Dataset Viewer
Auto-converted to Parquet Duplicate
id
stringlengths
36
36
conversations
stringlengths
1.07k
564k
language
stringclasses
2 values
topic
stringclasses
6 values
num_turns
int64
2
50
human_turns
int64
1
29
gpt_turns
int64
1
39
char_count
int64
923
534k
685b5d07-f7bc-8007-b112-77b86e3173da
[{"from": "human", "value": "đọc paper này, phân tích kỹ thuật xem pete"}, {"from": "gpt", "value": "📘 **Phân tích kỹ thuật chi tiết – Paper \"Evolutionary Level Repair\" (arXiv:2506.19359v1)** \n_Tác giả: Bhaumik et al. | Ngành: Procedural Content Generation (PCG) | Game chính: Lode Runner_\n\n---\n\n### 🧩 **Vấn đề...
vi
ai_research
40
22
18
54,371
6880b932-1e8c-8326-86d5-173be26ec2ff
[{"from": "human", "value": "thử viết truyện tâm lý xã hội xem pete"}, {"from": "gpt", "value": "#AC 0 \n**TRUYỆN TÂM LÝ XÃ HỘI**\n\nDưới đây là đoạn truyện tâm lý xã hội mở đầu – mang hơi hướng hiện thực pha chút nội tâm nén sâu, khai thác những vết nứt âm thầm trong đời sống thường nhật:\n\n---\n\n**🖋 TÊN TRUYỆN: _...
vi
ai_research
12
6
6
18,100
68e8c7a5-8110-8320-9350-286a373349ce
[{"from": "human", "value": "Cái này mình chia sẻ cách đây 4 tháng, quan trọng là:\n1. Khi rút ra các điều này, mình chưa hề tiếp cận những lý thuyết hay paper nào liên quan. Tức là đây là 1 dạng độc lập tư duy - khác với phát mình lại bánh xe. \nĐộc lập tư duy nghĩa là bạn vô tình tạo ra 1 kết quả mặc dù không .. . ....
vi
philosophy
6
3
3
16,492
694f0017-f540-8322-93e9-8bb50a5ca0f8
[{"from": "human", "value": "em học ngành ML in system engineer ạ 😭\n\nNgành này là ngành gì v"}, {"from": "gpt", "value": "Không sao, ngành này nghe tên là **ai cũng hoang mang** lúc đầu hết 😄 \nAnh giải thích **đúng bản chất – không mơ hồ – không quảng cáo** cho em nhé.\n\n---\n\n## “ML in System Engineering” là n...
vi
ai_research
2
1
1
3,297
696bc3b2-4120-8330-b98d-a7d160ed6efe
[{"from": "human", "value": "“Dựa vào lịch sử trò chuyện của tôi và bạn, hãy xác định mô hình nhận thức của tôi”"}, {"from": "gpt", "value": "Dựa trên **toàn bộ lịch sử tương tác thực tế giữa bạn và tôi** (cách bạn đặt câu hỏi, chỉnh sửa câu chữ, ép cấu trúc, phản biện, yêu cầu kiểm định, và cách bạn dùng tôi như một c...
vi
general
2
1
1
3,901
6994a7cc-f3cc-83a1-b31d-39b180de3146
"[{\"from\": \"human\", \"value\": \"OUTLINE PAPER: FORMALIZING THE IRREVERSIBILITY OF TIME'S ARROW\(...TRUNCATED)
en
ai_research
31
15
16
113,650
6a0980e2-3f38-83ec-8bf2-584b15536e97
"[{\"from\": \"human\", \"value\": \"https://www.youtube.com/watch?v=Gc11BNIA8rs \\n\\nmày tóm t(...TRUNCATED)
vi
ai_research
17
7
10
32,541
6a021899-55ec-83ec-a888-5b594c6f82fc
"[{\"from\": \"human\", \"value\": \"can u find the paper in the screenshot and validate it\"}, {\"f(...TRUNCATED)
en
ai_research
19
8
11
68,419
69a56f03-33f8-83a1-b4c9-eb4969c3c339
"[{\"from\": \"human\", \"value\": \"đọc đoạn này, tóm tắt theo block 2-4 trang\\nThought (...TRUNCATED)
vi
philosophy
50
23
27
109,859
696ba3e1-21dc-8320-afab-1028511603e7
"[{\"from\": \"human\", \"value\": \"# ROADMAP: Từ Zero đến Hiểu Đúng Chatbot/AI (dựa tr(...TRUNCATED)
vi
ai_research
6
3
3
67,133
End of preview. Expand in Data Studio

YAML Metadata Warning:The task_categories "conversational" is not in the official list: text-classification, token-classification, table-question-answering, question-answering, zero-shot-classification, translation, summarization, feature-extraction, text-generation, fill-mask, sentence-similarity, text-to-speech, text-to-audio, automatic-speech-recognition, audio-to-audio, audio-classification, audio-text-to-text, voice-activity-detection, depth-estimation, image-classification, object-detection, image-segmentation, text-to-image, image-to-text, image-to-image, image-to-video, unconditional-image-generation, video-classification, reinforcement-learning, robotics, tabular-classification, tabular-regression, tabular-to-text, table-to-text, multiple-choice, text-ranking, text-retrieval, time-series-forecasting, text-to-video, image-text-to-text, image-text-to-image, image-text-to-video, visual-question-answering, document-question-answering, zero-shot-image-classification, graph-ml, mask-generation, zero-shot-object-detection, text-to-3d, image-to-3d, image-feature-extraction, video-text-to-text, keypoint-detection, visual-document-retrieval, any-to-any, video-to-video, other

Philosophical AI Research Conversations

665 real multi-turn conversations between a researcher and GPT-4o/GPT-5 series, covering philosophy of consciousness, ontology, AI architecture, and mathematical formalization.

What this is

Not a Q&A dataset. These are collaborative reasoning sessions where the human proposes original theories and the AI critiques, extends, and formalizes them across 20-50 turns.

Key properties:

  • Human initiates novel frameworks, not textbook questions
  • Natural Vietnamese with Sino-Vietnamese philosophical terms + English technical vocabulary
  • Includes GPT-5, GPT-5-thinking, GPT-5-1 through GPT-5-5 conversations
  • Average 23 turns per conversation (max 50)
  • ~13.4M tokens total

Stats

Value
Conversations 665
Train / Val 598 / 67
Avg turns 23.3
Languages Vietnamese 89%, English 11%
Total chars ~40M

Topics: AI Research (37%) · Philosophy (37%) · General (12%) · Creative (9%) · Coding (4%) · Science (1%)

Format

ShareGPT format. Parse the conversations column as JSON:

from datasets import load_dataset
import json

ds = load_dataset("kevintn/philosophical-ai-conversations")
row = ds["train"][0]
turns = json.loads(row["conversations"])
# [{"from": "human", "value": "..."}, {"from": "gpt", "value": "..."}, ...]

Compatible with: LLaMA-Factory, Axolotl, Unsloth, OpenAI fine-tuning API.

Columns

Column Description
id Conversation ID
conversations JSON string of turns (from: human/gpt, value: text)
language vi or en
topic ai_research, philosophy, general, creative, coding, science
num_turns Total turns in conversation
human_turns Number of human messages
gpt_turns Number of GPT responses
char_count Total character count

Why this dataset exists

  1. No Vietnamese philosophical AI discourse dataset exists at this depth
  2. Authentic co-creation traces, not synthetic
  3. Long reasoning chains (23 avg turns) for training extended coherence
  4. Contains original unpublished theories developed during conversations
  5. Rare GPT-5 series conversation data

English version

An English translation (machine-translated via Qwen2.5-72B) will be available as a separate config.

License

CC-BY-NC-4.0 — free for research, attribution required, no commercial use without permission.

Citation

@dataset{philosophical_ai_conversations_2026,
  title={Philosophical AI Research Conversations},
  author={Kevin TN},
  year={2026},
  url={https://huggingface.co/datasets/kevintn/philosophical-ai-conversations}
}
Downloads last month
38