Stanford's agent won Databricks' inaugural Grounded Reasoning Cup with 63.3% accuracy on OfficeQA Pro V2, a benchmark built from 120,000 pages of U.S. Treasury financial documents. Out-of-the-box frontier agents averaged below 30% on the same benchmark. System design beat model choice.

The competition pitted 11 academic teams from the U.S. and Canada against each other under deliberately adversarial conditions. Each team partnered with OpenAI, Anthropic, or Google DeepMind and was locked to that provider's model family. Teams spent two months optimizing against OfficeQA, Databricks' grounded-reasoning benchmark. On competition day they received a new corpus: the U.S. Treasury's Accounts of Receipts and Expenditures, released 36 hours prior. The short window forced teams to ingest and index the data without overfitting.

Six 15-minute rounds, 15 questions each. Difficulty escalated; final-round questions doubled in weight. The first correct answer per question earned a 0.25-point speed bonus, measuring both accuracy and latency.

DimensionDetail
Rounds6
Questions per round15
Total questions90
Final-round weight2× (double points)
Speed bonus+0.25 pts — first correct answer per question
DifficultyEscalated across rounds
FIG. 02 Grounded Reasoning Cup competition format — Databricks Grounded Reasoning Cup

Stanford won. UMass Amherst and Yale took second and third. All three finished 20+ points clear of the field. Teams using the same underlying model differed by an average of 30.4 points, isolating system design as the primary lever. Top teams shared four tactics: document preprocessing before retrieval, parallel agent execution, structured tool use with explicit verification, and targeted fallbacks when primary document-representation methods failed. Stanford packaged these into a skill library with adaptive answer verification.

TacticWhat it doesStanford addition
Document preprocessing before retrievalCleans and structures pages before chunking/indexing
Parallel agent executionMultiple agent lanes run concurrently to improve throughput
Structured tool use + explicit verificationEach tool call is followed by a formal check stepAdaptive answer verification via skill library
Targeted fallbacksAlternative document-representation method fires when primary fails
FIG. 03 System-design tactics shared by top-finishing teams — Databricks Grounded Reasoning Cup

Across all 90 competition questions, 18.8% went unsolved by every team. These represent systematic retrieval and parsing gaps no architecture closed. Stanford's winning system, which outperformed baseline frontier agents by 35 points, still answered only 63.3% correctly. Enterprise grounded reasoning over complex financial corpora remains unsolved.

OfficeQA Pro V2 accuracy: Stanford vs. out-of-the-box frontier agents vs. unsolved-by-all-teams share
FIG. 04 OfficeQA Pro V2 accuracy: Stanford vs. out-of-the-box frontier agents vs. unsolved-by-all-teams share — Databricks Grounded Reasoning Cup. Baseline derived: Stanford (63.3%) outperformed baseline by 35 pts → baseline ≈ 28.3%, consistent with reported "below 30%".

Benchmark-driven agent development surfaces an old problem: techniques tuned to a known corpus don't generalize reliably. Approaches strong on OfficeQA degraded on OfficeQA Pro V2's different document structure and vocabulary. The held-out, same-day corpus made this failure visible. Static benchmarks don't force this test.

For production RAG pipelines and document-QA systems, Grounded Reasoning Cup results suggest a concrete checklist. Retrieval fallbacks — not single strategies — are structural. Parallel agent lanes with separate verification improve throughput and confidence. Latency is observable and carries cost: the speed bonus models production queuing. Model swaps within the same provider shift scores by points; system-level changes move them by tens.

Stanford's winning agent pipeline: from raw corpus to verified answer
FIG. 05 Stanford's winning agent pipeline: from raw corpus to verified answer — Databricks Grounded Reasoning Cup — Stanford system design

The OfficeQA benchmark suite is public. Practitioners building enterprise document agents now have a reproducible test for the failure modes the Grounded Reasoning Cup exposed.