Reward Atlas

Jump to

SQL-Detective-v1LLM / Agentic · A multi-turn agent answers analytics questions by querying a read-only SQLite sandbox. Tool use, verifiable rewards, 12-turn budget.
All field guidesEsc to close
Sign in
LLM / AgenticPublicPublishedv1.0Advanced

SQL-Detective-v1

A multi-turn agent answers analytics questions by querying a read-only SQLite sandbox. Tool use, verifiable rewards, 12-turn budget. Maintained by Bhaskar & Jaswanth · updated 28 Aug 2026.

Dataset
2,400 q · 12 DBs
Action
tool call | final answer
Turn cap
12
Best baseline
GRPO-7B 61.4%
12

Known issues & versions

Known issues

  • SELECT * dumps can earn partial credit on narrow tables (see Step 5).
  • Four eval questions have ambiguous date boundaries; they are listed in data/eval_ambiguous.json and excluded from reported numbers.
Guide details
Version
Type
LLM / Agentic
API
OpenEnv
License
Apache-2.0
Seeds
3
Domains
tool-usedatacoding
Install
pip install reward-atlas-sql-detective

Issue with this step? Suggest an edit