damru-knowledge · Dataset Card
damru-knowledge: Dataset Card
Written by Damru-AI, published under apache-2.0, revision 1b5c0371ff63, read 2026-09-29. Shown as written; SAVRN's own facts about this dataset are on its page.
A continuously growing, self-collected question-answer knowledge base that powers Damru AI — a self-learning assistant built for exam preparation and general-purpose help, with a focus on Indian students.
The dataset is harvested and quality-filtered automatically, 24x7, from multiple open sources and a self-evaluating reasoning engine. New rows are appended every hour as parquet shards under data/.
What's inside
| Column | Type | Description |
|---|---|---|
question |
string | A question, problem, or prompt |
answer |
string | A detailed, self-checked answer / solution |
intent |
string | Topic / category (e.g. math, coding, analysis, general) |
lang |
string | Language code (en, hi, ...) |
upvotes |
int | Internal quality score (higher = better) |
created_at |
string | Timestamp the row was generated |
Sources
- Wikipedia — encyclopedic facts across 80+ subjects
- arXiv — research-paper concepts and summaries
- StackExchange — real Q&A across many communities
- Math engine — symbolically-verified maths (SymPy) + olympiad / JEE / MIT-level problems solved and self-checked
- Coding engine — problems generated, solved, and executed to verify the solution actually runs (tests must pass)
- Analysis engine — deep reasoning and critical-analysis Q&A
How it's built
Damru runs a multi-worker learning loop that: 1. Picks a subject and masters it before moving to the next (curriculum-driven). 2. Generates Q&A, then self-evaluates each item; only high-quality rows are kept. 3. Raises its own quality bar over time (rising acceptance threshold). 4. Verifies maths symbolically and runs code to confirm correctness. 5. Deduplicates and stores rows, then syncs them here hourly.
Intended use
- Supervised fine-tuning / instruction-tuning of small open models (e.g. Qwen2.5-3B) for Damru AI.
- Retrieval-augmented generation (RAG) knowledge source.
- Educational Q&A research.
Limitations
- Content is machine-generated and machine-verified; occasional errors are possible.
- Coverage grows over time and may be uneven across subjects early on.
- Not a substitute for authoritative textbooks or professional advice.
Status
This is a living dataset — it grows continuously toward a multi-million-row target. Row counts and file sizes update as new hourly shards are added.
License
Released under the Apache-2.0 license. Source-derived content remains under its original terms; please respect upstream source licenses where applicable.