NVIDIA Spatial Intelligence Lab (SIL) NVIDIA Research

Test-Time Coverage: Test-Conditioned Data Curation for Deployment-Aware Learning

*Equal contribution  1NVIDIA  2University of Ottawa
TTCov pipeline for extracting L-APs, constructing Atlas and K-Atlas, and selecting training data for test coverage.

TTCov (Test-Time Coverage) moves test-time conditioning from online weight updates to offline data curation. It builds an Atlas and K-Atlas from open knowledge and unlabeled deployment videos, then selects training clips for test-relevant coverage.

Key Results


Label-Free
Matches human-curated driving data without manual labels.
Coverage
Curates toward deployment concepts instead of training-side proxies.
Adaptability
Expands to new cities without re-curation or forgetting.

Abstract


Deployed AI systems are often trained from broad candidate data pools, necessitating data curation towards the deployment test distribution. However, standard data curation methods score training-side criteria rather than directly optimizing deployment match. We introduce TTCov (Test-Time Coverage), a data-level test-conditioned curation method that uses test-side information before training instead of updating model weights at inference. TTCov decomposes deployment-conditioned curation into coverage and distribution. To represent coverage, it builds a task Atlas, a collection of LLM-based atomic propositions (L-APs) describing deployment-relevant concepts, seeded from open task knowledge and expanded with unmatched L-APs extracted from unlabeled deployment samples. To represent distribution, it instantiates the matched deployment L-APs with their frequencies, yielding a Knowledge Atlas (K-Atlas) that operationalizes the deployment distribution as a curation target. TTCov then selects a budgeted training set whose deployment L-AP distribution approximates this target. We apply TTCov towards autonomous driving (AD), keeping adaptation off the inference path while selecting data with greater deployment-relevant coverage, closer K-Atlas matching, and stronger downstream end-to-end driving performance than data-curation baselines, including seamless adaptability to novel domains via city-to-city expansion.

From Deployment Videos to Targeted Training Data


Represent

Videos become atomic propositions

TTCov avoids collapsing a clip into one caption or embedding. Each video becomes an unordered L-AP set describing scenes, agents, actions, rules, relations, and conditions. These L-APs can then be matched, counted, and composed across the dataset.

Target

A frequency-weighted task map

TTCov combines open driving knowledge with deployment-video L-APs to build the Atlas. Counting matched Atlas concepts across the deployment set yields the K-Atlas, a target distribution over what the training data should cover.

TTCov overview showing L-AP extraction, Atlas expansion, and Atlas L-AP unification.
Steer

Select clips against the target

Candidate clips are mapped into the same L-AP vocabulary. Under a budget, TTCov greedily selects clips that reduce the gap between the selected-set distribution and the deployment K-Atlas. The L-AP map also exposes coverage gaps and supports traversal by scenario, agent, action, and condition.

selected subset ≈ deployment K-Atlas

Results


TTCov is evaluated on autonomous-driving curation for Navhard, using OpenScene as the unlabeled pool and Latent Transfuser in NAVSIM for EPDMS. Across budgets, TTCov improves K-Atlas matching and outperforms random, Coreset, and SSE baselines.

End-to-end driving performance

Navhard EPDMS across different data budgets, reported relative to the manually curated Navtrain oracle.
Method 0.5x 0.75x 1x 1.25x 1.5x
Navtrain (oracle) -- -- 24.70 | 1.00 -- --
Random 18.95 | 0.77 18.76 | 0.76 20.15 | 0.82 21.95 | 0.89 22.95 | 0.93
Coreset 20.77 | 0.84 21.48 | 0.87 23.63 | 0.96 24.55 | 0.99 25.89 | 1.05
SSE 18.44 | 0.75 22.29 | 0.90 23.45 | 0.95 24.77 | 1.00 26.03 | 1.05
TTCov (ours) 20.62 | 0.83 23.00 | 0.93 24.42 | 0.99 25.49 | 1.03 26.40 | 1.07

Coverage after adding a city

Histogram comparing selected training points by distance to test points for City 1.

When deployment expands to a new city, TTCov covers 54.4% of its target test regions versus 20.4% for Coreset.

K-Atlas distribution match

KL divergence to the K-Atlas distribution across data budgets.

TTCov stays closest to the target K-Atlas across budgets.

Selected data examples

Visual examples of TTCov-selected data covering scenarios, agents, and behaviors.

Caption-level retrieval can mix a requested x + y + z with unrelated factors. Atomic L-APs retrieve and compose the axes that matter, such as scenario, action, and agent.

BibTeX


@misc{chang2026testtimecoverage,
  title = {Test-Time Coverage: Test-Conditioned Data Curation for Deployment-Aware Learning},
  author = {Chang, Nadine and Shen, Maying and Diao, Shizhe and Wang, Jialiang and Chen, Jingde and Breuel, Thomas and Molchanov, Pavlo and Mahmood, Rafid and Alvarez, Jose M.},
  year = {2026},
}