Select a task
Model mode uses the server-side HF token and auto-solves all steps in
the episode.
Select a task and start a new episode to begin interpreting Sanskrit manuscripts
How to use
- Glossary Anchoring — Choose the correct domain-specific meaning of a Sanskrit technical term (e.g. agni means different things in Ayurveda vs. philosophy).
- Sandhi Resolution — Identify the correct phonological split of a compound word into its component parts.
- Samāsa Classification — Classify a Sanskrit compound word into one of six Pāṇinian types: Tatpurusha, Karmadharaya, Dvigu, Dvandva, Bahuvrihi, or Avyayibhava.
- Referential Coherence — Track pronouns and implicit subjects across a multi-verse passage (multiple steps per episode).
- Manuscript Restoration (Task 5) — Use tools to gather evidence and restore a passage. Note: Tools are only available via the API/model. In the UI, play without tools to test your pure knowledge!
- Full Manuscript Session (Task 6) — A 5-phase long-horizon task chaining all challenges on a single passage.
Choose an interpretation
Episode complete
0.00
—
Session stats
0
Episodes
0
Correct
0.00
Best score
Model performance
Run model episodes to see overall and task-wise model score
summaries.