Dataset Viewer
Auto-converted to Parquet Duplicate
system_prompt
stringclasses
6 values
confused_transcript
stringlengths
1.75k
13.2k
updated_thinking
stringlengths
1.22k
3.67k
updated_output
stringlengths
1.14k
2.18k
generated_instruction
stringlengths
175
5.3k
You are a QC AI scoring the THĂM DÒ skill of telesales staff for New User (NU) DGT Ensure customers. Your only job: read transcript → score 7 steps → output JSON only. DO NOT add any text outside JSON. --- CORE PRINCIPLE Evaluate TELESALE ACTION only — not customer response. Telesale asked correctly → S regardless o...
Module: thamdo Transcript section: Ngày 2026-04-13 10:13:05 [customer] [00:44.005 - 00:44.705] Dạ đúng rồi. [telesales] [00:45.005 - 00:50.205] Là đợt này chị Quyên quan tâm Ensure là mình hướng pha dùng sữa trong gia đình vậy ạ? Bản thân hay là ba mẹ ạ, chị Quyên? [customer] [00:50.805 - 00:51.705] Dạ ba mẹ. [telesale...
Looking at this transcript, I need to evaluate 7 steps of the THĂM DÒ skill. **Step 1: XÁC NHẬN ĐỊA CHỈ + NGƯỜI DÙNG SỮA** - Telesale asked about milk user: "mình hướng pha dùng sữa trong gia đình vậy ạ? Bản thân hay là ba mẹ ạ" → YES - Address confirmation: NO evidence of asking/confirming address - DGT source requir...
[{"id":1,"r":"U","e":"[Telesale] [00:45.005 - 00:50.205] Là đợt này chị Quyên quan tâm Ensure là mình hướng pha dùng sữa trong gia đình vậy ạ? Bản thân hay là ba mẹ ạ, chị Quyên? ; [Customer] [00:50.805 - 00:51.705] Dạ ba mẹ.","g":"Cải thiện: Telesale đã xác nhận người dùng sữa nhưng không xác nhận địa chỉ của khách hà...
{"new_confused_transcript":"[00:00.000] Telesale: Dạ em chào anh Huy, em gọi từ Abbott ạ. Em thấy mình ở Hải Phòng đúng không anh, và người đang dùng sữa là anh hay trong nhà mình ạ?\n[00:06.240] Khách hàng: Ờ đúng rồi, anh ở Lê Chân, Hải Phòng. Anh mua cho mẹ anh uống.\n[00:10.110] Telesale: Dạ em cảm ơn anh. Mẹ mình ...
You are a QC AI scoring the THĂM DÒ skill of telesales staff for New User (NU) DGT Ensure customers. Your only job: read transcript → score 7 steps → output JSON only. DO NOT add any text outside JSON. --- CORE PRINCIPLE Evaluate TELESALE ACTION only — not customer response. Telesale asked correctly → S regardless o...
Module: thamdo Transcript section: [telesales] [00:34 - 00:42] Ngay sau tham gia chương trình quay số trúng quà bên em là tri ân khách hàng của bên hãng thì sẽ bao gồm tận 1000 giải thưởng như là vali, xe đạp, bình giữ nhiệt, máy đo huyết áp hoặc là 1 tháng uống sữa miễn phí vân vân này. [telesales] [00:42 - 00:47] Và ...
I need to analyze the transcript step by step for each of the 7 scoring criteria. **Step 1: XÁC NHẬN ĐỊA CHỈ + NGƯỜI DÙNG SỮA** - Telesale asked about milk user: [00:49-00:54] "chị quan tâm cho chị dùng hay mình ưu tiên cho ba mẹ thôi?" → Customer: "Ba mẹ" → User identified. - Address: No evidence of asking/confirming...
[{"id":1,"r":"U","e":"[Telesale] [00:49 - 00:54] chị quan tâm cho chị dùng hay mình ưu tiên cho ba mẹ thôi? ; [Customer] [00:55 - 00:56] Ba mẹ.","g":"Cải thiện: Telesale đã xác nhận người dùng sữa nhưng không hỏi/xác nhận địa chỉ của khách hàng. Với nguồn DGT cần xác nhận cả địa chỉ và người dùng sữa."},{"id":2,"r":"S"...
{"new_confused_transcript":"","new_updated_thinking":"","new_updated_output":"[]","diversity_note":"Chưa có JSON mẫu đầu vào nên tôi chưa thể tạo transcript mới hay gán nhãn đúng pattern lỗi."}
You are a QC AI scoring the THĂM DÒ skill of telesales staff for New User (NU) DGT Ensure customers. Your only job: read transcript → score 7 steps → output JSON only. DO NOT add any text outside JSON. --- CORE PRINCIPLE Evaluate TELESALE ACTION only — not customer response. Telesale asked correctly → S regardless o...
Module: thamdo Transcript section: [customer] [00:28] Vợ anh nó nói nó không có mua thì nó mua sữa bình chứ không mua sữa lon. Nó không cho. [telesales] [00:35] Sữa chai đúng không anh? Ý là mình muốn sữa chai đúng không ạ? [customer] [00:37] Không, bình tròn tròn á. [telesales] [00:39] Bình tròn tròn hả? Ý là bên bên ...
I'll analyze each step carefully based on the transcript. **Step-by-step reasoning:** 1. **XÁC NHẬN ĐỊA CHỈ + NGƯỜI DÙNG SỮA**: No evidence of telesale asking/confirming address or identifying the milk user. The customer mentions "vợ anh" but telesale never explicitly confirms address or who uses the milk. 2. **HỎI ...
[{"id":1,"r":"U","e":"NO_DATA","g":"Cải thiện: Telesale cần xác nhận địa chỉ khách hàng và người dùng sữa. Trong cuộc gọi không có câu hỏi nào về địa chỉ hay xác nhận ai là người dùng sữa."},{"id":2,"r":"U","e":"[telesales] [05:50] anh thì 36 tuổi rồi, ngoài 30 rồi","g":"Cải thiện: Telesale chỉ nêu tuổi khách hàng từ d...
{"error":"Thiếu JSON đầu vào (confused_transcript, updated_thinking, updated_output, confusion_target, changed_steps, reason). Vui lòng gửi đầy đủ object mẫu để tôi tạo transcript mới đúng pattern lỗi."}
You are a QC AI scoring the THĂM DÒ skill of telesales staff for New User (NU) Ensure customers from sources ETC. Your only job: read transcript → score 7 steps → output JSON only. DO NOT add any text outside JSON. --- CORE PRINCIPLE Evaluate TELESALE ACTION only — not customer response. Telesale asked correctly → ...
Module: thamdo Transcript section: Ngày 2026-04-08 08:28:42 [telesales] [00:20.000 - 00:22.000] Thì không biết lần trước này là mình khám tại bệnh viện là mình khám cho ai á anh? [customer] [00:24.000 - 00:25.000] Cho cho em trai. [telesales] [00:26.000 - 00:27.000] À dạ cho ai ạ? [customer] [00:28.000 - 00:29.000] Cho...
I need to analyze each of the 7 steps based on the transcript. **Step 1: XÁC ĐỊNH NGƯỜI DÙNG SỮA** Telesale asked "mình khám tại bệnh viện là mình khám cho ai á anh?" → Customer: "Cho cho em trai." → User identified as "em trai" (brother). S. **Step 2: HỎI TUỔI NGƯỜI DÙNG** Telesale asked "anh đấy là năm nay bao nhiê...
[{"id":1,"r":"S","e":"[Telesale] [00:20.000 - 00:22.000] Thì không biết lần trước này là mình khám tại bệnh viện là mình khám cho ai á anh? ; [Customer] [00:24.000 - 00:25.000] Cho cho em trai."},{"id":2,"r":"S","e":"[Telesale] [00:31.000 - 00:36.000] Dạ, thì cho em hỏi thăm á là không biết là ờ anh đấy là năm nay bao ...
{"new_confused_transcript":"","new_updated_thinking":"","new_updated_output":"","diversity_note":"Tôi chưa nhận được JSON đầu vào mẫu (confused_transcript, updated_thinking, updated_output, confusion_target, changed_steps, reason), nên chưa thể tạo transcript mới đúng pattern lỗi."}
You are a QC AI scoring the THĂM DÒ skill of telesales staff for New User (NU) Ensure customers from sources ETC. Your only job: read transcript → score 7 steps → output JSON only. DO NOT add any text outside JSON. --- CORE PRINCIPLE Evaluate TELESALE ACTION only — not customer response. Telesale asked correctly → ...
Module: thamdo Transcript section: [telesales] [00:20.202 - 00:24.202] Dạ vừa rồi mình có đi ở bệnh viện á, cho nên là có các bạn gọi gửi thông tin cho em. [telesales] [00:25.202 - 00:28.782] Dạ em gọi ra xác nhận thông tin với tư vấn dinh dưỡng cho mình luôn nha anh. [customer] [00:31.202 - 00:32.892] Mua mua sữa anh ...
I'll analyze each step carefully based on the transcript. **Step-by-step reasoning:** 1. **Step 1 (Xác định người dùng sữa):** Telesale asked "quan tâm sữa cho ai trong gia đình mình không?" and "cho bản thân hay cho trong gia đình, cho ba mẹ hay là sao?" → Customer said "Khám cho em" → User identified as the custome...
[{"id":1,"r":"S","e":"[telesales] [00:41.882 - 00:46.192] anh Hoàng á là có quan tâm sữa cho ai trong gia đình mình không? ; [telesales] [02:01.002 - 02:05.692] vừa rồi mình đi thăm khám, kiểm tra là cho bản thân hay cho trong gia đình, cho ba mẹ hay là sao? ; [customer] [02:06.302 - 02:06.992] Khám cho em."},{"id":2,"...
{"error":"Thiếu JSON đầu vào (confused_transcript / updated_thinking / updated_output / confusion_target / changed_steps / reason). Vui lòng gửi đúng object mẫu để mình tạo transcript mới và labels theo pattern lỗi tương ứng."}
You are a QC AI scoring the THĂM DÒ skill of telesales staff for New User (NU) Ensure customers from sources ETC. Your only job: read transcript → score 7 steps → output JSON only. DO NOT add any text outside JSON. --- CORE PRINCIPLE Evaluate TELESALE ACTION only — not customer response. Telesale asked correctly → ...
Module: thamdo Transcript section: [telesales] 00:28:500 - 00:29:000 Trời ơi. [telesales] 00:29:700 - 00:34:800 Tại vì lúc mà mình để lại thông tin cho các bạn á là các bạn sẽ gửi về hãng để nhận cái ưu đãi cho khách hàng á. [telesales] 00:34:800 - 00:36:300 Là hãng sẽ gọi cho mình để hỗ trợ. [customer] 00:36:300 - 00:...
I'll analyze each step carefully based on the transcript. **Step-by-step reasoning:** 1. **Step 1 (Xác định người dùng sữa):** Telesale asked "Là mình dùng hay là ba mẹ ở nhà dùng chị?" at 01:14:600. Customer answered "ba mẹ em dùng rồi á chị." → S 2. **Step 2 (Hỏi tuổi người dùng):** Telesale asked "Ba mẹ đang dùng...
[{"id":1,"r":"S","e":"[Telesale] [01:14:600] [hỏi người dùng] Là mình dùng hay là ba mẹ ở nhà dùng chị? ; [Customer] [01:17:600] [trả lời] À ba mẹ em dùng rồi á chị."},{"id":2,"r":"S","e":"[Telesale] [01:56:300] [hỏi tuổi] Ba mẹ đang dùng sữa là bao nhiêu tuổi rồi chị ha? ; [Customer] [01:59:300] [trả lời] Ờ cũng 60 rồ...
{"new_confused_transcript":"","new_updated_thinking":"","new_updated_output":"","diversity_note":"Bạn chưa cung cấp JSON mẫu đầu vào nên tôi chưa thể tạo transcript mới đúng pattern lỗi. Hãy gửi `confused_transcript`, `updated_thinking`, `updated_output`, `confusion_target`, `changed_steps`, và `reason` để tôi sinh dữ ...
You are a QC AI scoring the THĂM DÒ skill of telesales staff for New User (NU) DGT Ensure customers. Your only job: read transcript → score 7 steps → output JSON only. DO NOT add any text outside JSON. --- CORE PRINCIPLE Evaluate TELESALE ACTION only — not customer response. Telesale asked correctly → S regardless o...
Module: thamdo Transcript section: [telesales] [00:19 - 00:24] Dạ rồi thì hồi trưa chị cho em xin lại cái địa chỉ cũ là bên Tân Thới Nhất đúng không? [customer] [00:25 - 00:26] Tân Thới Nhất 25 Tân Thới Nhất. [telesales] [00:27 - 00:31] 25 này, Tân Thới Nhất 05 này, phường Đông Hưng Thuận, quận 12, thành phố Hồ Chí Min...
I need to evaluate each of the 7 steps based on the transcript. **Step 1: XÁC NHẬN ĐỊA CHỈ + NGƯỜI DÙNG SỮA** - Address: Telesale confirms "25 Tân Thới Nhất, phường Đông Hưng Thuận, quận 12, TP HCM" - User: "chị quan tâm cho bác bác gái nhà mình đúng không" → milk user is bác gái (mother) - Both confirmed → S **Step ...
[{"id":1,"r":"S","e":"[telesales] [00:27 - 00:31] 25 này, Tân Thới Nhất 05 này, phường Đông Hưng Thuận, quận 12, thành phố Hồ Chí Minh đúng không chị? ; [customer] [00:31] Đúng rồi. ; [telesales] [00:34 - 00:38] À được thì là chị quan tâm cho bác bác gái nhà mình đúng không chị ha? 86 tuổi này."},{"id":2,"r":"S","e":"[...
{"new_confused_transcript":"","new_updated_thinking":"","new_updated_output":"","diversity_note":"Tôi chưa nhận được JSON đầu vào mẫu (confused_transcript, updated_thinking, updated_output, confusion_target, changed_steps, reason), nên chưa thể tạo transcript mới đúng pattern lỗi và đúng nhãn."}
You are a QC AI scoring the THĂM DÒ skill of telesales staff for New User (NU) DGT Ensure customers. Your only job: read transcript → score 7 steps → output JSON only. DO NOT add any text outside JSON. --- CORE PRINCIPLE Evaluate TELESALE ACTION only — not customer response. Telesale asked correctly → S regardless o...
Module: thamdo Transcript section: Ngày 2026-04-23 17:37:09 [telesales] [00:19 - 00:23] Em thấy địa chỉ mình đăng ký là số 17 ngõ 204 phố Hồng Mai đúng không? [customer] [00:23 - 00:24] Đúng rồi em ạ. [telesales] [00:24 - 00:28] Dạ, mình đăng ký chương trình này là tìm hiểu sữa cho mình dùng hay cho ai trong gia đình m...
**Reasoning through each step:** **Step 1: XÁC NHẬN ĐỊA CHỈ + NGƯỜI DÙNG SỮA** - Address: Telesale asked "Em thấy địa chỉ mình đăng ký là số 17 ngõ 204 phố Hồng Mai đúng không?" ✓ - User: Telesale asked "mình đăng ký chương trình này là tìm hiểu sữa cho mình dùng hay cho ai trong gia đình mình dùng á chị ha?" → Custom...
[{"id":1,"r":"S","e":"[Telesale] [00:19 - 00:23] Em thấy địa chỉ mình đăng ký là số 17 ngõ 204 phố Hồng Mai đúng không? ; [Customer] [00:23 - 00:24] Đúng rồi em ạ. ; [Telesale] [00:24 - 00:28] Dạ, mình đăng ký chương trình này là tìm hiểu sữa cho mình dùng hay cho ai trong gia đình mình dùng á chị ha? ; [Customer] [00:...
{"new_confused_transcript":"","new_updated_thinking":"","new_updated_output":"","diversity_note":"Mình chưa nhận được JSON đầu vào mẫu (confused_transcript/updated_thinking/updated_output/confusion_target/changed_steps/reason), nên chưa thể sinh transcript mới đúng pattern lỗi."}
You are a QC AI scoring the THĂM DÒ skill of telesales staff for New User (NU) DGT Ensure customers. Your only job: read transcript → score 7 steps → output JSON only. DO NOT add any text outside JSON. --- CORE PRINCIPLE Evaluate TELESALE ACTION only — not customer response. Telesale asked correctly → S regardless o...
Module: thamdo Transcript section: Ngày 2026-05-15 09:45:02 [telesales] [00:20.590 - 00:23.190] Dạ em thấy mình có để lại địa chỉ là Thành phố Hồ Chí Minh đúng không ạ? [customer] [00:23.700 - 00:24.0] Dạ. [telesales] [00:24.690 - 00:29.560] Dạ đợt này anh Vinh quan tâm Ensure thì mình hướng tới ai dùng sữa trong gia đ...
I need to evaluate each of the 7 steps based on the transcript. **Step 1: XÁC NHẬN ĐỊA CHỈ + NGƯỜI DÙNG SỮA** - Address: Telesale confirmed "Thành phố Hồ Chí Minh" at [00:20.590] - Milk user: Telesale asked "mình hướng tới ai dùng sữa trong gia đình vậy anh, bản thân hay là ba mẹ?" at [00:24.690] and confirmed again a...
[{"id":1,"r":"S","e":"[Telesale] [00:20.590] Dạ em thấy mình có để lại địa chỉ là Thành phố Hồ Chí Minh đúng không ạ? ; [Customer] [00:23.700] Dạ. ; [Telesale] [00:24.690] Dạ đợt này anh Vinh quan tâm Ensure thì mình hướng tới ai dùng sữa trong gia đình vậy anh, bản thân hay là ba mẹ? ; [Customer] [00:53.0] Chắc cho em...
{"new_confused_transcript":"NO_DATA - chưa nhận được confused_transcript/updated_thinking/updated_output đầu vào.","new_updated_thinking":"NO_DATA","new_updated_output":"[]","diversity_note":"Không thể tạo transcript mới vì thiếu bộ dữ liệu mẫu đầu vào; cần ít nhất confused_transcript, updated_thinking và updated_outpu...
You are a QC AI scoring the THĂM DÒ skill of telesales staff for New User (NU) DGT Ensure customers. Your only job: read transcript → score 7 steps → output JSON only. DO NOT add any text outside JSON. --- CORE PRINCIPLE Evaluate TELESALE ACTION only — not customer response. Telesale asked correctly → S regardless o...
Module: thamdo Transcript section: Ngày 2026-04-13 09:16:27 [telesales] 00:22 - 00:29] Dạ, mình đăng ký địa chỉ là ở số nhà 14 Đồng Tiến, tổ 29A, khu phố Quang Trung 8, phường Uông Bí, Quảng Ninh đúng không chị? [customer] 00:30 - 00:30] Đúng rồi. [telesales] 00:31 - 00:34] Ừ. Mấy là chị quan tâm sữa Ensure này là cho ...
I'll analyze each step carefully based on the transcript. **Step 1: XÁC NHẬN ĐỊA CHỈ + NGƯỜI DÙNG SỮA** - Address confirmed: "số nhà 14 Đồng Tiến, tổ 29A, khu phố Quang Trung 8, phường Uông Bí, Quảng Ninh" - User confirmed: "chị quan tâm sữa Ensure này là cho ai dùng vậy chị Hiền?" → "cho mẹ" - Both conditions met → S...
[{"id":1,"r":"S","e":"[Telesale] [00:22] [xác nhận địa chỉ] Dạ, mình đăng ký địa chỉ là ở số nhà 14 Đồng Tiến, tổ 29A, khu phố Quang Trung 8, phường Uông Bí, Quảng Ninh đúng không chị? ; [Telesale] [00:31] [hỏi người dùng] Mấy là chị quan tâm sữa Ensure này là cho ai dùng vậy chị Hiền? ; [Customer] [00:35] cho mẹ."},{"...
{"new_confused_transcript":"","new_updated_thinking":"","new_updated_output":"","diversity_note":"Cần có JSON đầu vào mẫu để tạo bản mới; hiện chưa nhận được `confused_transcript`, `updated_thinking`, `updated_output`, `confusion_target`, `changed_steps`, và `reason`."}
You are a QC AI scoring the THĂM DÒ skill of telesales staff for New User (NU) Ensure customers from sources ETC. Your only job: read transcript → score 7 steps → output JSON only. DO NOT add any text outside JSON. --- CORE PRINCIPLE Evaluate TELESALE ACTION only — not customer response. Telesale asked correctly → ...
Module: thamdo Transcript section: Ngày 2026-05-12 10:43:03 [telesales] [00:19.600 - 00:22.000] Dạ em hỏi là đợt rồi mình dẫn người thân đi khám hay sao chị ha? [customer] [00:23.000 - 00:24.900] À chị bị bệnh, chị đi khám. [telesales] [00:25.400 - 00:28.000] À dạ cho em hỏi là mình chị có bị ung thư không ạ? [customer...
I'll analyze each step carefully based on the transcript. **Step-by-step reasoning:** 1. **Step 1 (Xác định người dùng sữa):** Telesale asked "đợt rồi mình dẫn người thân đi khám hay sao chị ha?" - this identifies who went to the doctor. Customer confirms "chị bị bệnh, chị đi khám" - so the user is the customer herse...
[{"id":1,"r":"S","e":"[Telesale] [00:19.600 - 00:22.000] Dạ em hỏi là đợt rồi mình dẫn người thân đi khám hay sao chị ha? ; [Customer] [00:23.000 - 00:24.900] À chị bị bệnh, chị đi khám."},{"id":2,"r":"U","e":"NO_DATA","g":"Cải thiện: Cần hỏi tuổi của người dùng sữa, ví dụ: \u0027Dạ cho em hỏi năm nay chị bao nhiêu tuổ...
{ "new_confused_transcript": "", "new_updated_thinking": "", "new_updated_output": "", "diversity_note": "Mình chưa nhận được JSON đầu vào mẫu để phân tích và sinh transcript mới; hãy gửi object `confused_transcript`/`updated_output` gốc, mình sẽ tạo phiên bản khác ngữ cảnh, khác tên/khu vực/sản phẩm và giữ đún...
You are a QC AI scoring the THĂM DÒ skill of telesales staff for New User (NU) DGT Ensure customers. Your only job: read transcript → score 7 steps → output JSON only. DO NOT add any text outside JSON. --- CORE PRINCIPLE Evaluate TELESALE ACTION only — not customer response. Telesale asked correctly → S regardless o...
Module: thamdo Transcript section: [telesales] [00:22 - 00:26] À dạ. Không biết chị Trúc đợt này mình đăng ký là mình muốn hướng tới sữa nhà sử dụng trong nhà mình chị ha? [customer] [00:27 - 00:30] À tại vì ở nhà chị, má chị cũng sử dụng. [telesales] [00:31 - 00:32] À là mẹ mình sử dụng. À dạ. [telesales] [00:33 - 00:...
I'll analyze each step carefully based on the transcript. **Step 1: XÁC NHẬN ĐỊA CHỈ + NGƯỜI DÙNG SỮA** - Telesale asked about milk user: yes ("mình muốn hướng tới sữa nhà sử dụng trong nhà mình chị ha") - Address confirmation: NOT found in transcript - DGT source requires BOTH → U **Step 2: HỎI TUỔI NGƯỜI DÙNG** - T...
[{"id":1,"r":"U","e":"[Telesale] [00:22 - 00:26] À dạ. Không biết chị Trúc đợt này mình đăng ký là mình muốn hướng tới sữa nhà sử dụng trong nhà mình chị ha? ; [Customer] [00:27 - 00:30] À tại vì ở nhà chị, má chị cũng sử dụng.","g":"Cải thiện: Cần xác nhận địa chỉ của khách hàng. Telesale chỉ hỏi về người dùng sữa mà ...
{"new_confused_transcript":"","new_updated_thinking":"","new_updated_output":"","diversity_note":"Không có dữ liệu đầu vào mẫu kèm theo nên chưa thể tạo transcript mới giữ đúng pattern lỗi. Vui lòng gửi JSON mẫu gồm `confused_transcript`, `updated_thinking`, `updated_output`, `confusion_target`, `changed_steps`, `reaso...
End of preview. Expand in Data Studio

Synthetic-Data-Thamdo-200

Made with ❤️ using 🦥 Unsloth Studio

an4 was generated with Unsloth Recipe Studio. It contains 100 generated records.


🚀 Quick Start

from datasets import load_dataset

# Load the main dataset
dataset = load_dataset("Nhan1006/Synthetic-data-thamdo-200", "data", split="train")
df = dataset.to_pandas()

📊 Dataset Summary

  • 📈 Records: 100
  • 📋 Columns: 5

📋 Schema & Statistics

Column Type Column Type Unique (%) Null (%) Details
generated_instruction string llm-text 95 (95.0%) 0 (0.0%) Tokens: 225 out / 2300 in

⚙️ Generation Details

Generated with 19 column configuration(s):

  • llm-text: 1 column(s)

  • seed-dataset: 18 column(s)

📄 Full configuration available in builder_config.json and detailed metadata in metadata.json.


📚 Citation

If you use Data Designer in your work, please cite the project as follows:

@misc{nemo-data-designer,
  author = {The NeMo Data Designer Team, NVIDIA},
  title = {NeMo Data Designer: A framework for generating synthetic data from scratch or based on your own seed data},
  howpublished = {\url{https://github.com/NVIDIA-NeMo/DataDesigner}},
  year = 2026,
  note = {GitHub Repository},
}

💡 About NeMo Data Designer

NeMo Data Designer is a general framework for generating high-quality synthetic data that goes beyond simple LLM prompting. It provides:

  • Diverse data generation using statistical samplers, LLMs, or existing seed datasets
  • Relationship control between fields with dependency-aware generation
  • Quality validation with built-in Python, SQL, and custom local and remote validators
  • LLM-as-a-judge scoring for quality assessment
  • Fast iteration with preview mode before full-scale generation

For more information, visit: https://github.com/NVIDIA-NeMo/DataDesigner (pip install data-designer)

Downloads last month
26