麓。", "clean": "家长会地点定在麓湖书院。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2009.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 919, "source_record_id": "explanation.jsonl:2009"}
{"oral": "我今晚住盛州...不对，嵊州，嵊是山字旁一个乘车的乘，另外明天想点个交白炒肉，不是，茭白，茭是草字头下面一个交叉的交。", "clean": "我今晚住嵊州，明天想点个茭白炒肉。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2004.wav", "scene": "explanation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 914, "source_record_id": "explanation.jsonl:2004"}
{"oral": "纪念品帮我带一方吸砚...不是，歙砚，歙是歙县的歙，下午再去李园，不对，蠡园，蠡是范蠡的蠡。", "clean": "纪念品帮我带一方歙砚，下午再去蠡园。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2010.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 920, "source_record_id": "explanation.jsonl:2010"}
{"oral": "快递先寄到宜南...不对，沂南，沂是沂河的沂，然后地址在浪苑路，不是，阆苑路，阆是阆中古城的阆，苑是公园的苑。", "clean": "快递先寄到沂南，地址在阆苑路。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2008.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 918, "source_record_id": "explanation.jsonl:2008"}
{"oral": "路线先到番阳湖...不是，鄱阳湖，鄱是番茄的番右边一个双耳旁，然后再转车去批州，不对，邳州，邳是一个丕字右边加双耳旁。", "clean": "路线先到鄱阳湖，然后再转车去邳州。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2011.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 921, "source_record_id": "explanation.jsonl:2011"}
{"oral": "周六上午先去七霞寺...不对，栖霞寺，栖是栖息的栖，下午再爬牢山，不是，崂山，崂是山字旁一个劳累的劳。", "clean": "周六上午先去栖霞寺，下午再爬崂山。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2015.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 925, "source_record_id": "explanation.jsonl:2015"}
{"oral": "实验室名称写甬江实验室，甬是甬道的甬，回来的路上顺便买两个金云烧饼，不对，缙云烧饼，缙是绞丝旁一个晋升的晋。", "clean": "实验室名称写甬江实验室，回来的路上顺便买两个缙云烧饼。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2016.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 926, "source_record_id": "explanation.jsonl:2016"}
{"oral": "药先买氯雷他定，氯是气字头那个氯化钠的氯，不是绿色的绿，再带一盒普地蓝，不对，蒲地蓝，蒲是蒲公英的蒲。", "clean": "药先买氯雷他定，再带一盒蒲地蓝。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2013.wav", "scene": "explanation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 923, "source_record_id": "explanation.jsonl:2013"}
{"oral": "晚饭我要一盘同蒿，不是，茼蒿，茼是草字头下面一个同学的同，蒿是草字头下面一个高低的高，然后再来一碗合络面，不对，饸饹面，饸是食字旁一个合适的合，饹是食字旁一个各自的各。", "clean": "晚饭我要一盘茼蒿，再来一碗饸饹面。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2012.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 922, "source_record_id": "explanation.jsonl:2012"}
{"oral": "我想在举南看房，不是，莒南，莒是莒国的莒，旁边那个小区叫若溪公馆...不对，栎溪公馆，栎是栎树的栎。", "clean": "我想在莒南看房，旁边那个小区叫栎溪公馆。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2018.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 928, "source_record_id": "explanation.jsonl:2018"}
{"oral": "嘴上起泡那个药你给我拿阿夕洛韦...不对，阿昔洛韦，昔是今昔的昔，再备两包蒙脱石散，蒙是蒙古的蒙。", "clean": "嘴上起泡那个药你给我拿阿昔洛韦，再备两包蒙脱石散。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2014.wav", "scene": "explanation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 924, "source_record_id": "explanation.jsonl:2014"}
{"oral": "船票订到圣泗...不对，嵊泗，嵊是山字旁一个乘车的乘，泗是泗水的泗。", "clean": "船票订到嵊泗。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2017.wav", "scene": "explanation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 927, "source_record_id": "explanation.jsonl:2017"}
{"oral": "酒店就在鹿湖边上，不对，麓湖，麓是山麓的麓，对面是落家创谷...不是，珞珈创谷，珞是王字旁一个各自的各，珈是王字旁一个加法的加。", "clean": "酒店就在麓湖边上，对面是珞珈创谷。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2019.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 929, "source_record_id": "explanation.jsonl:2019"}
{"oral": "导航直接到许树关...不是，浒墅关，浒是水浒的浒，墅是别墅的墅。", "clean": "导航直接到浒墅关。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2023.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 933, "source_record_id": "explanation.jsonl:2023"}
{"oral": "快递寄到秦童古镇...不对，溱潼古镇，溱是三点水一个秦国的秦，潼是三点水一个儿童的童。", "clean": "快递寄到溱潼古镇。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2024.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 934, "source_record_id": "explanation.jsonl:2024"}
{"oral": "明天下午先去庸和宫，不对，雍和宫，雍是雍容的雍，然后晚上住在齐连路附近，不对，祁连路，祁是祁连山的祁。", "clean": "明天下午先去雍和宫，然后晚上住在祁连路附近。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2020.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 930, "source_record_id": "explanation.jsonl:2020"}
{"oral": "这趟先去京县...不对，泾县，泾是泾渭分明的泾，然后晚上住旌德，旌是旌旗的旌。", "clean": "这趟先去泾县，然后晚上住旌德。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2021.wav", "scene": "explanation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 931, "source_record_id": "explanation.jsonl:2021"}
{"oral": "水果再买点泥猴桃...不对，猕猴桃，猕是反犬旁一个弥补的弥，嗓子不舒服的话再带一瓶枇杷膏，枇是木字旁一个比方的比。", "clean": "水果再买点猕猴桃，嗓子不舒服的话再带一瓶枇杷膏。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2026.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 936, "source_record_id": "explanation.jsonl:2026"}
{"oral": "地铁坐到申庄...不对，莘庄，莘是莘莘学子的莘，出站后去这林路，不是，柘林路，柘是柘树的柘。", "clean": "地铁坐到莘庄，出站后去柘林路。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2022.wav", "scene": "explanation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 932, "source_record_id": "explanation.jsonl:2022"}
{"oral": "晚饭我要一份欠实糕...不对，芡实糕，芡是草字头下面一个欠条的欠，再来个扶苓饼，不是，茯苓饼，茯是草字头下面一个伏天的伏。", "clean": "晚饭我要一份芡实糕，再来个茯苓饼。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2025.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 935, "source_record_id": "explanation.jsonl:2025"}
{"oral": "体检地点在鞋和医院...不对，协和医院，协是协调的协，和是和平的和。", "clean": "体检地点在协和医院。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2030.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 940, "source_record_id": "explanation.jsonl:2030"}
{"oral": "周末先去凡净山...不对，梵净山，梵是梵高的梵，然后再去九兹古镇，不是，鸠兹古镇，鸠是斑鸠的鸠，兹是兹事体大的兹。", "clean": "周末先去梵净山，然后再去鸠兹古镇。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2028.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 938, "source_record_id": "explanation.jsonl:2028"}
{"oral": "酒水就买泸州老教...不对，泸州老窖，窖是地窖的窖，再带一瓶羊韶，不是，仰韶，仰是仰头的仰，韶是韶华的韶。", "clean": "酒水就买泸州老窖，再带一瓶仰韶。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2027.wav", "scene": "explanation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 937, "source_record_id": "explanation.jsonl:2027"}
{"oral": "见面地点在到香村门口...不对，稻香村，稻是水稻的稻，晚饭再去鬼街，不是，簋街，簋是竹字头下面一个皿字。", "clean": "见面地点在稻香村门口，晚饭再去簋街。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2034.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 944, "source_record_id": "explanation.jsonl:2034"}
{"oral": "门票定岳路书院...不对，岳麓书院，麓是山麓的麓，旁边那个景点叫橘子洲，橘是橘子的橘。", "clean": "门票定岳麓书院，旁边那个景点叫橘子洲。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2033.wav", "scene": "explanation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 943, "source_record_id": "explanation.jsonl:2033"}
{"oral": "零食顺手买两包恰恰瓜子...不对，洽洽瓜子，洽是三点水一个合适的合，再带一盒许福记，不是，徐福记，徐是双人旁一个余下的余。", "clean": "零食顺手买两包洽洽瓜子，再带一盒徐福记。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2035.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 945, "source_record_id": "explanation.jsonl:2035"}
{"oral": "特产带两盒扶茶...不对，茯茶，茯是草字头下面一个伏天的伏，还有正糕，不是，甑糕，甑是上面一个曾经的曾下面一个瓦。", "clean": "特产带两盒茯茶，还有甑糕。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2031.wav", "scene": "explanation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 941, "source_record_id": "explanation.jsonl:2031"}
{"oral": "家长群通知我发到飞书，飞是飞翔的飞，书是图书的书，视频再传哔哩哔哩，哔是口字旁一个毕，哩是口字旁一个里外的里。", "clean": "家长群通知我发到飞书，视频再传哔哩哔哩。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2036.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 946, "source_record_id": "explanation.jsonl:2036"}
{"oral": "买鞋就上德物...不对，得物，得是得到的得，资料搜索用夸克，夸是夸奖的夸，克是克服的克。", "clean": "买鞋就上得物，资料搜索用夸克。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2037.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 947, "source_record_id": "explanation.jsonl:2037"}
{"oral": "旧手机我还想看看魅族，魅是魅力的魅，新能源那边再问一下哪吒汽车，哪就是哪个的哪，吒是哪吒的吒。", "clean": "旧手机我还想看看魅族，新能源那边再问一下哪吒汽车。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2039.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 949, "source_record_id": "explanation.jsonl:2039"}
{"oral": "给孩子装小猿口算，小猿就是猿猴的猿，再顺便下一个作业帮，作是作业的作，帮是帮助的帮。", "clean": "给孩子装小猿口算，再顺便下一个作业帮。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2040.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_wennuanahu_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 950, "source_record_id": "explanation.jsonl:2040"}
{"oral": "试驾先约蔚来，蔚是蔚蓝的蔚，另外岚图那边也帮我问一下，岚是山风的岚。", "clean": "试驾先约蔚来，另外岚图那边也帮我问一下。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2038.wav", "scene": "explanation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 948, "source_record_id": "explanation.jsonl:2038"}
{"oral": "合同收件人是洛宁...不是，珞宁，珞是王字旁一个各自的各，另一个叫砚秋，砚是砚台的砚。", "clean": "合同收件人是珞宁，另一个叫砚秋。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2042.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 952, "source_record_id": "explanation.jsonl:2042"}
{"oral": "签到表上写云川...不对，昀川，昀是日字旁一个匀称的匀，旁边那个是沐岑，岑是岑寂的岑。", "clean": "签到表上写昀川，旁边那个是沐岑。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2041.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 951, "source_record_id": "explanation.jsonl:2041"}
{"oral": "下午来取件的是肖然...不对，骁然，骁是骁勇的骁，还有一个瑞禾，不是，芮禾，芮是草字头下面一个内外的内。", "clean": "下午来取件的是骁然，还有一个芮禾。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2043.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 953, "source_record_id": "explanation.jsonl:2043"}
{"oral": "作品集发给许安...不对，栩安，栩是栩栩如生的栩，再抄送乔宁，不是，荞宁，荞是荞麦的荞。", "clean": "作品集发给栩安，再抄送荞宁。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2044.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 954, "source_record_id": "explanation.jsonl:2044"}
{"oral": "晚上的包厢订在陈屿...不对，岑屿，岑是岑寂的岑，最后到的人叫棠溪，棠是海棠的棠。", "clean": "晚上的包厢订在岑屿名下，最后到的人叫棠溪。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2045.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 955, "source_record_id": "explanation.jsonl:2045"}
{"oral": "讲义先给仪川...不是，沂川，沂是沂河的沂，另外纸宁，不对，祉宁，祉是福祉的祉，也要一份。", "clean": "讲义先给沂川，另外祉宁也要一份。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2046.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_vv_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 956, "source_record_id": "explanation.jsonl:2046"}
{"oral": "接站牌写穷音...不是，邛音，邛是工人的工右边一个双耳旁，同行的瑞岚，不对，芮岚，芮是草字头下面一个内外的内。", "clean": "接站牌写邛音，同行的是芮岚。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2048.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_female_xiaohe_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 958, "source_record_id": "explanation.jsonl:2048"}
{"oral": "预约名单里一个叫浪宁...不对，阆宁，阆是阆中古城的阆，另一个叫砚棠，砚是砚台的砚，棠是海棠的棠。", "clean": "预约名单里一个叫阆宁，另一个叫砚棠。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2049.wav", "scene": "explanation", "language": "zh", "voice_id": "zh_male_shaonianzixin_uranus_bigtts", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 959, "source_record_id": "explanation.jsonl:2049"}
{"oral": "民宿登记鹿遥...不对，麓遥，麓是山麓的麓，同行的是加禾，不是，珈禾，珈是王字旁一个加法的加。", "clean": "民宿登记麓遥，同行的是珈禾。", "usage_text_words": null, "audio": "audio/explanation__explanationjsonl2047.wav", "scene": "explanation", "language": "zh", "voice_id": "ICL_uranus_zh_male_yangyang_tob", "voice_role": "", "reference_audio": "", "reference_transcript": "", "source_file": "data/benchv2.jsonl", "source_line": 957, "source_record_id": "explanation.jsonl:2047"}