Coverage Aware Active Evaluation for Failure Discovery with Paired Systems

Anjali Parashar; Rachel Luo; Apoorva Sharma; Sushant Veer; Edward Schmerling; Carson Sobolewski; Mingxin Yu; Chuchu Fan

cs.AIAug 13, 2026

Coverage Aware Active Evaluation for Failure Discovery with Paired Systems

Authors: Anjali Parashar, Rachel Luo, Apoorva Sharma, Sushant Veer, Edward Schmerling, Carson Sobolewski, Mingxin Yu, Chuchu Fan,

Organizations: Laboratory of Information & Decision Systems (LIDS), MIT, USA · NVIDIA Research, USA

Abstract

Autonomous systems can fail in rare and heterogeneous ways, making real-world failure discovery difficult under limited testing budgets. Although cheaper proxies such as simulators, lower-fidelity systems, or related policies can be sampled extensively to find failures, proxy failures often do not transfer to the real world due to sim-to-real and system-to-system gaps. The key challenge is therefore to effectively leverage proxy system information for accurate prediction of severe target system failures. We propose an adaptive failure discovery method that combines proxy evaluations with limited target system results to guide scenario selection for target system testing. Our method learns a local predictor of target risk by correcting proxy failure signals using control-variate-inspired residual modeling. To find failures that are both likely and diverse, we combine this predictor with a support-aware mutual-information objective that favors realistic, well-supported regions while expanding coverage across failure modes. Across autonomous driving, manipulation, and quadruped velocity-tracking tasks, our method discovers up to 2 $\times$ as many failures as random sampling and active-learning baselines, including severe and diverse failures missed by competing methods.

Figures & tables

Figure 1: Overview of approach. Our method uses support-aware Mutual Information ( Equation 3 ) to subsample $K_{MI}$ scenarios $x$ , which are evaluated on the cheap proxy system to get metrics $y_{s}$ , and a surrogate model $q_{\theta}$ trained using data evaluated on target system. We subsequently optimize $\mu_{CV}(x)$ ( Equation 2 ) over these sub sampled candidates to select the next batch of evaluation points as likely failure scenarios. The process is repeated until evaluation budget $B$ is exhausted.

Task	$\gamma$	Random	BNN-C	BAMS	BNN-CV	Our
nuPlan	0.3	0.23 $\pm$ 0.01	0.25 $\pm$ 0.01	0.28 $\pm$ 0.01	0.27 $\pm$ 0.01	0.33 $\pm$ 0.01
	0.2	0.22 $\pm$ 0.00	0.24 $\pm$ 0.02	0.27 $\pm$ 0.01	0.26 $\pm$ 0.01	0.32 $\pm$ 0.01
	0.05	0.21 $\pm$ 0.01	0.21 $\pm$ 0.02	0.25 $\pm$ 0.01	0.22 $\pm$ 0.02	0.29 $\pm$ 0.01

Table 1: Positive Samples for different severity levels for nuPlan (mean

\pm

std across four seeds).

Figure 2: MI driven exploration for different forms of data acquisition on nuPlan task . 2D T-SNE embeddings of scenario dataset showing high and low MI regions for 200 datapoints sampled from (1) small concentrated region of low failures, (2) MC sampling, (3) data recovered from a specific type of scenarios. The plots show that our support aware MI strategy successfully identifies unobserved regions agnostic to the choice of sampling strategy.

Figure 4: Baseline comparison for Quadruped task, (avg $\pm$ std, two seeds) for $B=30$ samples, with $10$ initial random samples for training surrogate, and 20 samples acquired in batches of $b=5$ .

$\gamma$	0.7 (Our)	0.8 (Our)	1.0 (Our)	0.7 (BAMS)	0.8 (BAMS)	1.0 (BAMS)
P.S.	0.6 $\pm$ 0.1	0.3 $\pm$ 0.1	0.2 $\pm$ 0.1	0.5 $\pm$ 0.1	0.3 $\pm$ 0.2	0.0 $\pm$ 0.0

Table 2: Positive samples for different

\gamma

values in Quadruped task.

	Full	Greedy	$\beta=0(n+k)$	$\beta{=}0$	MI	$-$ MI-C	$-$ MI-G
Utility	0.38	0.38	0.36	0.36	0.32	0.4	0.41
P.S.-0.5	0.24	0.22	0.14	0.17	0.11	0.30	0.32
P.S.-0.7	0.03	0.01	0.02	0.01	0.02	0.02	0.01
Cov.	0.78	0.58	0.79	0.72	0.91	0.6	0.48

Table 3: KITTI component ablation

Panel of Figure 5 — Figure 5: Examples of grasping failures for SIMPLER task. Top- scenarios for which both proxy and target systems succeed (left) and fail (right) to grasp the can. Bottom- two examples for which the proxy system succeeds and target system fails to grasp the can, discovered by our approach.

Appendix figures & tables10 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 6: Overview of approach. Our method uses support-aware Mutual Information ( Equation 3 ) to subsample $K_{MI}$ scenarios $x$ , which are evaluated on the cheap proxy system to get metrics $y_{s}$ , and a surrogate model $q_{\theta}$ trained using data evaluated on target system. We subsequently optimize $\mu_{CV}(x)$ ( Equation 2 )over these sub sampled candidates to select the next batch of evaluation points as likely failure scenarios. The process is repeated until evaluation budget $B$ is exhausted. Figure on the left shows the surrogate model learnt after 40 samples of evaluation for the target system shown.

Figure 7: 1D synthetic task. Left-right: (1) Randomly sampled 10 scenarios and observed values of synthetic target, shown against ground truth target and proxy system. (2) Scenarios identified as high MI due to being outside support/novel candidates ( Section 4.2 ). (3) $\hat{\mu}_{CV}(x)$ estimated using different fixed local regions and adaptive local region selection plotted against ground truth target. Adaptively selecting local region shows minimum deviation from ground truth.

Setting	Avg. Cumulative TTC	Cumulative Coverage	Positive Samples
$n_{\text{init}}=5$	$1.09\pm 0.06$	$0.58\pm 0.04$	$0.27\pm 0.02$
$n_{\text{init}}=10$	$1.16\pm 0.13$	$0.48\pm 0.09$	$0.25\pm 0.01$
$n_{\text{init}}=20$	$1.06\pm 0.04$	$0.75\pm 0.02$	$0.26\pm 0.03$
$n_{\text{init}}=50$	$0.99\pm 0.06$	$1.10\pm 0.13$	$0.30\pm 0.05$

Appendix

Table 4: Ablation over different initialization sizes on the nuPlan task, 100 samples, four seeds, avg

\pm

std reported.

Setting	Avg. Cumulative TTC	Cumulative Coverage	Positive Samples
$\gamma=0.3$
nominal	$1.04\pm 0.0332$	$0.98\pm 0.06$	$0.32\pm 0.01$
$\beta=0$	$1.03\pm 0.05$	$0.95\pm 0.05$	$0.35\pm 0.01$
$\gamma=0.0$
nominal	$1.04\pm 0.03$	$0.87\pm 0.05$	$0.29\pm 0.01$
$\beta=0$	$1.03\pm 0.05$	$0.81\pm 0.08$	$0.31\pm 0.01$

Appendix

Table 5: Ablation comparing our approach with and without sim samples by setting

\beta=0

on the nuPlan task. Results reported for 20 initial samples,

B=150

, avg and std for four seeds.

Figure 8: Scatterplot of failure scenarios

x=(v_{x},v_{y},w_{z})

(

y\geq 0.7

) for Quadruped task, generated by our method (red), and BAMS (blue). Our method shows higher diversity of scenarios unlike BAMS that tends to concentrate around failures discovered in early iterations.

Figure 9: Heatmaps showing learnt surrogate model predictions projected over scenario space for Quadruped task. Projection value corresponds to median of third variable reported in parentheses. Scattered points correspond to initial 10 and subsequent 20 scenarios collected by our approach.

Figure 10: Baseline comparison for SIMPLER task, showing avg

\pm

std across three seeds for

B=70

samples, with

10

initial random samples for training all surrogate models, and 60 additional samples acquired in batches of

b=3

Panel of Figure 11 — Figure 11: Failures for Quadruped task. Quadrupeds shown at the end of tracking duration, failure scenarios correspond to Quadruped reaching near one of the boxed and stopping due to obstacle avoidance.

Panel of Figure 12 — Figure 12: More explains of failure scenarios uncovered by our approach in the SIMPLER task.

Task	Architecture	Input	Hidden layers	Activation
nuPlan surrogate	MLP	384	$96\to 24\to 6$	ELU + LN
nuPlan BNN-C head	+ sigmoid	—	—	$\sigma$
SIMPLER BNN	MLP	19	$96\to 24\to 6$	Tanh
SIMPLER classifier for GP-C	MLP + sigmoid	19	$96\to 24\to 6$	$\sigma$
Quadruped MLP	MLP	3	$8\to 8$	ReLU

Appendix

Table 6: Summary of neural network architectures across tasks.

Robot Failure Detection

Robot failure detection identifies when autonomous systems are failing, predicts emerging faults, and supports diagnosis or recovery.

See all 46 papers

Importance Sampling and PCA for Finding Failures in Commercial Autonomous Vehicles
Jul 20, 2026Hailey Warner, Duncan Eddy, Shreya Parjan +6Agent Failure Analysis Autonomous Driving Safety Evaluation
Department of Aeronautics and Astronautics, Stanford University, Palo Alto, CA · Torc Robotics, Blacksburg, VA
Sufficient and Necessary Causal Explanations for the Analysis of Autonomous Systems
Oct 21, 2026Tom Yaacov, Khen Elimelech, Hana Chockler +1Causal Explanation
King’s College London · Rice University
Risk-Aware Adaptive Evaluation: Finding High-Impact Failures Under Limited Budgets
Sep 30, 2026Priyanath Maji, Spandan Ghose ChowdhuryThompson Sampling AI Agent Evaluation
Georgia Institute of Technology Atlanta, GA 30332

Jul 20, 2026J/K move · Enter open · S save

Importance Sampling and PCA for Finding Failures in Commercial Autonomous Vehicles

Hailey Warner, Duncan Eddy, Shreya Parjan, Caroline Cahilly+5

Department of Aeronautics and Astronautics, Stanford University, Palo Alto, CA · Torc Robotics, Blacksburg, VA

Agent Failure Analysis Autonomous Driving Safety Evaluation Autonomous Driving Simulation

Open paper AskPDF arXiv

Methods for discovering rare failures in autonomous systems have so far been demonstrated almost exclusively in simulations with simple, academic driving stacks, leaving open whether they generalize to the more robust planners used in commercial systems. We address this gap by applying two rare-event discovery algorithms to a commercial autonomous trucking stack. Adaptive stress testing (AST) uses reinforcement learning to search for the most likely noise trajectories leading to a simulated collision, while diffusion-based failure sampling (DiFS) trains a denoising diffusion model to sample a diverse set of failures. We show that both algorithms find simulated collisions during merge and cut-in maneuvers where traditional Monte Carlo simulation does not. To make these failures actionable, we introduce a statistical analysis based on principal component analysis (PCA) that classifies failures into common modes and identifies the timesteps that most influence the outcome. We cluster the principal components and invert the PCA transform to recover generalized noise trajectories, and show that these trajectories reproduce failures in identical and similar scenarios. This provides a path from failure discovery to systematic diagnosis of perception-level flaws.

Abstract

Robot Failure Detection

Explore similar work

Importance Sampling and PCA for Finding Failures in Commercial Autonomous Vehicles

Sufficient and Necessary Causal Explanations for the Analysis of Autonomous Systems

Risk-Aware Adaptive Evaluation: Finding High-Impact Failures Under Limited Budgets