ScenarioBench: Trace-Grounded Compliance Evaluation for Text-to-SQL and RAG
Bibliographic record
Abstract
ScenarioBench is a policy-grounded, trace-aware benchmark for evaluating Text-to-SQL and retrieval-augmented generation in compliance settings. Each YAML scenario ships with a no-peek gold-standard package-expected decision, minimal witness trace, governing clause set, and canonical SQL-enabling end-to-end scoring of both what a system decides and why it decides so. Systems must justify outputs with clause IDs retrieved from the same policy canon, making explanations falsifiable and audit-ready. The evaluator reports core signals-decision/trace quality, retrieval and SQL correctness (by result-set equivalence), latency, and a hallucination rate. A normalized Scenario Difficulty Index (SDI) and its budgeted variant (SDI-R) aggregate these while pricing retrieval difficulty and time. Unlike Spider/BIRD or KILT/RAG-style setups, ScenarioBench enforces clause-ID grounding under a strict nopeek rule. On a seed synthetic suite ($N=16$), accuracy saturates (Acc$=\mathbf{M}-\mathbf{F} 1=1.000)$, and a single budgeted reflection closes trace gaps (TrC & Coverage$0.541 \rightarrow 1.000$) with zero hallucination at$\sim+1 \text{ms}$-showing that marginal gains come from why-quality under explicit time budgets. Artifact & Code (v0.1.0): https://github.com/ShabnamAtf/ScenarioBench/releases/tag/v0.1.0
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.010 | 0.054 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.002 | 0.004 |
| Open science | 0.004 | 0.003 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.008 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".